โŒ

Reading view

There are new articles available, click to refresh the page.

FedRAMP is getting faster, new automation and pilots promise approvals in months, not years

Interview transcript

Terry Gerton Weโ€™re going to talk about one of everybodyโ€™s favorite topics, FedRAMP. Itโ€™s been around for years, but agencies are still struggling to get modern tools. So from your perspective, why is the process so hard for software and service companies to get through?

Irina Denisenkoย  Itโ€™s a great question. Why is it so hard to get through FedRAMP? It is so hard to get through FedRAMP because at the end of the day, what is FedRAMP really here to do? Itโ€™s here to secure cloud software, to secure government data sitting in cloud software. You have to remember this all came together almost 15 years ago, which if you remember 15 years ago, 20 years ago, was kind of early days of all of us interacting with the internet. And we were still even, in some cases, scared to enter our credit card details onto an online website. Fast forward to today, we pay with our face when we get on our phone. Weโ€™ve come a long way. But the reality is cloud security hasnโ€™t always been the, of course, itโ€™s secure. In fact, it has been the opposite. Of course, its unsecure and itโ€™s the internet and thatโ€™s where you go to lose all your data and all your information. And so long story short, you have to understand thatโ€™s were the government is coming from. We need to lock everything down in order to make sure that whether itโ€™s VA patient data, IRS data on our taxpayers, obviously anything in the DoW, any sort of information data there, all of that stays secure. And so thatโ€™s why there are hundreds of controls that are applied to cloud environments in order make sure and double sure and triple sure that that data is secure.

Terry Gerton You lived the challenge first-hand with your own company. What most surprised you about the certification process when you tackled it yourself? What most surprise me?

Irina Denisenkoย  When we tackled FedRAMP ourselves for the first time was that even if you have the resources and specifically if you $3 million to spend, you know, $3 million burning a hole in your pocket doesnโ€™t happen often, but even if have that and you have staff on the U.S. Soil and you have the willingness to invest all of that for a three-year process to get certified, that is still not enough. What you need on top of that is an agency to say yes to sponsoring you. And when they say yes, to sponsoring you what they are saying yes to you is to take on your cyber risk. And specifically what theyโ€™re saying yes to is to spend half a million dollars of taxpayer money of agency budget, typically using contractors, to do an initial security review of your application. And then to basically get married to you and do something called continuous monitoring, which is a monthly meeting that theyโ€™re going to have with you forever. They, that agency is going to be your accountability partner and ultimately the risk bearer of you, the software provider, to make sure you are burning down all of the vulnerabilities, all of these CVEs, every finding in your cloud environment on the timeline that youโ€™re supposed to do that. And that ends up costing an agency about $250,000 a year, again, in the form of contractors, tooling, etc. That was the most surprising to me, that again, even as a cloud service provider, whoโ€™s already doing business with JP Morgan and Chase, you know, healthcare systems, you name it, even thatโ€™s not enough, you need an agency sponsor, because at the end of the day, itโ€™s the agencyโ€™s data and they have to protect it. And so they have do that triple assurance of, yes, you said youโ€™re doing the security stuff, but let us confirm that youโ€™re doing the the security stuff. That was the most surprising to me. And why, really, ultimately, we started Knox Systems, because what we do at Knox is we enable the inheritance model. So we are doing all of that with our sponsoring agencies, of which we have 15. Knox runs the largest FedRAMP managed cloud. And what that means is we host the production environment of our customers inside of our FedRAMP environment across AWS, Azure, and GCP. And our customers inherit our sponsors. So they inherit the authorization from the treasury, from the VA, from the Marines, etc., Which means that the Marines, the Treasury, the VA, didnโ€™t have to spend an extra half a million upfront and $250k ongoing with every new application that was authorized. They are able to get huge bang for their buck by just investing that authorization, that sponsorship into the Knox boundary. And then Knox does the work and the hard work to ensure the security and ongoing authorization and compliance of all of the applications that we bring into our environment.

Terry Gerton Iโ€™m speaking with Irina Denisenko. Sheโ€™s the CEO of Knox Systems. So it sounds like you found a way through the maze that was shorter, simpler, less expensive. Is FedRAMP 20X helping to normalize that kind of approach? How do you see it playing out?

Irina Denisenkoย  Great question. FedRAMP 20X is a phenomenal initiative coming out of OMB-GSA. And really the crux of that is all about machine-readable and continuous authorization. Today, when I talked about continuous monitoring, thatโ€™s a monthly meeting that happens. And I kid you not, we, as a cloud service provider, again, we secure Adobeโ€™s environment and many others, we come with a spreadsheet, an actual spreadsheet that has all of the vulnerabilities listed from all the scans weโ€™ve done over the last month, and anything that is still open from anything prior months. And we review that spreadsheet, that actual Excel document, and then after the meet with our agencies and then, after that meeting, we upload that spreadsheet into a system called USDA on the FedCiv side, eMass, DOW side, DISA side. And then they, on their side, download that spreadsheet and they put it into other systems. And I mean, thatโ€™s the process. I think no one is confused, or no one would argue that surely thereโ€™s a better way. And a better would be a machine readable way, whether thatโ€™s over an API, using a standard language like OSCAL. Thereโ€™s lots of ways to standardize, but it doesnโ€™t have to be basically the equivalent of a clipboard and a pencil. And thatโ€™s what FedRAMP 20X is doing. Itโ€™s automating that information flow so that not only is it bringing down the amount of just human labor that needs to be done to do all this tracking, but more importantly, this is cloud security. Just because youโ€™re secure one second doesnโ€™t mean youโ€™re secure five seconds from now, right? You need to be actively monitoring this, actively reporting this. And if itโ€™s taking you 30 days to let an agency know that you have a critical vulnerability, thatโ€™s crazy. You, you got to tell them in, you know, five minutes after you find out or, you know to put a respectable buffer, a responsible buffer to allow you to mitigate remediate before you notify more parties, maybe itโ€™s a four day buffer but itโ€™s certainly not 30 days. Thatโ€™s what FedRAMP20X is doing. Weโ€™re super excited about it. We are very supportive of it and have been actively involved in phase I and all subsequent phases.

Terry Gerton Right, so phase II is scheduled to start shortly in 2026. What are you expecting to see as a result?

Irina Denisenkoย  Well, phase I was all about FedRAMP low, phase II is all about FedRAMP moderate. And we expect that, you know, itโ€™s going to really โ€” FedRAMP moderate is realistically where most cloud service offerings sit, FedRAMP moderate and high. And so thatโ€™s really the one that the FedRAMP needs to get right. What we expect to see and hope to see is to have agencies actually authorized off of these new frameworks. The key is really going to be what shape does FedRAMP 20x take in terms of machine readable reporting on the security posture of any cloud environment? And then of course, the industry will standardize around that. So weโ€™re excited to see what that looks like. And also how much AI does the agency, the GSA, OMB and ultimately FedRAMP leverage because there is a tremendous amount of productivity, but also security that AI can provide. It can also introduce a lot of risks. And so weโ€™re all collaborating with that agency, as well as weโ€™re excited to see what, you know, where they draw the bright red lines and where they embrace AI.

Terry Gerton So phase II is only gonna incorporate 10 companies, right? So for the rest of the world whoโ€™s waiting on these results, what advice do you have for them in the meantime? How can companies prepare better or how can companies who want to get FedRAMP certified now best proceed?

Irina Denisenkoย  I think the end of the day the inheritance model that Knox provides โ€” and, you know, weโ€™re not the only ones, actually thereโ€™s two key players.; itโ€™s ourselves and Palantir. Thereโ€™s a reason hat large companies like Celonis like OutSystems like BigID like Armis who was just bought by ServiceNow for almost $8 billion. Thereโ€™s reason that all those guys choose Knox and thereโ€™s a reason Anthropic chose Palantir and Grafana chose Palantir, because regardless, FedRAMP 20X, Rev 5, doesnโ€™t matter, there is a massive, massive premium put on getting innovative technology in the hands of our government faster. We have a window right now with the current administration prioritizing innovative technology and commercial off-the-shelf. You know, take the best out of Silicon Valley and use it in the government or out of Europe, out of Israel, you name it, rather than build it yourself, customize it until youโ€™re blue in the face and still get an inferior product. Just use the best and breed, right? But you need it to be secure. And we have this window as a country. We have a window as country for the next few years here to get these technologies in. It takes a while to adopt new technologies. It takes awhile to do a quantum leap, but Iโ€™ll give you a perfect example. Celonis, since becoming FedRAMPed on August 19th with Knox โ€” they had been trying to get FedRAMPed for five years โ€” since getting FedRAMPed on august 19th, has implemented three agencies. And what do they do? They do process mining and intelligence. Theyโ€™re an $800 million company thatโ€™s 20 years old that competes, by the way, head on with Palantirโ€™s core product, Foundry and Gotham and so on. Theyโ€™ve implemented three agencies already to drive efficiency, to drive visibility, to drive process mining, to driving intelligence, to drive AI-powered decision-making. And thatโ€™s during the holidays, during a government shutdown, itโ€™s speed that weโ€™ve never seen before. If you want outcomes, you need to get these technologies into the hands of our agencies today. And so thatโ€™s why, you know, weโ€™re such big proponents of this model, and also why, our agencies, our federal advisory board, which includes the DHS CISO, the DOW CIO, the VA CIO are also supportive of this because ultimately itโ€™s about serving the mission and doing it now. Rather than waiting for some time in the future.

The post FedRAMP is getting faster, new automation and pilots promise approvals in months, not years first appeared on Federal News Network.

ยฉ Getty Images/iStockphoto/Kalawin

Cloud

๋กœ์ปฌ ์ปดํ“จํŒ…์œผ๋กœ ๋„˜์–ด๊ฐ€๋Š” AI ์ถ”๋ก ยทยทยทโ€˜์—ฃ์ง€ AIโ€™ ํŠธ๋ Œ๋“œ ํ•œ๋ˆˆ์— ๋ณด๊ธฐ

์—ฃ์ง€ AI์™€ ๊ด€๋ จํ•ด ๋ ˆ๋“œํ–‡์˜ ์—ฃ์ง€ ํ”„๋กœ์ ํŠธ ๊ด€๋ฆฌ ๋ถ€๋ฌธ ์‹œ๋‹ˆ์–ด ๋””๋ ‰ํ„ฐ ์กฐ์Šˆ์•„ ๋ฐ์ด๋น„๋“œ๋Š” โ€œ๊ธ€๋กœ๋ฒŒ ์‹œ์žฅ์€ ํ˜„์žฌ ๊ฐ€ํŒŒ๋ฅธ ์„ฑ์žฅ ๊ถค๋„์— ์˜ฌ๋ผ ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ์‹ค์ œ๋กœ ์—ฃ์ง€ AI ์‹œ์žฅ ๊ทœ๋ชจ๋Š” 2034๋…„๊นŒ์ง€ 1,430์–ต ๋‹ฌ๋Ÿฌ์— ์ด๋ฅผ ๊ฒƒ์œผ๋กœ ์˜ˆ์ƒ๋˜๊ณ  ์žˆ๋‹ค.

์—ฃ์ง€ AI์˜ ์„ฑ์žฅ์„ธ๋Š” ์—…๊ณ„ ์ „๋ฐ˜์ด ๋จธ์‹ ๋Ÿฌ๋‹(ML) ๋ชจ๋ธ์— ๋ฐ์ดํ„ฐ๋ฅผ ํ•™์Šต์‹œํ‚ค๋Š” AI ํ•™์Šต ์ค‘์‹ฌ์—์„œ ๋ฒ—์–ด๋‚˜, ์šด์˜ ํ™˜๊ฒฝ์—์„œ ๋ชจ๋ธ์„ ์‹ค์ œ๋กœ ํ™œ์šฉํ•ด ์ง€์‹์„ ์ ์šฉํ•˜๊ฑฐ๋‚˜ ์˜ˆ์ธก์„ ์ˆ˜ํ–‰ํ•˜๋Š” ์ถ”๋ก  ์ค‘์‹ฌ์œผ๋กœ ์ด๋™ํ•˜๋Š” ํ๋ฆ„๊ณผ ๋งž๋ฌผ๋ ค ์žˆ๋‹ค.

์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ๋ฐ์ดํ„ฐ ๊ด€๋ฆฌยทํ†ตํ•ฉ ๊ธฐ์—… ์ธํฌ๋งคํ‹ฐ์นด์˜ ์ œํ’ˆ ๊ด€๋ฆฌ ๋ถ€๋ฌธ ๋ถ€์‚ฌ์žฅ ์ˆ˜๋ฐ‹ ์•„๊ทธ๋ผ์™ˆ์€ โ€œ๊ฐ•๋ ฅํ•˜๋ฉด์„œ๋„ ์—๋„ˆ์ง€ ํšจ์œจ์ ์ธ AI ํ”„๋กœ์„ธ์„œ์˜ ๋ฐœ์ „๊ณผ ์‚ฌ๋ฌผ์ธํ„ฐ๋„ท(IoT) ๊ธฐ๊ธฐ์˜ ํ™•์‚ฐ์ด ์ด๋Ÿฌํ•œ ํ๋ฆ„์„ ๊ฐ€์†ํ•˜๊ณ  ์žˆ๋‹ค. ๋ณต์žกํ•œ AI ๋ชจ๋ธ์„ ์—ฃ์ง€ ๋””๋ฐ”์ด์Šค์—์„œ ์ง์ ‘ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ํ™˜๊ฒฝ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ด์™€ ๊ด€๋ จํ•ด ๋ชจ๊ฑด์Šคํƒ ๋ฆฌ ๋ธ”๋กœ๊ทธ๋Š” AI ์‚ฐ์—…์ด โ€œ์ƒˆ๋กญ๊ณ , ์ž ์žฌ์ ์œผ๋กœ ํ›จ์”ฌ ๋” ํฐ ๋‹จ๊ณ„์ธ AI ์ถ”๋ก ์œผ๋กœ ์ ‘์–ด๋“ค๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ•˜๋ฉฐ, ์ด ๋‹จ๊ณ„์˜ ํ•ต์‹ฌ ํŠน์ง•์œผ๋กœ ์†Œ๋น„์ž์šฉ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜๊ณผ ๊ธฐ์—… ์ „๋ฐ˜์— ๊ฑธ์ณ AI ๋ชจ๋ธ์ด ๊ด‘๋ฒ”์œ„ํ•˜๊ฒŒ ์ฑ„ํƒ๋œ๋‹ค๋Š” ์ ์„ ์งš์—ˆ๋‹ค.

์—ฃ์ง€ ์ปดํ“จํŒ…์€ ํผ๋ธ”๋ฆญ ํด๋ผ์šฐ๋“œ์™€๋Š” ๋‹ค๋ฅธ ์žฅ์ ์„ ์ œ๊ณตํ•œ๋‹ค. ํผ๋ธ”๋ฆญ ํด๋ผ์šฐ๋“œ๋Š” ํ™•์žฅ์„ฑ๊ณผ ์‚ฌ์šฉ ํŽธ์˜์„ฑ ์ธก๋ฉด์—์„œ ๊ฐ•์ ์ด ์žˆ์ง€๋งŒ, ์ถ”๋ก  ๊ด€์ ์—์„œ๋Š” ํ•œ๊ณ„๋„ ๋ถ„๋ช…ํ•˜๋‹ค. ์ง€์—ฐ ์‹œ๊ฐ„์ด ๋Š˜์–ด๋‚˜๊ณ  ๋ฐ์ดํ„ฐ ํ”„๋ผ์ด๋ฒ„์‹œ ์šฐ๋ ค๊ฐ€ ์ปค์ง€๋ฉฐ, ์—ฐ์‚ฐ ์ฒ˜๋ฆฌ์™€ ๋ฐ์ดํ„ฐ ์œ ์ž…ยท์œ ์ถœ ๊ณผ์ •์—์„œ ๋น„์šฉ ๋ถ€๋‹ด์ด ์ฆ๊ฐ€ํ•œ๋‹ค๋Š” ์ ์ด ๋Œ€ํ‘œ์ ์ด๋‹ค. ๋ฐ˜๋ฉด AI๋ฅผ ์—ฃ์ง€์—์„œ ์‹คํ–‰ํ•˜๋ฉด ์ด๋Ÿฌํ•œ ๋ฌธ์ œ ์ƒ๋‹น ๋ถ€๋ถ„์„ ํ•ด์†Œํ•  ์ˆ˜ ์žˆ๋‹ค. ๋ฐ์ด๋น„๋“œ๋Š” โ€œ์—ฃ์ง€ AI๋Š” ์ง€์—ฐ ์‹œ๊ฐ„ ๊ฐ์†Œ, ๋น„์šฉ ์ ˆ๊ฐ, ๋ณด์•ˆ๊ณผ ํ”„๋ผ์ด๋ฒ„์‹œ ๊ฐ•ํ™” ๋“ฑ ์—ฌ๋Ÿฌ ํ•ต์‹ฌ์ ์ธ ์ด์ ์„ ์ œ๊ณตํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์•„๋งˆ์กด์ด ์ตœ๊ทผ ์ผ๋ถ€ ML ํ•™์Šต ์ž‘์—…์— ์ฃผ๋กœ ์‚ฌ์šฉ๋˜๋Š” GPU ์ด์šฉ ์š”๊ธˆ์„ 15% ์ธ์ƒํ•œ ์‚ฌ๋ก€์ฒ˜๋Ÿผ, ์ค‘์•™ ์ง‘์ค‘ํ˜• ํ•™์Šต์„ ์ค‘์‹ฌ์œผ๋กœ ํ•œ ํด๋ผ์šฐ๋“œ AI ๋น„์šฉ์€ ์˜ˆ์ธกํ•˜๊ธฐ ์–ด๋ ค์šด ๋ฐฉํ–ฅ์œผ๋กœ ํ˜๋Ÿฌ๊ฐ€๊ณ  ์žˆ๋‹ค. IDC๋Š” 2027๋…„๊นŒ์ง€ CIO์˜ 80%๊ฐ€ AI ์ถ”๋ก  ์ˆ˜์š”๋ฅผ ์ถฉ์กฑํ•˜๊ธฐ ์œ„ํ•ด ํด๋ผ์šฐ๋“œ ์—…์ฒด์˜ ์—ฃ์ง€ ์„œ๋น„์Šค๋ฅผ ํ™œ์šฉํ•  ๊ฒƒ์œผ๋กœ ์ „๋งํ–ˆ๋‹ค.

๋‹ค๋งŒ ์ด๋Ÿฌํ•œ ์ „ํ™˜์ด ์ˆœํƒ„ํ•˜๊ฒŒ ์ด๋ค„์ง€์ง€๋Š” ์•Š์„ ์ „๋ง์ด๋‹ค. ์‹ค์‹œ๊ฐ„ ์„ฑ๋Šฅ์— ๋Œ€ํ•œ ๋†’์€ ์š”๊ตฌ, AI ์Šคํƒ์ด ์ฐจ์ง€ํ•˜๋Š” ํฐ ์‹œ์Šคํ…œ ์ž์› ๋ถ€๋‹ด, ํŒŒํŽธํ™”๋œ ์—ฃ์ง€ ์ƒํƒœ๊ณ„๋Š” ์—ฌ์ „ํžˆ ์ฃผ์š” ๊ณผ์ œ๋กœ ๋‚จ์•„์žˆ๋‹ค.

์ด ๊ธ€์—์„œ๋Š” ์—ฃ์ง€ AI๋ฅผ ๋‘˜๋Ÿฌ์‹ผ ๊ธฐ์ˆ  ๊ฐœ๋ฐœ ํ˜„ํ™ฉ์„ ์‚ดํŽด๋ณด๊ณ , ์—ฃ์ง€ ํ™˜๊ฒฝ์—์„œ AI๋ฅผ ์‹คํ–‰ํ•˜๊ธฐ ์œ„ํ•œ ์ƒˆ๋กœ์šด ๊ธฐ์ˆ ๊ณผ ์šด์˜ ๋ฐฉ์‹, ๊ทธ๋ฆฌ๊ณ  AI ์‹œ๋Œ€์— ์ปดํ“จํŒ…์˜ ๋ฏธ๋ž˜๊ฐ€ ์–ด๋–ป๊ฒŒ ์ง„ํ™”ํ• ์ง€ ์ „๋ฐ˜์ ์œผ๋กœ ์งš์–ด๋ณธ๋‹ค.

์—ฃ์ง€ AI ์„ฑ์žฅ์„ ์ด๋„๋Š” ์š”์ธ

๋ฐ์ด๋น„๋“œ๋Š” โ€œ์—ฃ์ง€ AI ํ™•์‚ฐ์˜ ๊ฐ€์žฅ ํฐ ๋™์ธ์€ ์‹ค์‹œ๊ฐ„ ๋ฐ์ดํ„ฐ ์ฒ˜๋ฆฌ์— ๋Œ€ํ•œ ์š”๊ตฌโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ค‘์•™ํ™”๋œ ํด๋ผ์šฐ๋“œ ๊ธฐ๋ฐ˜ AI ์›Œํฌ๋กœ๋“œ์— ์˜์กดํ•˜๋Š” ๋Œ€์‹ , ์—ฃ์ง€์—์„œ ์ง์ ‘ ๋ฐ์ดํ„ฐ๋ฅผ ๋ถ„์„ํ•˜๋ฉด ๋ฐ์ดํ„ฐ๊ฐ€ ์ƒ์„ฑ๋˜๋Š” ์ง€์ ์—์„œ ์ฆ‰๊ฐ์ ์ธ ์˜์‚ฌ๊ฒฐ์ •์„ ๋‚ด๋ฆด ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ์ด ํ•ต์‹ฌ์ด๋ผ๋Š” ๋ถ„์„์ด๋‹ค.

์—ฌ๋Ÿฌ ์ „๋ฌธ๊ฐ€๋„ ๊ฐ™์€ ์˜๊ฒฌ์„ ๋‚ด๋†“๊ณ  ์žˆ๋‹ค. ์ธํฌ๋งคํ‹ฐ์นด์˜ ์•„๊ทธ๋ผ์™ˆ์€ โ€œ์—ฃ์ง€ AI์— ๋Œ€ํ•œ ๊ด€์‹ฌ์ด ํญ๋ฐœ์ ์œผ๋กœ ์ฆ๊ฐ€ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์ „ํ•˜๋ฉฐ, ํŠนํžˆ ์‚ฐ์—… ํ˜„์žฅ์ด๋‚˜ ์ž๋™์ฐจ ๋ถ„์•ผ์ฒ˜๋Ÿผ ์ฐฐ๋‚˜์˜ ํŒ๋‹จ์ด ์ค‘์š”ํ•œ ํ™˜๊ฒฝ์—์„œ๋Š” ์ง€์—ฐ ์‹œ๊ฐ„ ๊ฐ์†Œ๊ฐ€ ๊ฐ€์žฅ ํฐ ์žฅ์ ์œผ๋กœ ์ž‘์šฉํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ์ง„๋‹จํ–ˆ๋‹ค.

๋˜ํ•œ ํด๋ผ์šฐ๋“œ๋กœ ๋ฐ์ดํ„ฐ๋ฅผ ๋ณด๋‚ด์ง€ ์•Š๊ณ ๋„ ML ๋ชจ๋ธ์— ๊ฐœ์ธ์ ์ด๊ฑฐ๋‚˜ ๊ธฐ์—… ๊ณ ์œ ์˜ ๋งฅ๋ฝ ์ •๋ณด๋ฅผ ์ œ๊ณตํ•˜๋ ค๋Š” ์ˆ˜์š” ์—ญ์‹œ ์ปค์ง€๊ณ  ์žˆ๋‹ค. ์˜คํ”ˆ์†Œ์Šค ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ํ”Œ๋žซํผ ์ œ๊ณต์‚ฌ ํ…œํฌ๋Ÿด ํ…Œํฌ๋†€๋กœ์ง€์˜ ์‹œ๋‹ˆ์–ด ์Šคํƒœํ”„ ์†Œํ”„ํŠธ์›จ์–ด ์—”์ง€๋‹ˆ์–ด์ด์ž AI ๊ธฐ์ˆ  ์ฑ…์ž„์ž์ธ ์š”ํ•œ ์А๋ผ์ด์–ด-์Šค๋ฏธ์Šค๋Š” โ€œํ”„๋ผ์ด๋ฒ„์‹œ๋Š” ๋งค์šฐ ๊ฐ•๋ ฅํ•œ ๋™์ธโ€์ด๋ผ๋ฉด์„œ, ์˜๋ฃŒ๋‚˜ ๊ธˆ์œต์ฒ˜๋Ÿผ ๊ทœ์ œ๊ฐ€ ์—„๊ฒฉํ•œ ์‚ฐ์—…์—์„œ๋Š” ๋ฏผ๊ฐํ•œ ์ •๋ณด๋ฅผ ๋กœ์ปฌ์—์„œ ์ฒ˜๋ฆฌํ•˜๋Š” ๊ฒƒ์ด ๊ทœ์ œ ์ค€์ˆ˜๋ฅผ ์œ„ํ•ด ํ•„์ˆ˜์ ์ด๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

์ˆ˜์„ธ(SUSE)์˜ ์—ฃ์ง€ ์‚ฌ์—… ๋ถ€๋ฌธ ๋ถ€์‚ฌ์žฅ ๊ฒธ ์ด๊ด„ ์ฑ…์ž„์ž ํ‚ค์Šค ๋ฐ”์‹ค ์—ญ์‹œ โ€œ์—ฃ์ง€ AI์— ๋Œ€ํ•œ ๊ด€์‹ฌ์€ ๋ถ„๋ช…ํžˆ ์ฆ๊ฐ€ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค. ๊ทธ๋Š” ์ œ์กฐ ๋ถ„์•ผ๋ฅผ ๋Œ€ํ‘œ์ ์ธ ์‚ฌ๋ก€๋กœ ๊ผฝ์œผ๋ฉฐ, ๊ธฐ์—…์ด ์ƒ์‚ฐ ๋ผ์ธ์„ ์šด์˜ํ•˜๋Š” ๋Œ€ํ˜• ์„œ๋ฒ„๋ถ€ํ„ฐ ์†Œํ˜• ์„ผ์„œ์—์„œ ๋ฐœ์ƒํ•˜๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ์ฒ˜๋ฆฌํ•˜๋Š” ์˜์—ญ๊นŒ์ง€ ๋‹ค์–‘ํ•œ ์‚ฌ์šฉ๋ก€์— ์—ฃ์ง€ AI ๋„์ž…์„ ๊ฒ€ํ† ํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋กœํฌ์›ฐ ์˜คํ† ๋ฉ”์ด์…˜์— ๋”ฐ๋ฅด๋ฉด ์ œ์กฐ ๊ธฐ์—…์˜ 95%๊ฐ€ ํ–ฅํ›„ 5๋…„ ๋‚ด์— AI/ML, ์ƒ์„ฑํ˜• AI, ์ธ๊ณผ ๊ธฐ๋ฐ˜ AI์— ์ด๋ฏธ ํˆฌ์žํ–ˆ๊ฑฐ๋‚˜ ํˆฌ์ž๋ฅผ ๊ณ„ํšํ•˜๊ณ  ์žˆ๋‹ค. ๋˜ 2024๋…„ ์ธํ…”์˜ CIO ๋ณด๊ณ ์„œ์—์„œ๋Š” ์ œ์กฐ ๋ถ„์•ผ ๋ฆฌ๋”์˜ 74%๊ฐ€ AI๊ฐ€ ๋งค์ถœ ์„ฑ์žฅ์— ๊ธฐ์—ฌํ•  ์ž ์žฌ๋ ฅ์ด ์žˆ๋‹ค๊ณ  ๋‹ตํ–ˆ๋‹ค.

๋กœ์ปฌ AI ์—ฐ์‚ฐ์˜ ๊ฐ€์žฅ ํฐ ํšจ๊ณผ๋Š” ๋น„์šฉ ์ ˆ๊ฐ์ด๋‹ค. ์•„๊ทธ๋ผ์™ˆ์€ โ€œ์ „์†กํ•ด์•ผ ํ•  ๋ฐ์ดํ„ฐ ์–‘์ด ์ค„์–ด๋“ค๋ฉด์„œ ๋น„์šฉ๊ณผ ๋Œ€์—ญํญ์„ ํฌ๊ฒŒ ์ตœ์ ํ™”ํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

ํŠน์ • ์›Œํฌ๋กœ๋“œ๋ฅผ ์—ฃ์ง€์—์„œ ์ฒ˜๋ฆฌํ•˜๋Š” ๋ฐฉ์‹์€ ๋น„์šฉ ์ ˆ๊ฐ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ์—๋„ˆ์ง€ ์†Œ๋น„ ๊ฐ์†Œ์™€๋„ ๋ฐ€์ ‘ํ•˜๊ฒŒ ์—ฐ๊ฒฐ๋œ๋‹ค. 2025๋…„ 1์›” ์•„์นด์ด๋ธŒ(Arxiv)์— ๋ฐœํ‘œ๋œ ๋…ผ๋ฌธ โ€˜ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ์—ฃ์ง€ ํด๋ผ์šฐ๋“œ์˜ ์—๋„ˆ์ง€ ๋ฐ ๋น„์šฉ ์ ˆ๊ฐ ํšจ๊ณผ ์ •๋Ÿ‰ํ™”โ€™์—์„œ๋Š” ์ˆœ์ˆ˜ ํด๋ผ์šฐ๋“œ ์ฒ˜๋ฆฌ ๋ฐฉ์‹๊ณผ ๋น„๊ตํ•ด, ์—์ด์ „ํŠธ ๊ธฐ๋ฐ˜ AI ์›Œํฌ๋กœ๋“œ์— ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ์—ฃ์ง€ ํด๋ผ์šฐ๋“œ๋ฅผ ์ ์šฉํ•  ๊ฒฝ์šฐ ์กฐ๊ฑด์— ๋”ฐ๋ผ ์ตœ๋Œ€ 75%์˜ ์—๋„ˆ์ง€ ์ ˆ๊ฐ๊ณผ 80%๋ฅผ ์›ƒ๋„๋Š” ๋น„์šฉ ์ ˆ๊ฐ ํšจ๊ณผ๋ฅผ ๊ฑฐ๋‘˜ ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

ํ•ด๋‹น ๋…ผ๋ฌธ์˜ ์ €์ž์ธ ์‹œ์•„๋ฐ”์‹œ ์•Œ๋ผ๋ฌดํ‹ฐ๋Š” โ€œ์—ฃ์ง€ ์ฒ˜๋ฆฌ๋Š” ๋กœ์ปฌ ์ปดํ“จํŒ…์˜ ๋งฅ๋ฝ์„ ์ง์ ‘ ํ™œ์šฉํ•ด ์—ฐ์‚ฐ ๋ณต์žก๋„๋ฅผ ๋‚ฎ์ถ”๊ณ , ํด๋ผ์šฐ๋“œ ๊ทœ๋ชจ์—์„œ ๋ฐœ์ƒํ•˜๋Š” ๋ง‰๋Œ€ํ•œ ์—๋„ˆ์ง€ ์ˆ˜์š”๋ฅผ ํ”ผํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์ง„๋‹จํ–ˆ๋‹ค.

๋กœ์ปฌ AI๋ฅผ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ๊ธฐ์ˆ 

์—ฃ์ง€ AI ์„ฑ์žฅ์„ ์ด๋„๋Š” ์š”์ธ์€ ๋ถ„๋ช…ํ•ด ๋ณด์ด์ง€๋งŒ, ์ด๋ฅผ ์‹ค์ œ๋กœ ๊ตฌํ˜„ํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์–ด๋–ค ๊ธฐ์ˆ ์ด ํ•„์š”ํ• ๊นŒ? ์ž์›์ด ์ œํ•œ๋œ ์—ฃ์ง€ ํ™˜๊ฒฝ์—์„œ AI ์—ฐ์‚ฐ์„ ์‹คํ–‰ํ•˜๋ ค๋ฉด ์†Œ๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ, ๊ฒฝ๋Ÿ‰ ํ”„๋ ˆ์ž„์›Œํฌ, ์ตœ์ ํ™”๋œ ๋ฐฐํฌ ๋ฐฉ์‹์ด ๊ฒฐํ•ฉ๋ผ์•ผ ํ•  ๊ฐ€๋Šฅ์„ฑ์ด ํฌ๋‹ค.

์†Œ๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(SLM)

์ง€๊ธˆ๊นŒ์ง€ ๋Œ€๋ถ€๋ถ„์˜ ๊ธฐ์—…์€ ์•คํŠธ๋กœํ”ฝ์˜ ํด๋กœ๋“œ, ๊ตฌ๊ธ€์˜ ์ œ๋ฏธ๋‚˜์ด, ์˜คํ”ˆAI์˜ GPT ๋ชจ๋ธ๊ณผ ๊ฐ™์€ ๋ฒ”์šฉ ์„œ๋น„์Šค ๊ธฐ๋ฐ˜์˜ LLM์„ ํ™œ์šฉํ•ด AI๋ฅผ ์ค‘์•™์—์„œ ์šด์˜ํ•ด ์™”๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์ตœ๊ทผ AI ๋ชจ๋ธ ๊ธฐ์ˆ ์ด ๋ฐœ์ „ํ•˜๋ฉด์„œ ์ด๋Ÿฐ ๊ตฌ์กฐ์—๋„ ๋ณ€ํ™”๊ฐ€ ๋‚˜ํƒ€๋‚˜๊ณ  ์žˆ๋‹ค.

ํŠนํžˆ ์ž์ฒด ๋ฐฐํฌ๊ฐ€ ๊ฐ€๋Šฅํ•œ SLM์€ ํŠน์ • ์‚ฌ์šฉ๋ก€์—์„œ ํด๋ผ์šฐ๋“œ AI ํ”Œ๋žซํผ์— ๋Œ€ํ•œ ์˜์กด๋„๋ฅผ ๋‚ฎ์ถ”๊ณ  ์žˆ๋‹ค. ํ…œํฌ๋Ÿด ํ…Œํฌ๋†€๋กœ์ง€์˜ ์А๋ผ์ด์–ด-์Šค๋ฏธ์Šค๋Š” โ€œSLM์˜ ์„ฑ๋Šฅ์ด ๋น ๋ฅด๊ฒŒ ํ–ฅ์ƒ๋˜๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ, ์ตœ๊ทผ ๊ธฐ์ˆ  ์ง„์ „์˜ ์‚ฌ๋ก€๋กœ ์˜คํ”ˆAI์˜ GPT-OSS์™€ ๊ณ„์ธต์  ์ถ”๋ก  ๋ชจ๋ธ์„ ์–ธ๊ธ‰ํ–ˆ๋‹ค.

์ตœ์ ํ™” ์ „๋žต

๋กœ์ปฌ AI์˜ ์‹œ์Šคํ…œ ๋ถ€๋‹ด์„ ์ค„์ด๋Š” ๊ฒƒ์€ ์—ฐ์‚ฐ ๋Šฅ๋ ฅ๊ณผ ๋Œ€์—ญํญ์ด ์ œํ•œ๋œ ์—ฃ์ง€ ๋””๋ฐ”์ด์Šค ํ™˜๊ฒฝ์—์„œ ํŠนํžˆ ์ค‘์š”ํ•˜๋‹ค. ์ด์— ๋”ฐ๋ผ SLM์„ ์ตœ์ ํ™”ํ•˜๋Š” ๊ธฐ์ˆ ์ด ์—ฃ์ง€ AI ๊ตฌํ˜„์„ ๋’ท๋ฐ›์นจํ•˜๋Š” ํ•ต์‹ฌ ์˜์—ญ์œผ๋กœ ๋– ์˜ค๋ฅด๊ณ  ์žˆ๋‹ค.

๋Œ€ํ‘œ์ ์ธ ์ „๋žต ๊ฐ€์šด๋ฐ ํ•˜๋‚˜๋Š” ๋ชจ๋ธ ์••์ถ• ๊ธฐ๋ฒ•์ธ ์–‘์žํ™”๋‹ค. ์–‘์žํ™”๋Š” ๋ชจ๋ธ ํฌ๊ธฐ์™€ ์—ฐ์‚ฐ ์š”๊ตฌ๋Ÿ‰์„ ์ค„์ด๋Š” ๋ฐฉ์‹์œผ๋กœ, ์—ฃ์ง€ ํ™˜๊ฒฝ์— ์ ํ•ฉํ•œ ํ˜•ํƒœ๋กœ AI ๋ชจ๋ธ์„ ๊ฒฝ๋Ÿ‰ํ™”ํ•œ๋‹ค. ์•„๊ทธ๋ผ์™ˆ์€ โ€œ์ด๋Ÿฌํ•œ ๊ธฐ๋ฒ•์„ ํ†ตํ•ด NPU, ๊ตฌ๊ธ€์˜ ์—ฃ์ง€ TPU, ์• ํ”Œ์˜ ๋‰ด๋Ÿด ์—”์ง„, ์—”๋น„๋””์•„ ์ ฏ์Šจ ๋””๋ฐ”์ด์Šค์™€ ๊ฐ™์€ ํŠนํ™” ํ•˜๋“œ์›จ์–ด์—์„œ SLM์„ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ž์ฒด ์™„๊ฒฐํ˜• ํŒจํ‚ค์ง€ ์—ญ์‹œ ์—ฃ์ง€ AI๋ฅผ ๋Œ€๊ทœ๋ชจ๋กœ ์šด์˜ํ•˜๋Š” ๋ฐ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ๋‹ค. ๋ฐ์ด๋น„๋“œ๋Š” ์šด์˜์ฒด์ œ, ํ•˜๋“œ์›จ์–ด ๋“œ๋ผ์ด๋ฒ„, AI ๋ชจ๋ธ์„ ํ•˜๋‚˜๋กœ ๋ฌถ์€ ์ฆ‰์‹œ ๋ฐฐํฌ ๊ฐ€๋Šฅํ•œ ๋ฒ ์ด์Šค ์ด๋ฏธ์ง€๊ฐ€ ์—ฃ์ง€ AI ์ƒ์šฉํ™”๋ฅผ ํ˜„์‹คํ™”ํ•˜๋Š” ์ ‘๊ทผ ๋ฐฉ์‹์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์—ฃ์ง€ ๋Ÿฐํƒ€์ž„ ๋ฐ ํ”„๋ ˆ์ž„์›Œํฌ

์ƒˆ๋กœ์šด ๋Ÿฐํƒ€์ž„ ๋ฐ ํ”„๋ ˆ์ž„์›Œํฌ ์—ญ์‹œ ์—ฃ์ง€ ํ™˜๊ฒฝ์—์„œ์˜ AI ์ถ”๋ก ์„ ์ตœ์ ํ™”ํ•˜๋Š” ๋ฐ ์ค‘์š”ํ•œ ์—ญํ• ์„ ํ•œ๋‹ค. ๋ฐ์ด๋น„๋“œ๋Š” ๊ฒฝ๋Ÿ‰ ์ƒ์„ฑํ˜• AI ๋Ÿฐํƒ€์ž„์ธ llama.cpp์™€ ํ•จ๊ป˜, ๋กœ์ปฌ ํ•˜๋“œ์›จ์–ด์—์„œ ๋ชจ๋ธ ์ถ”๋ก ์„ ์ง€์›ํ•˜๋Š” ์˜คํ”ˆ๋น„๋…ธ(OpenVINO)์™€ ๋ผ์ดํŠธRT(LiteRT, ์ด์ „ ํ…์„œํ”Œ๋กœ ๋ผ์ดํŠธ) ๊ฐ™์€ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์–ธ๊ธ‰ํ–ˆ๋‹ค.

์•„๊ทธ๋ผ์™ˆ์€ โ€œllama.cpp์™€ GGUF ๋ชจ๋ธ ํฌ๋งท ๊ฐ™์€ ํ”„๋กœ์ ํŠธ๋Š” ๋‹ค์–‘ํ•œ ์†Œ๋น„์ž์šฉ ๋””๋ฐ”์ด์Šค์—์„œ ๊ณ ์„ฑ๋Šฅ ์ถ”๋ก ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๊ณ  ์žˆ๋‹ค. MLC LLM๊ณผ ์›นLLM(WebLLM) ์—ญ์‹œ ์›น ๋ธŒ๋ผ์šฐ์ €์™€ ๋‹ค์–‘ํ•œ ๋„ค์ดํ‹ฐ๋ธŒ ํ”Œ๋žซํผ์—์„œ AI๋ฅผ ์ง์ ‘ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ๊ฐ€๋Šฅ์„ฑ์„ ํ™•์žฅํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

ํด๋ผ์šฐ๋“œ ๋„ค์ดํ‹ฐ๋ธŒ ํ˜ธํ™˜์„ฑ

์—ฃ์ง€ AI๊ฐ€ ํด๋ผ์šฐ๋“œ ๋„ค์ดํ‹ฐ๋ธŒ ์ƒํƒœ๊ณ„ ๋ฐ ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค์™€์˜ ํ˜ธํ™˜์„ฑ์„ ํ™•๋ณดํ•˜๋Š” ๊ฒƒ ์—ญ์‹œ ์ค‘์š”ํ•œ ๊ณผ์ œ๋กœ ๋– ์˜ค๋ฅด๊ณ  ์žˆ๋‹ค. ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค๊ฐ€ ์ด๋ฏธ ์—ฃ์ง€ ํ™˜๊ฒฝ์œผ๋กœ ๋น ๋ฅด๊ฒŒ ํ™•์‚ฐ๋˜๊ณ  ์žˆ๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค. ๋Œ€ํ‘œ์ ์ธ ์‚ฌ๋ก€๋กœ๋Š” โ€˜์ž์ฒด ํ˜ธ์ŠคํŒ… AI๋ฅผ ์œ„ํ•œ ์˜คํ”ˆ์†Œ์Šค ํ‘œ์ค€โ€™์œผ๋กœ ์†Œ๊ฐœ๋˜๋Š” ์ผ€์ด์„œ๋ธŒ(KServe)๊ฐ€ ์žˆ๋‹ค. ์ผ€์ด์„œ๋ธŒ๋Š” ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค ํ™˜๊ฒฝ์—์„œ ์—ฃ์ง€ ์ถ”๋ก ์„ ์ง€์›ํ•˜๋Š” ํ”„๋ ˆ์ž„์›Œํฌ๋‹ค.

๋˜ ๋‹ค๋ฅธ ํ•ต์‹ฌ ๊ธฐ์ˆ ๋กœ๋Š” ํด๋ผ์šฐ๋“œ ๋„ค์ดํ‹ฐ๋ธŒ ์ปดํ“จํŒ… ์žฌ๋‹จ(CNCF)์ด ์šด์˜ํ•˜๋Š” ์ƒŒ๋“œ๋ฐ•์Šค ํ”„๋กœ์ ํŠธ ์•„ํฌ๋ฆฌ(Akri)๊ฐ€ ๊ผฝํžŒ๋‹ค. ์ˆ˜์„ธ์˜ ๋ฐ”์‹ค์€ โ€œ์•„ํฌ๋ฆฌ๋Š” ์—ฃ์ง€ ํ™˜๊ฒฝ์—์„œ ๋‹ค์–‘ํ•œ ๋™์  ๋””๋ฐ”์ด์Šค์™€ ๊ฐ„ํ—์ ์œผ๋กœ ์—ฐ๊ฒฐ๋˜๋Š” ๋ฆฌํ”„ ๋””๋ฐ”์ด์Šค๋ฅผ ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค์—์„œ ์‰ฝ๊ฒŒ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ๋งŒ๋“ ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์•„ํฌ๋ฆฌ๋ฅผ ํ™œ์šฉํ•˜๋ฉด IP ์นด๋ฉ”๋ผ, ์„ผ์„œ, USB ๋””๋ฐ”์ด์Šค ๋“ฑ ๋‹ค์–‘ํ•œ ์—”๋“œํฌ์ธํŠธ๋ฅผ ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค ๋ฆฌ์†Œ์Šค๋กœ ๋…ธ์ถœํ•  ์ˆ˜ ์žˆ์–ด, ํ•ด๋‹น ํ•˜๋“œ์›จ์–ด์— ์˜์กดํ•˜๋Š” ์—ฃ์ง€ AI ์›Œํฌ๋กœ๋“œ๋ฅผ ๋ณด๋‹ค ์†์‰ฝ๊ฒŒ ๋ฐฐํฌํ•˜๊ณ  ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค์—์„œ ๋ชจ๋‹ˆํ„ฐ๋งํ•  ์ˆ˜ ์žˆ๋‹ค.

๊ฐœ๋ฐฉํ˜• ํ‘œ์ค€

๋งˆ์ง€๋ง‰์œผ๋กœ ์—ฃ์ง€ AI ํ™•์‚ฐ ๊ณผ์ •์—์„œ ๊ฐœ๋ฐฉํ˜• ์‚ฐ์—… ํ‘œ์ค€์ด ์ค‘์š”ํ•œ ์—ญํ• ์„ ํ•  ๊ฒƒ์œผ๋กœ ๋ณด์ธ๋‹ค. ๋ฐ”์‹ค์€ โ€œ๋น ๋ฅด๊ฒŒ ํ™•์žฅ๋˜๊ณ  ์žˆ๋Š” ์—ฃ์ง€ AI ํ•˜๋“œ์›จ์–ด์™€ ์†Œํ”„ํŠธ์›จ์–ด ํ™˜๊ฒฝ์€ ์‹ฌ๊ฐํ•œ ์ƒํ˜ธ์šด์šฉ์„ฑ ๋ฌธ์ œ๋ฅผ ์•ˆ๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” ๋ฆฌ๋ˆ…์Šค ์žฌ๋‹จ์ด ์ฃผ๋„ํ•˜๋Š” ๋งˆ๊ณ (Margo)์™€ ๊ฐ™์€ ํ”„๋กœ์ ํŠธ๊ฐ€ ์‚ฐ์—…์šฉ ์—ฃ์ง€ ์ž๋™ํ™” ๋ถ„์•ผ์—์„œ ํ‘œ์ค€์„ ์ •๋ฆฝํ•˜๋Š” ๋ฐ ์ค‘์š”ํ•œ ์—ญํ• ์„ ํ•  ๊ฒƒ์œผ๋กœ ๋‚ด๋‹ค๋ดค๋‹ค.

์ด์™€ ํ•จ๊ป˜ ONNX๋„ ์˜จ๋””๋ฐ”์ด์Šค AI ์ถ”๋ก ์„ ์œ„ํ•œ ๊ฒฝ์Ÿ ํ”„๋ ˆ์ž„์›Œํฌ ๊ฐ„ ์ƒํ˜ธ์šด์šฉ์„ฑ ๋ฌธ์ œ ํ•ด๊ฒฐ์— ๋„์›€์ด ๋  ํ‘œ์ค€์œผ๋กœ ์ฃผ๋ชฉ๋ฐ›๊ณ  ์žˆ๋‹ค.

์—ฃ์ง€ AI์˜ ํ˜„์‹ค์  ์žฅ๋ฒฝ

๊ธฐ์ˆ  ์ž์ฒด๋Š” ์ด๋ฏธ ๋งˆ๋ จ๋ผ ์žˆ์ง€๋งŒ, ์—ฃ์ง€ AI๋ฅผ ์‹ค์ œ๋กœ ์šด์˜ํ•˜๋Š” ๋ฐฉ์‹์€ ์•„์ง ์ดˆ๊ธฐ ๋‹จ๊ณ„์— ๋จธ๋ฌผ๋Ÿฌ ์žˆ๋‹ค. ๊ฐœ๋… ์ฆ๋ช… ์ˆ˜์ค€์—์„œ ๋ฒ—์–ด๋‚˜ ์—ฃ์ง€ AI ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ๋ณธ๊ฒฉ์ ์ธ ์šด์˜ ํ™˜๊ฒฝ์œผ๋กœ ์˜ฎ๊ธฐ๊ธฐ๊นŒ์ง€๋Š” ์—ฌ๋Ÿฌ ๊ณผ์ œ๋ฅผ ๊ทน๋ณตํ•ด์•ผ ํ•œ๋‹ค.

์•„๊ทธ๋ผ์™ˆ์€ โ€œ๊ฐ€์žฅ ๊ทผ๋ณธ์ ์ธ ํ•œ๊ณ„๋Š” ์—ฃ์ง€ ๋””๋ฐ”์ด์Šค๊ฐ€ ์ง€๋‹Œ ์ž์› ์ œ์•ฝโ€์ด๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ๋ฉ”๋ชจ๋ฆฌ์™€ ์—ฐ์‚ฐ ๋Šฅ๋ ฅ์ด ์ œํ•œ์ ์ด๊ธฐ ๋•Œ๋ฌธ์—, ๋ง‰๋Œ€ํ•œ ์—ฐ์‚ฐ ์ž์›์„ ์š”๊ตฌํ•˜๋Š” ํฌ๊ณ  ๋ณต์žกํ•œ AI ๋ชจ๋ธ์„ ๋ฐฐํฌํ•˜๋Š” ๋ฐ ์–ด๋ ค์›€์ด ๋”ฐ๋ฅธ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์ž์›์ด ์ œํ•œ๋œ ํ•˜๋“œ์›จ์–ด ํ™˜๊ฒฝ์— ๋งž๊ฒŒ ๋ชจ๋ธ ํฌ๊ธฐ๋ฅผ ์ตœ์ ํ™”ํ•˜๋ฉด์„œ๋„, ์‚ฌ์šฉ์ž๊ฐ€ ๊ธฐ๋Œ€ํ•˜๋Š” ๊ณ ์„ฑ๋Šฅ ์ƒ์œ„ ๋ชจ๋ธ ์ˆ˜์ค€์˜ ์ •ํ™•๋„๋ฅผ ์œ ์ง€ํ•ด์•ผ ํ•œ๋‹ค๋Š” ์  ์—ญ์‹œ ์—ฌ์ „ํžˆ ํ•ด๊ฒฐํ•ด์•ผ ํ•  ๊ณผ์ œ๋‹ค.

๋˜ํ•œ ์—ฃ์ง€ AI ์šด์˜์— ๋Œ€ํ•œ ์‹ค๋ฌด ๊ฒฝํ—˜๋„ ์•„์ง ์ถฉ๋ถ„ํžˆ ์ถ•์ ๋˜์ง€ ์•Š์•˜๋‹ค. ๋ฐ์ด๋น„๋“œ๋Š” โ€œํŠนํ™”๋œ ์—ฃ์ง€ ๋””๋ฐ”์ด์Šค๋Š” ํ•˜๋“œ์›จ์–ด ๊ตฌ์„ฑ์ด ๋ณต์žกํ•˜๊ณ , ๋Œ€๋ถ€๋ถ„ ์ฆ‰์‹œ ์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ์ƒํƒœ๋กœ ๋™์ž‘ํ•˜์ง€ ์•Š๋Š”๋‹ค๋Š” ์ ์ด ์ฃผ์š” ์žฅ์• ๋ฌผโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” ์ตœ์ „๋ฐฉ ์—ฃ์ง€ ํ™˜๊ฒฝ์—์„œ ๋ชจ๋ธ์„ ๋ฐฐํฌํ•˜๊ณ  ๋ชจ๋‹ˆํ„ฐ๋งํ•˜๋ฉฐ ๊ด€๋ฆฌํ•  ์ˆ˜ ์žˆ๋Š” ์—”๋“œํˆฌ์—”๋“œ ํ”Œ๋žซํผ์ด ๋ถ€์กฑํ•ด, ํ˜„์žฌ๋กœ์„œ๋Š” ๋ณต์žกํ•œ ์ˆ˜์ž‘์—… ๋ฐฉ์‹์— ์˜์กดํ•  ์ˆ˜๋ฐ–์— ์—†๋Š” ์ƒํ™ฉ์ด๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค.

๋ฐ”์‹ค์€ โ€œ์—ฃ์ง€ AI์˜ ์ฃผ์š” ๊ณผ์ œ๋Š” ํŒŒํŽธํ™”๋œ ์ƒํƒœ๊ณ„โ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค. ๊ทธ๋Š” โ€œํ‘œ์ค€ํ™”๋˜๊ณ  ์„ฑ์ˆ™ํ•œ ํด๋ผ์šฐ๋“œ ์ปดํ“จํŒ… ํ™˜๊ฒฝ๊ณผ ๋‹ฌ๋ฆฌ, ์—ฃ์ง€ AI๋Š” ํ•˜๋“œ์›จ์–ด์™€ ์†Œํ”„ํŠธ์›จ์–ด, ํ†ต์‹  ํ”„๋กœํ† ์ฝœ ์ „๋ฐ˜์— ๊ฑธ์ณ ๊ณตํ†ต๋œ ํ”„๋ ˆ์ž„์›Œํฌ๊ฐ€ ๋ถ€์กฑํ•˜๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด๋Ÿฌํ•œ ์‚ฐ์—… ์ „๋ฐ˜์˜ ํŒŒํŽธํ™”๋Š” ๋””๋ฐ”์ด์Šค๋ณ„๋กœ ๊ฒฝ์Ÿ์ ์ธ ์†Œํ”„ํŠธ์›จ์–ด์™€ ๊ธฐ์ˆ ์„ ์–‘์‚ฐํ•˜๊ณ , ๊ทธ ๊ฒฐ๊ณผ ์—ฃ์ง€ ํ™˜๊ฒฝ์—์„œ๋Š” ํ˜ธํ™˜์„ฑ ๋ฌธ์ œ์™€ ๊ฐœ๋ณ„ ๋งž์ถคํ˜• ํ•ด๊ฒฐ์ฑ…์ด ๋’ค๋”ฐ๋ฅด๊ฒŒ ๋œ๋‹ค.

๋งˆ์ง€๋ง‰์œผ๋กœ ๋ถ„์‚ฐ๋œ AI ๋ชจ๋ธ ๋„คํŠธ์›Œํฌ๋ฅผ ๊ด€๋ฆฌํ•˜๋Š” ๋ฌธ์ œ๋„ ๋ณต์žกํ•œ ์šด์˜ ๊ณผ์ œ๋กœ ๊ผฝํžŒ๋‹ค. ์•„๊ทธ๋ผ์™ˆ์€ โ€œ์ˆ˜๋งŽ์€ ๋””๋ฐ”์ด์Šค์— ๋ฐฐํฌ๋œ ๋ชจ๋ธ์„ ๋Œ€์ƒ์œผ๋กœ ๋ณด์•ˆ ์—…๋ฐ์ดํŠธ์™€ ๋ฒ„์ „ ๊ด€๋ฆฌ, ์„ฑ๋Šฅ ๋ชจ๋‹ˆํ„ฐ๋ง์„ ๋™์‹œ์— ์ˆ˜ํ–‰ํ•˜๋Š” ๊ฒƒ์€ ๋งค์šฐ ์–ด๋ ค์šด ์ž‘์—…โ€์ด๋ผ๋ฉฐ โ€œ์—ฃ์ง€ AI๋ฅผ ํšจ๊ณผ์ ์œผ๋กœ ํ™•์žฅํ•˜๊ธฐ ์œ„ํ•ด ๊ธฐ์—…์ด ๋ฐ˜๋“œ์‹œ ํ•ด๊ฒฐํ•ด์•ผ ํ•  ๊ณผ์ œโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ด๋Ÿฌํ•œ ์žฅ๋ฒฝ์„ ๊ทน๋ณตํ•˜๊ธฐ ์œ„ํ•ด ์ „๋ฌธ๊ฐ€๋“ค์€ ๋ช‡ ๊ฐ€์ง€ ์‹ค์ฒœ ๋ฐฉ์•ˆ์„ ์ œ์‹œํ–ˆ๋‹ค.

  • ์—ฐ๊ฒฐ์„ฑ์ด ๋‚ฎ์€ ํ™˜๊ฒฝ์—์„œ์˜ ์ถ”๋ก ๊ณผ ๊ฐ™์ด ์—ฃ์ง€ AI๊ฐ€ ์ ํ•ฉํ•œ ๊ฒฝ์šฐ์— ํ•œํ•ด ๋„์ž….
  • ๋น„๊ธฐ์ˆ  ๋ถ„์•ผ ๋ฆฌ๋”๋ฅผ ๋Œ€์ƒ์œผ๋กœ ๋น„์ฆˆ๋‹ˆ์Šค ๊ฐ€์น˜๋ฅผ ์ง€์†์ ์œผ๋กœ ์ „๋‹ฌ.
  • ์ „๋ฉด์ ์ธ ์—ฃ์ง€ ๋˜๋Š” ํด๋ผ์šฐ๋“œ ๋ฐฐํฌ ๋Œ€์‹  ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ์ „๋žต์„ ๊ณ ๋ ค.
  • ์•„ํ‚คํ…์ฒ˜ ์ฐจ์›์—์„œ ์†Œํ”„ํŠธ์›จ์–ด ๊ณ„์ธต์„ ํŠน์ • ํ•˜๋“œ์›จ์–ด ์˜์กด์„ฑ์œผ๋กœ๋ถ€ํ„ฐ ๋ถ„๋ฆฌ.
  • ์—ฃ์ง€ ํ™˜๊ฒฝ์˜ ์ œ์•ฝ์— ์ตœ์ ํ™”๋œ ๋ชจ๋ธ์„ ์„ ํƒ.
  • ์—…๋ฐ์ดํŠธ, ๋ชจ๋‹ˆํ„ฐ๋ง, ์œ ์ง€๋ณด์ˆ˜๋ฅผ ํฌํ•จํ•œ ์ „์ฒด ๋ชจ๋ธ ์ˆ˜๋ช… ์ฃผ๊ธฐ๋ฅผ ์ดˆ๊ธฐ ๋‹จ๊ณ„๋ถ€ํ„ฐ ๊ตฌ์ƒ.

์ค‘์•™ ์ง‘์ค‘ํ˜•์—์„œ ๋ถ„์‚ฐ ์ง€๋Šฅ์œผ๋กœ

์—ฃ์ง€ AI์— ๋Œ€ํ•œ ๊ด€์‹ฌ์ด ๋น ๋ฅด๊ฒŒ ๋†’์•„์ง€๊ณ  ์žˆ์ง€๋งŒ, ๋Œ€์•ˆํ˜• ํด๋ผ์šฐ๋“œ๋กœ์˜ ์ „ํ™˜ ํ๋ฆ„๊ณผ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ ์ „๋ฌธ๊ฐ€๋“ค์€ ๋กœ์ปฌ ์ฒ˜๋ฆฌ๊ฐ€ ์ค‘์•™ ์ง‘์ค‘ํ˜• ํด๋ผ์šฐ๋“œ์— ๋Œ€ํ•œ ์˜์กด๋„๋ฅผ ์˜๋ฏธ ์žˆ๊ฒŒ ๋‚ฎ์ถ”์ง€๋Š” ์•Š์„ ๊ฒƒ์œผ๋กœ ๋ณด๊ณ  ์žˆ๋‹ค. ์А๋ผ์ด์–ด-์Šค๋ฏธ์Šค๋Š” โ€œ์—ฃ์ง€ AI๊ฐ€ ๋ณธ๊ฒฉ์ ์œผ๋กœ ์ฃผ๋ชฉ๋ฐ›๋Š” ์‹œ์ ์€ ์˜ค๊ฒ ์ง€๋งŒ, ๋„์ž… ์†๋„๋Š” ํด๋ผ์šฐ๋“œ์— ๋น„ํ•ด ๋’ค์ฒ˜์งˆ ๊ฒƒโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์—ฃ์ง€ AI๋Š” ํผ๋ธ”๋ฆญ ํด๋ผ์šฐ๋“œ๋ฅผ ๋Œ€์ฒดํ•˜๊ธฐ๋ณด๋‹ค๋Š” ์ƒˆ๋กœ์šด ์—ญ๋Ÿ‰์„ ๋”ํ•ด ๋ณด์™„ํ•˜๋Š” ๋ฐฉํ–ฅ์œผ๋กœ ๋ฐœ์ „ํ•  ๊ฐ€๋Šฅ์„ฑ์ด ํฌ๋‹ค. ๋ฐ”์‹ค์€ โ€œ๊ธฐ์กด ์ธํ”„๋ผ๋ฅผ ๋Œ€์ฒดํ•˜๋Š” ๊ฒƒ์ด ์•„๋‹ˆ๋ผ, ์—ฃ์ง€์— AI๋ฅผ ๋ฐฐ์น˜ํ•ด ๋” ๋˜‘๋˜‘ํ•˜๊ณ  ํšจ์œจ์ ์ด๋ฉฐ ๋ฐ˜์‘์„ฑ์ด ๋†’์€ ํ™˜๊ฒฝ์„ ๋งŒ๋“œ๋Š” ๋ฐฉ์‹์ด ๋  ๊ฒƒโ€์ด๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ์ด๋Š” ๊ธฐ์กด ์šด์˜์ฒด์ œ๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ์—”๋“œํฌ์ธํŠธ๋ฅผ ๋ณด์™„ํ•˜๊ฑฐ๋‚˜, ์˜จํ”„๋ ˆ๋ฏธ์Šค ์„œ๋ฒ„ ์šด์˜์„ ์ตœ์ ํ™”ํ•˜๋Š” ํ˜•ํƒœ๋กœ ์ด์–ด์งˆ ์ˆ˜ ์žˆ๋‹ค.

์ „๋ฌธ๊ฐ€๋“ค์€ ์—ฃ์ง€ ๋””๋ฐ”์ด์Šค์˜ ์—ญํ• ๊ณผ ์—ญ๋Ÿ‰์ด ๋‹จ๊ธฐ๊ฐ„ ๋‚ด ํฌ๊ฒŒ ๊ฐ•ํ™”๋  ๊ฒƒ์ด๋ผ๊ณ  ์˜๊ฒฌ์„ ๋ชจ์•˜๋‹ค. ์•„๊ทธ๋ผ์™ˆ์€ โ€œํ•˜๋“œ์›จ์–ด์™€ ์ตœ์ ํ™”๋œ ๋ชจ๋ธ, ๋ฐฐํฌ ํ”Œ๋žซํผ์ด ๋น ๋ฅด๊ฒŒ ๋ฐœ์ „ํ•˜๋ฉด์„œ ์‚ฌ๋ฌผ์ธํ„ฐ๋„ท, ๋ชจ๋ฐ”์ผ ๋””๋ฐ”์ด์Šค, ์ผ์ƒ์ ์ธ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ์ „๋ฐ˜์— AI๊ฐ€ ๋”์šฑ ๊นŠ์ด ํ†ตํ•ฉ๋˜๋Š” ๋ชจ์Šต์„ ๋ณด๊ฒŒ ๋  ๊ฒƒโ€์ด๋ผ๊ณ  ์ „๋งํ–ˆ๋‹ค.

์ด์–ด ๊ทธ๋Š” โ€œ์•ž์œผ๋กœ ์—ฃ์ง€ AI๋Š” ๋น ๋ฅธ ์„ฑ์žฅ์„ ์•ž๋‘๊ณ  ์žˆ์œผ๋ฉฐ, ๋ถ„์‚ฐ๋˜๊ณ  ์‚ฌ์šฉ์ž ์ค‘์‹ฌ์ ์ธ ์ง€๋Šฅ์œผ๋กœ์˜ ๊ทผ๋ณธ์ ์ธ ์ „ํ™˜์„ ์ด๋Œ ๊ฒƒโ€์ด๋ผ๊ณ  ๋‚ด๋‹ค๋ดค๋‹ค.
dl-ciokorea@foundryco.com

์˜คํ”ˆํ…์ŠคํŠธ, ๊ธฐ๊ฐ€์˜ด โ€˜2025 ํด๋ผ์šฐ๋“œ ์„ฑ๋Šฅ ํ…Œ์ŠคํŠธ ๋ ˆ์ด๋”โ€™์—์„œ 5๋…„ ์—ฐ์† ์ตœ๊ณ  ํ‰๊ฐ€ ํš๋“

์ด๋ฒˆ ๋ณด๊ณ ์„œ์—์„œ ๊ธฐ๊ฐ€์˜ด(GigaOm)์€ ์˜คํ”ˆํ…์ŠคํŠธ๋ฅผ ํด๋ผ์šฐ๋“œ ์„ฑ๋Šฅ ํ…Œ์ŠคํŠธ ์‹œ์žฅ์—์„œ ๊ธฐ์ˆ ๋ ฅ๊ณผ ๋น„์ „, ์‹คํ–‰ ์—ญ๋Ÿ‰์„ ๋ชจ๋‘ ๊ฐ–์ถ˜ ์„ ๋„ ๊ธฐ์—…์œผ๋กœ ํ‰๊ฐ€ํ–ˆ๋‹ค. ํŠนํžˆ ๋‹จ์ˆœํ•œ ๋ถ€ํ•˜ ํ…Œ์ŠคํŠธ๋ฅผ ๋„˜์–ด, ํด๋ผ์šฐ๋“œ ๋„ค์ดํ‹ฐ๋ธŒ ํ™˜๊ฒฝ๊ณผ AI ๊ธฐ์ˆ ์„ ๊ฒฐํ•ฉํ•œ ์„ฑ๋Šฅ ์—”์ง€๋‹ˆ์–ด๋ง ์—ญ๋Ÿ‰์„ ํ†ตํ•ด ์‹œ์žฅ์„ ์ฃผ๋„ํ•˜๊ณ  ์žˆ๋‹ค๋Š” ์ ์—์„œ ๋†’์€ ์ ์ˆ˜๋ฅผ ์ œ์‹œํ–ˆ๋‹ค.

๊ฐ€๊ธฐ์˜ด์€ ์˜คํ”ˆํ…์ŠคํŠธ์˜ ์„ฑ๋Šฅ ์—”์ง€๋‹ˆ์–ด๋ง ์†”๋ฃจ์…˜์ด ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ ์ƒ์• ์ฃผ๊ธฐ ์ „๋ฐ˜์— ๊ฑธ์ณ ์„ฑ๋Šฅ ๊ฒ€์ฆ์„ ๋‚ด์žฌํ™”ํ•  ์ˆ˜ ์žˆ๋„๋ก ์„ค๊ณ„๋๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค. ๊ฐœ๋ฐœ ์ดˆ๊ธฐ ๋‹จ๊ณ„๋ถ€ํ„ฐ ์šด์˜ ํ™˜๊ฒฝ์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ์ผ๊ด€๋œ ์„ฑ๋Šฅ ํ…Œ์ŠคํŠธ๋ฅผ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ด๋ฅผ ํ†ตํ•ด ๊ธฐ์—…์ด ์‚ฌ์šฉ์ž ๊ฒฝํ—˜ ์ €ํ•˜์™€ ์„œ๋น„์Šค ์žฅ์• ๋ฅผ ์‚ฌ์ „์— ์˜ˆ๋ฐฉํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ์ด ๊ฐ•์ ์œผ๋กœ ๊ผฝํ˜”๋‹ค.

๋˜ํ•œ ๋ณด๊ณ ์„œ๋Š” ์˜คํ”ˆํ…์ŠคํŠธ์˜ AI ๊ธฐ๋ฐ˜ ์„ฑ๋Šฅ ํ…Œ์ŠคํŠธ ๊ธฐ๋Šฅ์— ์ฃผ๋ชฉํ–ˆ๋‹ค. AI๋ฅผ ํ™œ์šฉํ•œ ํ…Œ์ŠคํŠธ ์Šคํฌ๋ฆฝํŠธ ์ƒ์„ฑ๊ณผ ๋ถ„์„ ๊ธฐ๋Šฅ์ด ํ…Œ์ŠคํŠธ ์ค€๋น„ ์‹œ๊ฐ„์„ ๋‹จ์ถ•ํ•˜๊ณ , ์ด์ƒ ์ง•ํ›„ ํƒ์ง€์™€ ๊ทผ๋ณธ ์›์ธ ๋ถ„์„์„ ์ž๋™ํ™”ํ•จ์œผ๋กœ์จ ์„ฑ๋Šฅ ๋ฌธ์ œ๋ฅผ ๋ณด๋‹ค ์‹ ์†ํ•˜๊ฒŒ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•œ๋‹ค๋Š” ํ‰๊ฐ€๋‹ค. ๊ธฐ๊ฐ€์˜ด์€ ์ด๋Ÿฌํ•œ ์ ‘๊ทผ ๋ฐฉ์‹์ด ์ˆ˜๋™ ํ…Œ์ŠคํŠธ์˜ ํ•œ๊ณ„๋ฅผ ๊ทน๋ณตํ•˜๊ณ , ๋Œ€๊ทœ๋ชจ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ๋„ ๋†’์€ ์ •ํ™•์„ฑ๊ณผ ํšจ์œจ์„ฑ์„ ์ œ๊ณตํ•œ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์˜คํ”ˆํ…์ŠคํŠธ์˜ ํด๋ผ์šฐ๋“œ ๋„ค์ดํ‹ฐ๋ธŒ ์•„ํ‚คํ…์ฒ˜ ์—ญ์‹œ ์ฃผ์š” ๊ฒฝ์Ÿ๋ ฅ์œผ๋กœ ์–ธ๊ธ‰๋๋‹ค. ์ž๋™ ํ™•์žฅ๊ณผ ๊ณ ๊ฐ€์šฉ์„ฑ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ์†Œ๊ทœ๋ชจ ํ…Œ์ŠคํŠธ๋ถ€ํ„ฐ ๊ธ€๋กœ๋ฒŒ ๊ทœ๋ชจ์˜ ๋Œ€๊ทœ๋ชจ ์‹œ๋‚˜๋ฆฌ์˜ค๊นŒ์ง€ ์•ˆ์ •์ ์ธ ํ…Œ์ŠคํŠธ ์‹คํ–‰์ด ๊ฐ€๋Šฅํ•˜๋ฉฐ, ์‹ค์ œ ์šด์˜ ํ™˜๊ฒฝ๊ณผ ์œ ์‚ฌํ•œ ์กฐ๊ฑด์—์„œ ์„ฑ๋Šฅ์„ ๊ฒ€์ฆํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•œ๋‹ค๋Š” ์ ์—์„œ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ๊ณ ๊ฐ์˜ ์š”๊ตฌ๋ฅผ ์ถฉ์กฑํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๊ธฐ๊ฐ€์˜ด์€ ํ‰๊ฐ€ํ–ˆ๋‹ค.

๊ธฐ๊ฐ€์˜ด์€ ์˜คํ”ˆํ…์ŠคํŠธ๊ฐ€ ์ƒ์„ฑํ˜• AI, ๋ฐ๋ธŒ์˜ต์Šค(DevOps) ํ†ตํ•ฉ, ์ง€๋Šฅํ˜• ์ž๋™ํ™”๋ฅผ ์ง€์†์ ์œผ๋กœ ๊ฐ•ํ™”ํ•˜๋ฉฐ ์„ฑ๋Šฅ ์—”์ง€๋‹ˆ์–ด๋ง ์˜์—ญ์—์„œ ํ˜์‹ ์„ ์ด์–ด๊ฐ€๊ณ  ์žˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์˜คํ”ˆํ…์ŠคํŠธ๊ฐ€ ํ…Œ์ŠคํŠธ ์ž๋™ํ™”์™€ ์šด์˜ ํšจ์œจ์„ฑ์„ ๋†’์ด๋Š” ๋™์‹œ์—, ๋น ๋ฅด๊ฒŒ ๋ณ€ํ™”ํ•˜๋Š” ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ๋„ ์‹ ๋ขฐํ•  ์ˆ˜ ์žˆ๋Š” ๋””์ง€ํ„ธ ์‚ฌ์šฉ์ž ๊ฒฝํ—˜์„ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ๋Š” ๊ธฐ๋ฐ˜์„ ๋งˆ๋ จํ•˜๊ณ  ์žˆ๋‹ค๋Š” ๋ถ„์„์ด๋‹ค.

์˜คํ”ˆํ…์ŠคํŠธ๋Š” ์ด๋ฒˆ ํ‰๊ฐ€๋ฅผ ํ†ตํ•ด ํด๋ผ์šฐ๋“œ ์„ฑ๋Šฅ ํ…Œ์ŠคํŠธ ์‹œ์žฅ์—์„œ ๋‹จ์ˆœํ•œ ์†”๋ฃจ์…˜ ์—…์ฒด๋ฅผ ๋„˜์–ด, ์„ฑ๋Šฅ ์—”์ง€๋‹ˆ์–ด๋ง์˜ ๋ฐฉํ–ฅ์„ฑ๊ณผ ๊ธฐ์ค€์„ ์ œ์‹œํ•˜๋Š” ๊ธฐ์—…์œผ๋กœ์„œ์˜ ์ž…์ง€๋ฅผ ๋‹ค์กŒ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.
dl-ciokorea@foundryco.com

How adaptive infrastructure is evolving capabilities at the speed of business

Iโ€™m not normally fond of year-end technology retrospectives, but 2025 was indeed a year of quantum leaps in the art of the possible and it has filled us all with measured optimism paired with some healthy and well-earned skepticism where AI is concerned. When I put architecture in perspective, Iโ€™m inclined to take a longer view of automation in all its variations over a decade. Thatโ€™s why 2025 feels more like a footnote in a long series of events culminating in the perfect storm of opportunities weโ€™ve been contemplating for some time now.

The composable infrastructure revolution

Weโ€™ve been moving toward self-aware, composable infrastructure in architecture for a while now and infrastructure-as-code was merely the first major inflection.

Letโ€™s be honest, the old way of building IT infrastructure is breaking down. As an enterprise architect, the vicious cycle is very familiar. Tying agentic architecture demand-patterns to legacy infrastructure without careful consideration is fraught with peril. The old pattern is really predictable now: You provision systems, maintain them reactively and eventually retire them. Rinse and repeat.

That model is now officially unsustainable in the age of AI. Whatโ€™s taking its place? Composable and intelligent infrastructure that can proactively self-assemble, reconfigure and optimize on the fly to match what the business needs.

For IT leaders, this shift from rigid systems to modular, agent-driven infrastructure is both a breakthrough opportunity and a serious transformation challenge. And the numbers back this up: the global composable infrastructure market sits at USD $8.3 billion in 2025 and is projected to grow at 24.9% annually through 2032.

Whatโ€™s driving this hyper-accelerated growth? Geopolitical disruptions, supply chain chaos and AI advances are reshaping how and where companies operate. Business environments are being driven by reactive and dynamic agentic experiences, transactions and digital partnerships everywhere, all the time. Static infrastructure just canโ€™t deliver that kind of flexibility based on marketing exercises that describe solution offerings as โ€œon-demand,โ€ โ€œutility-based,โ€ โ€œadaptiveโ€ and โ€œcomposable.โ€ These are little more than half-truths.

A 2025 Forrester study commissioned by Microsoft found that 84% of IT leaders want solutions that consolidate edge and cloud operations across systems, sites and teams. As an architect in the consumer goods space, I found that our IT team would produce endless slide decks about composable enterprises ad nauseam, but infrastructure-as-code was the level of actual capability for some time.

Leaders wanted composable architecture that can pull together diverse components without hyperextended interoperability efforts. IBMโ€™s research reinforces this, showing that companies with modular architectures are more agile, more resilient and faster to market โ€” while also reducing the technical debt that slows everyone down.

The problem has been one of capacity and fitness for purpose. Legacy infrastructure and the underlying systems of record simply werenโ€™t designed with agentic AI patterns in mind. My conversations with pan-industry architecture colleagues reflect the same crisis of expectation and resilience around agentic architectures.

Consider McKinseyโ€™s 2025 AI survey that demonstrated 88% of organizations now use AI regularly in at least one business function and 62% are experimenting with AI agents. But most are stuck in pilot mode because their infrastructure canโ€™t scale AI across the business.

If there are any winners in this race, theyโ€™ve broken apart their monolithic systems into modular pieces that AI agents can orchestrate based on whatโ€™s actually happening in real time.

AI agents: The new orchestration layer

So, whatโ€™s driving this shift? Agentic AI โ€” systems that understand business context, figure out optimal configurations and execute complex workflows by pulling together infrastructure components on demand. This isnโ€™t just standard automation following rigid, brittle scripts. Agents reason about what to assemble, how to configure it and when to reconfigure as conditions change.

The adoption curve is steep. BCG and MIT Sloan Management Review found that 35% of organizations already use agentic AI, with another 44% planning to jump in soon. The World Economic Forum reports 82% of executives plan to adopt AI agents within three years. McKinseyโ€™s abovementioned State of AI research further highlights agentic AI as an emerging focus area for enterprise investment and describes AI agents as systems that can plan, take actions and orchestrate multi-step workflows with less human intervention than traditional automation.

As McKinsey puts it: โ€œWeโ€™re entering an era where enterprise productivity is no longer just accelerated by AI โ€” itโ€™s orchestrated by it.โ€ Thatโ€™s a fundamental change in how infrastructure works.

IBM is betting big on this future, stating that โ€œthe future of IT operations is autonomous, policy-driven and hybrid by design.โ€ Theyโ€™re building environments where AI agents can orchestrate everything โ€” public cloud, private infrastructure, on-premises systems, edge deployments โ€” assembling optimal configurations for specific workloads and contexts. The scope of automation ranges from helpful recommendations to closed-loop fixes to fully autonomous optimization.

What composable architecture actually looks like

I recall no shortage of Lego-induced architecture references to composability over the last decade. Sadly, we conflated them with domain services and not how business capabilities and automation could and should inform how the Legos are pieced together to solve problems. Traditional infrastructure comes as tightly integrated stacks โ€” hard to decompose, inflexible and reactive. The new composable model flips this, offering modular building blocks that agents can intelligently assemble and reassemble dynamically based on whatโ€™s needed right now.

Composability demands modularity and responsive automation

The foundation is extreme modularity โ€” breaking monolithic systems into discrete, independently deployable pieces with clean interfaces. Composable infrastructure lets you dynamically assemble and disassemble resources based on application demands, optimizing how pooled resources get allocated and improving overall efficiency.

This goes far beyond physical infrastructure to include services, data pipelines, security policies and workflows. When everything is modular and API-accessible, agents can compose complex solutions from simple building blocks and adapt in real time.

Bringing cloud and edge together

Enterprise organizations are no longer treating cloud and edge as separate worlds requiring manual integration. The new approach treats all infrastructure โ€” from hyperscale data centers to network edge โ€” as a unified resource pool that agents can compose into optimal configurations.

McKinsey identifies edge-cloud convergence as essential for agentic AI: โ€œAgents need real-time data access and low-latency environments. Combining edge compute (for inference and responsiveness) with cloud-scale training and storage is essential.โ€ They further highlight how Hewlett Packard Enterprise (HPE) expanded its GreenLake platform in late 2024 with composable infrastructure hardware for hybrid and AI-driven workloads โ€” modular servers and storage that let enterprises dynamically allocate resources based on real-time demand.

Agents running the show

Even IBM with its storied fixed-infrastructure history is all-in on agentic AI infrastructure capabilities โ€” including agents and Model Context Protocol (MCP) servers โ€” across its portfolio, making infrastructure components discoverable and composable by AI agents. These agents donโ€™t just watch the infrastructure state; they actively orchestrate resources across enterprise data and applications, creating optimal configurations for specific workloads.

Management interfaces across IBM cloud, storage, power and Z platforms are becoming MCP-compatible services โ€” turning infrastructure into building blocks that agents can reason about and orchestrate. Vendor-native agentic management solutions introduced similar AI-driven orchestration enhancements in 2024, letting large enterprises dynamically allocate resources across compute, storage and networking.

Self-aware and self-correcting infrastructure

Instead of manually configuring every component, composable architectures enable intent-based interfaces. You specify business objectives โ€” support 10,000 concurrent users with sub-100ms latency at 99.99% availability โ€” and agents figure out the infrastructure composition to make it happen.

Emerging intelligent infrastructure player Quali describes this as โ€œinfrastructure that understands itselfโ€ โ€” systems where agentic AI doesnโ€™t just demand infrastructure that keeps up, but infrastructure built from composable components that agents can understand and orchestrate.

Getting scale and flexibility in real time

Traditional infrastructure forces a choice: optimize for scale or build for adaptability. As architects, there are clear opposing trade-offs we must navigate successfully: Scale relative to adaptability, investment versus sustaining operations, tight oversight versus autonomy and process refactoring versus process reinvention.

Composable architectures solve this by delivering both. The dual nature of agentic AI โ€” part tool, part human-like โ€” doesnโ€™t fit traditional management frameworks. People are flexible but donโ€™t scale. Tools scale but canโ€™t adapt. Agentic AI on composable infrastructure gives you scalable adaptability โ€” handling massive workloads while continuously reconfiguring for changing contexts.

Self-composability and evolved governance

Agent-orchestrated infrastructure demands governance that balances autonomy with control. The earlier-mentioned MIT Sloan Management Review and BCG study found that most agentic AI leaders anticipate significant changes to governance and decision rights as they adopt agentic AI. They recommend creating governance hubs with enterprise-wide guardrails and dynamic decision rights rather than approving individual AI decisions one by one.

The answer lies in policy-based composition, defining constraints that bound agent decisions without prescribing exact configurations. Within those boundaries, agents compose and recompose infrastructure autonomously.

When AI agents continuously compose and recompose resources, you need governance frameworks that look nothing like traditional change management. A model registry that includes MCP connects different large language models while implementing guardrails for analytics, security, privacy and compliance. This treats AI as an agent whose decisions must be understood, managed and learned from โ€” not as an infallible tool.

Making it happen in 2026

What should IT leaders do? Here are the most critical moves from my perspective.

Redesign work around agents first. Use agentic AIโ€™s capacity to implement scalability and adapt broadly within parameterized governance automation rather than automating isolated tasks. Almost two-thirds of agentic AI leaders expect operating model changes. Build workflows that shift smoothly between efficiency and problem-solving modes.

Rethink roles for human-agent collaboration. Agents are an architectโ€™s new partner. Reposition your role as an architect in the enterprise to adopt and embrace portfolios of AI agents to coordinate workflows, and traditional management layers change. Expect fewer middle management layers, with managers evolving to orchestrate hybrid human-AI teams. Consider dual career paths for generalist orchestrators and AI-augmented specialists.

Keep investments tied to value. Agentic AI leaders anchor investments to value โ€” whether efficiency, innovation, revenue growth or some combination. Agentic systems are evolving from finite function agents to multi-agent collaborators, from narrow to broadly orchestrated tasks and other ecosystems and agents, and from operational to strategic human-mediated partnership.

The bottom line

The companies that will win in the next decade will recognize composability as the foundation of adaptive infrastructure. When every part of the technology stack becomes a modular building block and intelligent agents compose those blocks into optimal configurations based on real-time context, infrastructure becomes a competitive advantage instead of a constraint.

Organizations that understand agentic AIโ€™s dual nature and align their processes, governance, talent and investments accordingly will realize its full business value. My architectโ€™s perspective is that agentic AI will challenge established management approaches and, yes, even convince many of its ability to defy gravity. But with the right strategy and execution, it wonโ€™t just offer empty promises โ€” it will deliver results. Further, our grounded expectations around the capacity of aging infrastructure and legacy demand patterns must guide us in ensuring we make intelligent decisions.

The question isnโ€™t whether to embrace composable, agent-orchestrated infrastructure. Itโ€™s how fast you can decompose monolithic systems, build orchestration capabilities and establish the governance to make it work.

This article was made possible by our partnership with the IASA Chief Architect Forum. The CAFโ€™s purpose is to test, challenge and support the art and science of Business Technology Architecture and its evolution over time, as well as grow the influence and leadership of chief architects both inside and outside the profession. The CAF is a leadership community of the IASA, the leading non-profit professional association for business technology architects.ย 

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œโ€“์œ„์ฆˆ, ํด๋ผ์šฐ๋“œ ๋ณด์•ˆ ํ”Œ๋žซํผ ์—ฐ๊ณ„ ํ˜‘๋ ฅ ์ถ”์ง„

์–‘์‚ฌ๋Š” 19์ผ ์„œ์šธ ์—ญ์‚ผ๋™ ๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ ์—ฐ๋ฝ์‚ฌ๋ฌด์†Œ์—์„œ ํด๋ผ์šฐ๋“œ ๋ณด์•ˆ ๊ฐ•ํ™”๋ฅผ ์œ„ํ•œ ํŒŒํŠธ๋„ˆ์‹ญ์„ ์ฒด๊ฒฐํ–ˆ๋‹ค. ์ฒด๊ฒฐ์‹์—๋Š” ๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ ์—ผ๋™ํ›ˆ ๋Œ€ํ‘œ์™€ ์œ„์ฆˆ ๋Œ€ํ‘œ ๊ฒธ ์ตœ๊ณ ์šด์˜์ฑ…์ž„์ž(COO) ๋‹ฌ๋ฆฌ ๋ผ์ง์„ ๋น„๋กฏํ•œ ์–‘์‚ฌ ์ฃผ์š” ๊ด€๊ณ„์ž๋“ค์ด ์ฐธ์„ํ–ˆ๋‹ค.

์œ„์ฆˆ๋Š” ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ ์ „๋ฐ˜์˜ ๋ณด์•ˆ ์ƒํƒœ๋ฅผ ํ†ตํ•ฉ์ ์œผ๋กœ ๋ถ„์„ํ•˜๊ณ  ์‹œ๊ฐํ™”ํ•˜๋Š” ํด๋ผ์šฐ๋“œ ๋ณด์•ˆ ํ”Œ๋žซํผ์ด๋‹ค. ํด๋ผ์šฐ๋“œ ์ž์‚ฐ, ๊ตฌ์„ฑ ์˜ค๋ฅ˜, ๊ถŒํ•œ, ์„ค์ • ๋“ฑ ๋‹ค์–‘ํ•œ ๋ณด์•ˆ ์ •๋ณด๋ฅผ ๊ด€๊ณ„์™€ ๋งฅ๋ฝ์— ๋”ฐ๋ผ ์—ฐ๊ฒฐํ•ด ๋ถ„์„ํ•˜๋ฉฐ, ์ด๋ฅผ ์‹œํ๋ฆฌํ‹ฐ ๊ทธ๋ž˜ํ”„(Security Graph) ๊ธฐ๋ฐ˜์œผ๋กœ ์ œ๊ณตํ•œ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์ž ์žฌ์ ์ธ ์œ„ํ—˜ ์š”์†Œ์™€ ๊ณต๊ฒฉ ๊ฒฝ๋กœ๋ฅผ ํŒŒ์•…ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•ด, ๋ณต์žกํ•œ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์˜ ๋ณด์•ˆ ํ˜„ํ™ฉ์„ ๋ณด๋‹ค ์ฒด๊ณ„์ ์ธ ํ˜•ํƒœ๋กœ ์ œ๊ณต๋œ๋‹ค.

์–‘์‚ฌ๋Š” ์œ„์ฆˆ์˜ ํด๋ผ์šฐ๋“œ ๋ณด์•ˆ ํ”Œ๋žซํผ๊ณผ ๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ์˜ ํด๋ผ์šฐ๋“œ ์šด์˜ยท๊ตฌ์ถ• ์—ญ๋Ÿ‰์„ ์—ฐ๊ณ„ํ•ด, ๊ณ ๊ฐ์ด ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ ๋ณด์•ˆ ๋ฆฌ์Šคํฌ๋ฅผ ์‹๋ณ„ํ•˜๊ณ  ์šฐ์„ ์ˆœ์œ„๋ฅผ ์„ค์ •ํ•ด ๋Œ€์‘ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•  ๊ณ„ํš์ด๋‹ค. ๋ณด๋„์ž๋ฃŒ์— ๋”ฐ๋ฅด๋ฉด, ํŠนํžˆ ๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ์˜ ๋ณด์•ˆ์‚ฌ์—… ์กฐ์ง์ธ HALO๋Š” ์œ„์ฆˆ์˜ ์‹œํ๋ฆฌํ‹ฐ ๊ทธ๋ž˜ํ”„ ๊ธฐ๋ฐ˜ ๋ถ„์„ ์ •๋ณด๋ฅผ ๊ตญ๋‚ด ๊ณ ๊ฐ ํ™˜๊ฒฝ์— ๋งž๊ฒŒ ์ ์šฉํ•˜๋ฉฐ, ์ด๋ฅผ ํด๋ผ์šฐ๋“œ ์šด์˜ ๊ณผ์ •์—์„œ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•  ์˜ˆ์ •์ด๋‹ค.

๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ๋Š” ์ด๋Ÿฌํ•œ ๊ธฐ์ˆ ๋กœ โ–ฒํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ ์ „๋ฐ˜์˜ ๋ณด์•ˆ ์ƒํƒœ ํŒŒ์•… โ–ฒ๋ณด์•ˆ ์ง„๋‹จ ๊ฒฐ๊ณผ์˜ ์šด์˜ ํ™˜๊ฒฝ ์ ์šฉ โ–ฒ๋ณด์•ˆ ์ ๊ฒ€ ๊ฒฐ๊ณผ์˜ ์šด์˜ ํŒ๋‹จ ๋ฐ ๋ณด์•ˆ ๊ฐœ์„  ํ™œ๋™ ๋ฐ˜์˜์— ํ™œ์šฉ๋  ์ˆ˜ ์žˆ์„ ๊ฒƒ์œผ๋กœ ๊ธฐ๋Œ€ํ–ˆ๋‹ค.

์ด์™ธ์—๋„ ์–‘์‚ฌ๋Š” ๊ธฐ์ˆ  ํ˜‘์—…๊ณผ ๊ณต๋™ ์‹œ์žฅ ํ™œ๋™์„ ํ™•๋Œ€ํ•˜๊ณ , ๊ตญ๋‚ด ๊ธฐ์—… ๋ณด์•ˆ ์ฑ…์ž„์ž๋ฅผ ๋Œ€์ƒ์œผ๋กœ ํ•œ ์„ธ๋ฏธ๋‚˜ ๋“ฑ ์‹ค๋ฌด ์ค‘์‹ฌ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜ ํ”„๋กœ๊ทธ๋žจ๋„ ์ถ”์ง„ํ•  ์˜ˆ์ •์ด๋‹ค.

์—ผ๋™ํ›ˆ ๋Œ€ํ‘œ๋Š” โ€œํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ๋Š” ๋ณด์•ˆ ์†”๋ฃจ์…˜ ์ž์ฒด๋ณด๋‹ค ๊ทธ ์†”๋ฃจ์…˜์ด ์‹ค์ œ ์šด์˜ ๊ณผ์ •์—์„œ ์–ด๋–ป๊ฒŒ ํ™œ์šฉ๋˜๋Š”์ง€๊ฐ€ ๋”์šฑ ์ค‘์š”ํ•ด์ง€๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ๋Š” ํด๋ผ์šฐ๋“œ ์•„ํ‚คํ…์ฒ˜์™€ ์šด์˜ ์ „๋ฐ˜์— ๋Œ€ํ•œ ์ดํ•ด๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ, ์œ„์ฆˆ์˜ ๋ณด์•ˆ ํ”Œ๋žซํผ์ด ๊ณ ๊ฐ ํ™˜๊ฒฝ์—์„œ ํšจ๊ณผ์ ์œผ๋กœ ์ž‘๋™ํ•˜๋„๋ก ์„ค๊ณ„ยท์šด์˜์„ ์ง€์›ํ•ด ๋ณด์•ˆ ํ™˜๊ฒฝ ๊ฐœ์„ ์— ๊ธฐ์—ฌํ•˜๊ฒ ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋ผ์ง์€ โ€œํด๋ผ์šฐ๋“œ์™€ AI๋Š” ๊ธฐ์—…์˜ ํ˜์‹  ๋ฐฉ์‹์„ ๋ณ€ํ™”์‹œํ‚ค๊ณ  ์žˆ์œผ๋ฉฐ, ์œ„์ฆˆ๋Š” ์ด๋ฅผ ๋ณด๋‹ค ์•ˆ์ „ํ•˜๊ฒŒ ๋„์ž…ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ๋ฉ”๊ฐ€์กดํด๋ผ์šฐ๋“œ์™€ ํŒŒํŠธ๋„ˆ์‹ญ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ˜‘๋ ฅ์„ ๊ฐ•ํ™”ํ•ด ํ•œ๊ตญ ๊ณ ๊ฐ๋“ค์ด ํด๋ผ์šฐ๋“œ ์„ฑ์žฅ ๊ณผ์ •์—์„œ๋„ ๋†’์€ ์ˆ˜์ค€์˜ ๋ณด์•ˆ์„ ์œ ์ง€ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•˜๊ฒŒ ๋˜์–ด ๊ธฐ๋Œ€๊ฐ€ ํฌ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค.
dl-ciokorea@foundryco.com

Vega Cloud enters receivership, with millions in debt, in surprise turn for Spokane tech standout

Vega Cloudโ€™s technology helps companies track and manage their cloud spending. (Vega Cloud Images, GeekWire Illustration)

Vega Cloud, a Spokane-area tech startup that makes software to help companies manage their cloud spending, has been placed into the hands of a receiver after declaring it could no longer pay its debts.

Among those debts: nearly $830,000 owed to cloud giant Amazon Web Services.ย 

Vega Cloud, founded in 2018 and based in Liberty Lake, Wash., had raised $12.2 million and reached about $7 million in annual revenue as of 2023, according to PitchBook data. It had also cracked the GeekWire 200 โ€” ranking #181 in the most recent quarterly update of our Pacific Northwest startup index.

What brought Vega Cloud to this point isnโ€™t clear. Responding to our email inquiry this weekend, co-founder and CEO Kris Bliesner said the company is going through a restructuring via receivership, and said he wished he could say more about the situation.

The company had less than $17,000 in the bank when it was placed into receivership Thursday, Jan. 15, in King County Superior Court in Seattle, the filing shows. It employed about 35 people as of earlier this month, down from about 65 two years ago, according to LinkedIn.ย 

Receivership is a state-level process often used as an alternative to bankruptcy. In this case, Vega Cloud executed whatโ€™s known as an Assignment for the Benefit of Creditors, which puts a neutral party in charge of the company, pauses creditor collections, and places decisions about asset sales and payments under court supervision.

Sometimes those assets sell mostly intact, allowing new investors to give a business another try. But at this point, itโ€™s not yet clear what will happen to the companyโ€™s employees or product.

Past ambitions for an IPO

In a March 2024 interview for GeekWireโ€™s special series on Spokane, Bliesner described Vega Cloudโ€™s trajectory in optimistic terms, saying the company was planning a $20 million to $30 million funding round and eyeing the public markets.

โ€œWeโ€™re trying to push the envelope at Vega to maybe do the IPO route,โ€ Bliesner said at the time. โ€œWe think thatโ€™s a viable thing for us.โ€

Vega Cloud operates in the sector known as FinOps, short for financial operations, helping companies get a handle on their cloud spending by bringing together finance and technical teams to track costs and avoid waste.

This is becoming more and more important as businesses pour money into cloud computing, often without realizing how much theyโ€™re spending on unused resources. Vega Cloud focused specifically on helping mid-sized companies manage spending across AWS, Azure, and Google Cloud, using automated tools to spot problems and recommend fixes.ย 

In the tight-knit Spokane tech community, Vega Cloud has been seen as a startup with the potential to make it big. We took note of the company in 2022, when it raised $9 million.

Investor and entrepreneur Martin Tobias, a longtime fixture in Pacific Northwest enterprise tech, invested in Vega Cloud shortly after moving from Seattle to Spokane during the pandemic. He told us in early 2024 that it would probably be one of his most successful investments.ย 

Tobias said Bliesner was exactly the kind of founder he looks for: someone with deep experience in a market who had tried to solve something one way, realized it wasnโ€™t going to scale, and came up with a better solution.

โ€œHe took a new approach to an old problem,โ€ Tobias said at the time.ย 

Bliesner previously co-founded cloud migration startup 2nd Watch, which raised about $56 million before selling a majority interest to Singapore-based investor ST Telemedia.

Financial details from the filing

Vega Cloudโ€™s court filings give an inside look at the privately held business.

First, the company had real customers and revenue. The filings list contracts with companies including Paramount, Hearst, Deloitte, Molina Healthcare, John Wiley & Sons, and Cal Poly, among others. It lists roughly $264,000 in accounts receivable.

The largest secured creditor is Sun Mountain Private Credit Fund I, owed $3.5 million. That debt is backed by Vega Cloudโ€™s intellectual property โ€” its software, patents, trademarks, and domain names. Any proceeds from a sale of those assets would go first to that lender.

In addition to the roughly $830,000 owed to AWS, the court records show convertible promissory notes totaling about $2.5 million that were issued to investors throughout 2025.

The records list current and former employees who are owed unpaid commissions, bonuses, and expense reimbursements, with some bonus obligations dating back to 2023. The company also owes payroll and withholding taxes to the IRS and multiple state tax agencies.

Bliesner is the largest shareholder at about 30%. Other significant investors include Album Ventures (10%), Cowles Company (3%), Rudeen & Company (3%), Kick-Start III and IV (combined 4%), Tacoma Venture Fund (1.5%), and Pitbull Ventures (1%).ย 

The shareholder list also includes Voyager Capital, Alliance of Angels, Incisive Ventures, and Morning Star Foundation, along with dozens of individual investors.

Under court supervision, the receiver can now take possession of Vega Cloudโ€™s assets and records, secure its bank accounts and data, evaluate and sell assets such as intellectual property, collect remaining receivables, and distribute proceeds to creditors in priority order.

The filings do not include a timeline for asset sales or any plan for the business to continue operating. Those details typically emerge later through receiver reports.

โ€œ์œ„์น˜ ๊ด€๊ณ„์—†์ด ์ฃผ๊ถŒ ๊ตฌํ˜„ํ•œ๋‹คโ€ยทยทยทIBM, ์ƒˆ๋กœ์šด ํ•ด๋ฒ•์œผ๋กœ โ€˜์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ดโ€™ ๊ณต๊ฐœ

IBM์€ ๊ธฐ์—… ๋ฐ ์ •๋ถ€๊ฐ€ ํด๋ผ์šฐ๋“œ ์—…์ฒด์˜ ๋ฐ์ดํ„ฐ์„ผํ„ฐ ์œ„์น˜์— ์˜์กดํ•˜์ง€ ์•Š๊ณ ๋„ ์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ ๋ฐฐํฌ์— ๋Œ€ํ•œ ์šด์˜ ํ†ต์ œ๊ถŒ์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋„๋ก ์„ค๊ณ„๋œ ์†Œํ”„ํŠธ์›จ์–ด ์Šคํƒ โ€˜์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด(Sovereign Core)โ€™๋ฅผ ์ถœ์‹œํ–ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด CIO๊ฐ€ ๊ฐ•ํ™”๋˜๋Š” ๊ทœ์ œ ์‹ฌ์‚ฌ์— ๋Œ€์‘ํ•˜๊ณ  ์ปดํ”Œ๋ผ์ด์–ธ์Šค๋ฅผ ์ž๋™ํ™”ํ•˜๋ฉฐ, ๋ฐ์ดํ„ฐ์˜ ์—„๊ฒฉํ•œ ์œ„์น˜ ์กฐ๊ฑด ์•„๋ž˜์—์„œ ๋ฏผ๊ฐํ•œ AI ์›Œํฌ๋กœ๋“œ๋ฅผ ์‹ค์ œ ์šด์˜ ํ™˜๊ฒฝ์— ๋ฐฐ์น˜ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•˜๋Š” ๊ฒƒ์„ ๋ชฉํ‘œ๋กœ ํ•˜๊ณ  ์žˆ๋‹ค.

์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ๋Š” ์ผ๋ฐ˜์ ์œผ๋กœ ํด๋ผ์šฐ๋“œ์˜ ํšจ์œจ์„ฑ์„ ํ™œ์šฉํ•˜๋ฉด์„œ๋„ ๋ฐ์ดํ„ฐ์™€ IT ์šด์˜์— ๋Œ€ํ•œ ํ†ต์ œ๊ถŒ์„ ์œ ์ง€ํ•˜๋Š” ๋ฐ ์ดˆ์ ์„ ๋งž์ถ˜๋‹ค. ์ด๋Š” ๋ฐ์ดํ„ฐ ์œ„์น˜ ๊ทœ์ œ์™€ ๊ฐ™์€ ํ˜„์ง€ ๋ฒ•๊ทœ๋ฅผ ์ค€์ˆ˜ํ•˜๋Š” ๋™์‹œ์—, ๋ฐ์ดํ„ฐ์™€ ์šด์˜, ๋ณด์•ˆ์— ๋Œ€ํ•ด ๊ตญ๊ฐ€ ๋˜๋Š” ์กฐ์ง ์ฐจ์›์˜ ์™„์ „ํ•œ ํ†ต์ œ๋ฅผ ๋ณด์žฅํ•˜๊ธฐ ์œ„ํ•ด ๋Œ€๋ถ€๋ถ„ ํŠน์ • ์ง€์—ญ์— ๊ตฌ์ถ•๋œ๋‹ค. ์ด์ƒ์ ์œผ๋กœ๋Š” ๊ฒฉ๋ฆฌ๋œ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ ์šด์˜๋˜๋Š” IT ์ธํ”„๋ผ๋ฅผ ์˜๋ฏธํ•œ๋‹ค.

๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋‚˜ ๊ตฌ๊ธ€์˜ ์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ๊ฐ€ ์ „์šฉ ๋ฐ์ดํ„ฐ์„ผํ„ฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์„ค๊ณ„๋˜๋Š” ๊ฒƒ๊ณผ ๋‹ฌ๋ฆฌ, IBM์€ ๊ธฐ์—…์ด๋‚˜ ์ •๋ถ€๊ฐ€ ๋ฐฐํฌํ•˜๋ ค๋Š” ๋ชจ๋“  ์†Œํ”„ํŠธ์›จ์–ด์™€ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์— ์ฃผ๊ถŒ์„ ๊ธฐ๋ณธ์ ์œผ๋กœ ํƒ‘์žฌํ•˜๊ฒ ๋‹ค๋Š” ์ž…์žฅ์ด๋‹ค. IBM์€ ์˜ค๋Š” 2์›” ๊ธฐ์ˆ  ํ”„๋ฆฌ๋ทฐ ๊ณต๊ฐœ๊ฐ€ ์˜ˆ์ •๋œ ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๋ฅผ ํ†ตํ•ด, ๊ณ ๊ฐ์ด ์ž์ฒด ํ•˜๋“œ์›จ์–ด๋Š” ๋ฌผ๋ก  ์ง€์—ญ ํด๋ผ์šฐ๋“œ ์—…์ฒด๋‚˜ ๋‹ค๋ฅธ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ๋„ ์›Œํฌ๋กœ๋“œ๋ฅผ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

ํ“จ์ฒ˜๋Ÿผ ๊ทธ๋ฃน(Futurum Group)์˜ CIO ์‹ค๋ฌด ์ฑ…์ž„์ž ๋””์˜จ ํžŒ์น˜ํด๋ฆฌํ”„๋Š” โ€œ์ด๋Š” ์ „ํ†ต์ ์ธ ์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ๋ผ๊ธฐ๋ณด๋‹ค๋Š”, ๊ฐ ์กฐ์ง์ด ์ž์ฒด์ ์œผ๋กœ ํด๋ผ์šฐ๋“œ๋ฅผ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•˜๋Š” ์†Œํ”„ํŠธ์›จ์–ด ์Šคํƒ์— ๊ฐ€๊น๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๊ฐ€ ์˜จํ”„๋ ˆ๋ฏธ์Šค ๋ฐ์ดํ„ฐ์„ผํ„ฐ, ์ง€์—ญ ๋‚ด์—์„œ ์ง€์›๋˜๋Š” ํด๋ผ์šฐ๋“œ ์ธํ”„๋ผ, IT ์„œ๋น„์Šค ์—…์ฒด๋ฅผ ํ†ตํ•œ ํ™˜๊ฒฝ ๋“ฑ ๋‹ค์–‘ํ•œ ์šด์˜ ํ™˜๊ฒฝ์—์„œ ํ™œ์šฉ๋  ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

๋ฒค๋” ์ข…์†์„ฑ ์ œ๊ฑฐ

๋ถ„์„๊ฐ€๋“ค์€ ์ด๋Ÿฌํ•œ ์ ‘๊ทผ ๋ฐฉ์‹์ด ์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ ๊ด€๋ฆฌ ๋ฐฉ์‹์„ ์žฌ์ •์˜ํ•˜๊ณ , ๋ฒค๋” ์ข…์†์„ฑ์„ ํ”ผํ•˜๋Š” ๋ฐ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ๋‹ค๊ณ  ์ง„๋‹จํ–ˆ๋‹ค.

ํžŒ์น˜ํด๋ฆฌํ”„๋Š” ๊ธฐ์กด ์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ๋Š” ํด๋ผ์šฐ๋“œ ์—…์ฒด๊ฐ€ ์—…๋ฐ์ดํŠธ๋‚˜ ์ ‘๊ทผ ๊ถŒํ•œ๊ณผ ๊ฐ™์€ ํ•ต์‹ฌ ์šด์˜ ์š”์†Œ๋ฅผ ๊ณ„์† ํ†ต์ œํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค. ์ด๋กœ ์ธํ•ด ๊ทœ์ œ ๋ฆฌ์Šคํฌ๊ฐ€ ์ปค์งˆ ๋ฟ ์•„๋‹ˆ๋ผ, ๊ณ ๊ฐ์ด ํŠน์ • ์—…์ฒด์˜ ์•„ํ‚คํ…์ฒ˜์™€ API, ์ปดํ”Œ๋ผ์ด์–ธ์Šค ๋„๊ตฌ์— ์ข…์†๋˜๋Š” ๊ตฌ์กฐ๊ฐ€ ํ˜•์„ฑ๋  ์ˆ˜ ์žˆ๋‹ค๋Š” ๊ฒƒ์ด๋‹ค.

๋˜ํ•œ ์›Œํฌ๋กœ๋“œ๋ฅผ ๋‹ค๋ฅธ ํ™˜๊ฒฝ์œผ๋กœ ์ด์ „ํ•  ๊ฒฝ์šฐ, ๊ธฐ์กด ์—…์ฒด์˜ ์‹ ์› ๊ด€๋ฆฌ ์ฒด๊ณ„์™€ ์•”ํ˜ธํ™” ํ‚ค, ๊ฐ์‚ฌ ์ถ”์  ์ •๋ณด๊ฐ€ ๋งค๋„๋Ÿฝ๊ฒŒ ์ด์ „๋˜์ง€ ์•Š๋Š” ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋‹ค. ํžŒ์น˜ํด๋ฆฌํ”„๋Š” ์ด๋กœ ์ธํ•ด CIO๊ฐ€ ์ƒˆ๋กœ์šด ํ™˜๊ฒฝ์—์„œ๋„ ๊ทœ์ œ ์š”๊ฑด์„ ์ถฉ์กฑํ•˜๊ธฐ ์œ„ํ•ด ๊ฑฐ๋ฒ„๋„Œ์Šค ์ฒด๊ณ„๋ฅผ ๋‹ค์‹œ ๊ตฌ์ถ•ํ•ด์•ผ ํ•˜๋Š” ๋ถ€๋‹ด์„ ๋– ์•ˆ๊ฒŒ ๋œ๋‹ค๊ณ  ์ง€์ ํ–ˆ๋‹ค.

๋ฐ˜๋ฉด IBM์˜ ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๋Š” ์•”ํ˜ธํ™” ํ‚ค์™€ ์‹ ์› ๊ด€๋ฆฌ, ์šด์˜ ๊ถŒํ•œ์„ ๊ฐ ์กฐ์ง์˜ ๊ด€ํ•  ์˜์—ญ ์•ˆ์— ์œ ์ง€ํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•จ์œผ๋กœ์จ CIO์—๊ฒŒ ๋” ๋งŽ์€ ํ†ต์ œ๊ถŒ์„ ๋ถ€์—ฌํ•  ์ˆ˜ ์žˆ๋‹ค. ์ด๋Ÿฐ ๊ตฌ์กฐ๋กœ ์ธํ•ด CIO๋Š” ๊ฑฐ๋ฒ„๋„Œ์Šค ์ฒด๊ณ„๋ฅผ ๋‹ค์‹œ ๊ตฌ์ถ•ํ•˜์ง€ ์•Š๊ณ ๋„ ํด๋ผ์šฐ๋“œ ์—…์ฒด๋ฅผ ์ „ํ™˜ํ•  ์ˆ˜ ์žˆ๋‹ค.

ํ•˜์ดํผํ”„๋ ˆ์ž„ ๋ฆฌ์„œ์น˜(HyperFRAME Research)์˜ AI ์Šคํƒ ์ด๊ด„์ธ ์Šคํ…ŒํŒŒ๋‹ˆ ์›”ํ„ฐ๋Š” ๊ทœ์ œ ๊ธฐ๊ด€ ์ฃผ๋„์˜ ๊ฐ์‚ฌ๊ฐ€ ์ ์  ๋” ๋นˆ๋ฒˆํ•ด์ง€๊ณ , ์š”๊ตฌ ์ˆ˜์ค€๋„ ๊ฐ•ํ™”๋˜๊ณ  ์žˆ๋‹ค๊ณ  ์ง„๋‹จํ–ˆ๋‹ค. ํŠนํžˆ ์œ ๋Ÿฝ์—ฐํ•ฉ(EU)์˜ ๊ทœ์ œ ๋‹น๊ตญ์€ ๊ธฐ์—…์˜ ๊ทœ์ œ ์ค€์ˆ˜ ์•ฝ์†๋งŒ์œผ๋กœ๋Š” ์ถฉ๋ถ„ํ•˜์ง€ ์•Š๋‹ค๊ณ  ๋ณด๊ณ , ์‹ค์ œ ์ค€์ˆ˜ ์—ฌ๋ถ€๋ฅผ ์ž…์ฆํ•  ์ˆ˜ ์žˆ๋Š” ์ฆ๊ฑฐ์™€ ๊ฐ์‚ฌ ๊ธฐ๋ก, ์ƒ์‹œ์ ์ธ ์ปดํ”Œ๋ผ์ด์–ธ์Šค ๋ณด๊ณ ๋ฅผ ์š”๊ตฌํ•˜๊ณ  ์žˆ๋‹ค.

ํžŒ์น˜ํด๋ฆฌํ”„๋Š” ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๊ฐ€ ์ž๋™ํ™”๋œ ์ฆ๊ฑฐ ์ˆ˜์ง‘๊ณผ ์ง€์†์ ์ธ ๋ชจ๋‹ˆํ„ฐ๋ง์„ ํ†ตํ•ด ์ด๋Ÿฐ ์š”๊ตฌ์— ๋Œ€์‘ํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์€ํ–‰๊ณผ ์ •๋ถ€ ๊ธฐ๊ด€, ๋ฐฉ์œ„ ์‚ฐ์—…๊ณผ ์—ฐ๊ด€๋œ ๋ถ„์•ผ์—์„œ ๋ฐœ์ƒํ•˜๋Š” ์šด์˜ ๋ถ€๋‹ด์„ ์ค„์ด๋Š” ๋ฐ๋„ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ๋‹ค๊ณ  ํ‰๊ฐ€ํ–ˆ๋‹ค.

์†Œ๋ฒ„๋ฆฐ AI ํŒŒ์ผ๋Ÿฟ์˜ ์‹ค์ œ ๋ฐฐํฌ ์ง€์›

๋ถ„์„๊ฐ€๋“ค์€ ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๊ฐ€ ๊ธฐ์—…์˜ AI ํŒŒ์ผ๋Ÿฟ ํ”„๋กœ๊ทธ๋žจ์„ ์‹ค์ œ ์šด์˜ ํ™˜๊ฒฝ์— ๋ฐฐํฌํ•˜๋Š” ๋ฐ๋„ ํž˜์„ ์‹ค์–ด์ค„ ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ดค๋‹ค. ํŠนํžˆ ์—„๊ฒฉํ•œ ๋ฐ์ดํ„ฐ ์œ„์น˜ ์กฐ๊ฑด๊ณผ ์ปดํ”Œ๋ผ์ด์–ธ์Šค ํ†ต์ œ๊ฐ€ ์š”๊ตฌ๋˜๋Š” AI ํ”„๋กœ์ ํŠธ์—์„œ ํšจ๊ณผ๊ฐ€ ํด ๊ฒƒ์ด๋ผ๋Š” ๋ถ„์„์ด๋‹ค.

HFS ๋ฆฌ์„œ์น˜(HFS Research)์˜ CEO ํ•„ ํผ์ŠˆํŠธ๋Š” ๋Œ€๋ถ€๋ถ„์˜ ๊ธฐ์—…๊ณผ ์กฐ์ง์ด ์ž์ฒด ๋ฐ์ดํ„ฐ๋ฅผ ๋ฒ”์šฉ AI ๋ชจ๋ธ์— ์ „๋‹ฌํ•˜๋Š” ๋ฐ ์—ฌ์ „ํžˆ ๋ถ€๋‹ด์„ ๋А๋ผ๊ณ  ์žˆ๋‹ค๊ณ  ์ง„๋‹จํ•˜๋ฉด์„œ, ๋™์‹œ์— GPU ๊ธฐ๋ฐ˜ ์ถ”๋ก ์„ ์™„์ „ํžˆ ์ž์ฒด ์ฃผ๊ถŒ ๊ฒฝ๊ณ„ ์•ˆ์—์„œ๋งŒ ์‹คํ–‰ํ•˜๋Š” ๊ฒƒ๋„ ํ˜„์‹ค์ ์œผ๋กœ ์ œ์•ฝ์ด ๋งŽ์€ ์ƒํ™ฉ์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ด์— ๋น„ํ•ด ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด์˜ ๊ธฐ๋Šฅ๊ณผ ์—ญ๋Ÿ‰์€ ๊ธฐ์—… ๋ฐ ์ •๋ถ€ ์กฐ์ง์ด ๋‚ด๋ถ€ ํ™˜๊ฒฝ์—์„œ AI ์ถ”๋ก ์„ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•œ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์ฒ˜๋ฆฌ๋˜๋Š” ๋ฐ์ดํ„ฐ๋ฟ ์•„๋‹ˆ๋ผ AI ๋ชจ๋ธ ์ž์ฒด๋„ ์ฃผ๊ถŒ ์š”๊ตฌ์‚ฌํ•ญ์„ ์ถฉ์กฑํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ๊ฒฐ๊ณผ์ ์œผ๋กœ CIO๊ฐ€ ์ฃผ๊ถŒ์„ ํ™•๋ณดํ•˜๋ฉด์„œ AI๋ฅผ ํŒŒ์ผ๋Ÿฟ ๋‹จ๊ณ„์—์„œ ์šด์˜ ๋‹จ๊ณ„๋กœ ์˜ฎ๊ธธ ์ˆ˜ ์žˆ๋Š” ๊ธฐ๋ฐ˜์„ ์ œ๊ณตํ•œ๋‹ค๊ณ  ํผ์ŠˆํŠธ๋Š” ์„ค๋ช…ํ–ˆ๋‹ค.

์‹œ์žฅ ํ™˜๊ฒฝ์˜ ๋ณ€ํ™”

์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๋Š” IBM์ด ํ–ฅํ›„ AI ๊ทœ์ œ ๊ฐ•ํ™” ํ๋ฆ„์„ ์—ผ๋‘์— ๋‘๊ณ  ์†Œ๋ฒ„๋ฆฐ ํด๋ผ์šฐ๋“œ ์‹œ์žฅ ๊ณต๋žต์„ ๋ณธ๊ฒฉํ™”ํ•˜๋ ค๋Š” ์ „๋žต์œผ๋กœ ํ’€์ด๋œ๋‹ค. ๋™์‹œ์— ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ์™€ AWS, ๊ตฌ๊ธ€ ๋“ฑ ์ฃผ์š” ํด๋ผ์šฐ๋“œ ์—…์ฒด๋ณด๋‹ค ํ•œ๋ฐœ ์•ž์„œ ์ฃผ๋„๊ถŒ์„ ์žก์œผ๋ ค๋Š” ์˜๋„๋„ ๋‹ด๊ฒจ ์žˆ๋‹ค.

ํžŒ์น˜ํด๋ฆฌํ”„๋Š” โ€œ์œ ๋Ÿฝ์ด ๊ทœ์ œ๋ฅผ ๊ฐ•ํ™”ํ•˜๊ณ  ์•„์‹œ์•„ํƒœํ‰์–‘(APAC) ์ง€์—ญ๋„ ์ด๋ฅผ ๋’ค๋”ฐ๋ฅด๋Š” ์ƒํ™ฉ์—์„œ, IBM์€ ์ฃผ๊ถŒ ๋ฌธ์ œ๊ฐ€ ๊ธฐ์—…์˜ AI ๋„์ž… ์—ฌ๋ถ€๋ฅผ ๊ฐ€๋ฅด๋Š” ํ•ต์‹ฌ ์š”์ธ์ด ๋  ๊ฒƒ์œผ๋กœ ๋ณด๊ณ  ์žˆ๋‹ค. ์ผ๋ถ€ ๊ธฐ์—…์—์„œ๋Š” ๋น„์šฉ์ด๋‚˜ ์„ฑ๋Šฅ๋ณด๋‹ค๋„ ํ›จ์”ฌ ๋” ์ค‘์š”ํ•œ ์š”์†Œ๊ฐ€ ๋  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

ํŠนํžˆ EU๋Š” ์ฃผ์š” ํด๋ผ์šฐ๋“œ ์—…์ฒด ๋Œ€๋ถ€๋ถ„์ด ๋ฏธ๊ตญ์— ๋ณธ์‚ฌ๋ฅผ ๋‘๊ณ  ์žˆ๋‹ค๋Š” ์ ์—์„œ, ์™ธ๊ตญ ๊ธฐ์—…์ด ๋ฐ์ดํ„ฐ์— ์ ‘๊ทผํ•˜๊ฑฐ๋‚˜ ํ•ต์‹ฌ IT ์‹œ์Šคํ…œ์„ ํ†ต์ œํ•˜๋Š” ๊ฒƒ์„ ์—„๊ฒฉํ•˜๊ฒŒ ๊ทœ์ œํ•˜๊ณ  ์žˆ๋‹ค.

EU ๊ทœ์ œ๋ฅผ ์ถฉ์กฑํ•˜๊ธฐ ์œ„ํ•ด ํด๋ผ์šฐ๋“œ ์—…์ฒด๋Š” ๋ณดํ†ต ์ง€์—ญ ํ†ตํ•ฉ ์—…์ฒด๋‚˜ ๊ด€๋ฆฌํ˜• ์„œ๋น„์Šค ์—…์ฒด์™€ ํ˜‘๋ ฅํ•œ๋‹ค. ๋‹ค๋งŒ ํžŒ์น˜ํด๋ฆฌํ”„์— ๋”ฐ๋ฅด๋ฉด, ์ด ๊ฒฝ์šฐ์—๋„ ๊ธฐ๋ณธ ํ”Œ๋žซํผ์— ๋Œ€ํ•œ ์šด์˜ ํ†ต์ œ๊ถŒ์€ ํด๋ผ์šฐ๋“œ ์—…์ฒด๊ฐ€ ์œ ์ง€ํ•˜๊ณ , ํŒŒํŠธ๋„ˆ๋Š” ๊ทธ ์œ„์—์„œ ์„œ๋น„์Šค ๊ตฌ์ถ•๊ณผ ์šด์˜์„ ๋งก๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋Œ€๋ถ€๋ถ„์ด๋‹ค.

IBM์˜ ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด๋Š” ํŒŒํŠธ๋„ˆ๊ฐ€ ๊ณ ๊ฐ์„ ๋Œ€์‹ ํ•ด ์ „์ฒด ํ™˜๊ฒฝ์„ ์ง์ ‘ ์šด์˜ํ•  ์ˆ˜ ์žˆ๊ณ , IBM์€ ์šด์˜ ๊ณผ์ •์— ์ „ํ˜€ ๊ฐœ์ž…ํ•˜์ง€ ์•Š๋Š” ๊ตฌ์กฐ๋‹ค. ํžŒ์น˜ํด๋ฆฌํ”„๋Š” ์ด๋Ÿฌํ•œ ์ ‘๊ทผ์ด ๊ทœ์ œ ์ค€์ˆ˜ ์ธก๋ฉด์—์„œ ๋” ๋†’์€ ์‹ ๋ขฐ์„ฑ์„ ์ œ๊ณตํ•œ๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

์ด์™€ ๊ด€๋ จํ•ด IBM์€ ๋…์ผ์˜ ์ปดํ“จํƒ€์„ผํ„ฐ(Computacenter) ๋ฐ ์œ ๋Ÿฝ ์ง€์—ญ์„ ์‹œ์ž‘์œผ๋กœ ์ „ ์„ธ๊ณ„ IT ์„œ๋น„์Šค ์—…์ฒด์™€ ํ˜‘๋ ฅ์„ ํ™•๋Œ€ํ•  ๊ณ„ํš์ด๋ผ๊ณ  ๋ฐํ˜”๋‹ค. IBM์€ ์†Œ๋ฒ„๋ฆฐ ์ฝ”์–ด์— ์ถ”๊ฐ€ ๊ธฐ๋Šฅ์„ ๋”ํ•ด 2026๋…„ ์ค‘๋ฐ˜ ์ •์‹ ์ถœ์‹œํ•  ๊ณ„ํš์ด๋‹ค.
dl-ciokorea@foundryco.com

โ€œAI ๋ฐ์ดํ„ฐ์„ผํ„ฐ ํ™•์žฅ์˜ ์ตœ๋Œ€ ๊ฑธ๋ฆผ๋Œ์€ ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ ์ง€์—ฐโ€ ๊ตฌ๊ธ€

๊ตฌ๊ธ€์€ ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ ์ง€์—ฐ์ด ๋ฐ์ดํ„ฐ์„ผํ„ฐ์— ์ „๋ ฅ์„ ๊ณต๊ธ‰ํ•˜๋Š” ๋ฐ ๊ฐ€์žฅ ํฐ ์žฅ์• ๋ฌผ์ด ๋˜๊ณ  ์žˆ๋‹ค๋ฉฐ, ์ง€์—ฐ ๊ธฐ๊ฐ„์ด ์ตœ๋Œ€ 12๋…„์— ์ด๋ฅผ ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋‚ด๋‹ค๋ดค๋‹ค. AI ์›Œํฌ๋กœ๋“œ๊ฐ€ ๊ธ‰์ฆํ•˜๋Š” ๊ฐ€์šด๋ฐ, ๊ตฌ๊ธ€์€ ํด๋ผ์šฐ๋“œ ์šฉ๋Ÿ‰ ํ™•์žฅ์„ ์ œํ•œํ•˜๋Š” ์ธํ”„๋ผ ์œ„๊ธฐ๊ฐ€ ๋ณธ๊ฒฉํ™”๋˜๊ณ  ์žˆ๋‹ค๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค.

๊ตฌ๊ธ€์˜ ๊ธ€๋กœ๋ฒŒ ์ง€์†๊ฐ€๋Šฅ์„ฑ ๋ฐ ๊ธฐํ›„ ์ •์ฑ… ์ด๊ด„์ธ ๋งˆ์Šค๋“  ํ•œ๋‚˜๋Š” ์ง€๋‚œ 15์ผ ์—ด๋ฆฐ ๋ฏธ๊ตญ๊ธฐ์—…์—ฐ๊ตฌ์†Œ(AEI) ํ–‰์‚ฌ์—์„œ โ€œ์ „๋ ฅ๋ง์˜ ํ•ต์‹ฌ ๊ณผ์ œ๋Š” ์†ก์ „ ์žฅ๋ฒฝโ€์ด๋ผ๋ฉฐ, ์—ฌ๋Ÿฌ ์ง€์—ญ์˜ ์ „๋ ฅ ๊ธฐ์—…์ด ๋ฐ์ดํ„ฐ์„ผํ„ฐ ์‹ ๊ทœ ์—ฐ๊ฒฐ๊นŒ์ง€ 4๋…„์—์„œ 10๋…„์ด ๊ฑธ๋ฆฐ๋‹ค๊ณ  ์•ˆ๋‚ดํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ์ค‘ ์ผ๋ถ€ ๊ธฐ์—…์€ ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ ์š”์ฒญ์„ ๊ฒ€ํ† ํ•˜๋Š” ๋ฐ๋งŒ 12๋…„์ด ํ•„์š”ํ•˜๋‹ค๊ณ  ํ†ต๋ณดํ•œ ๊ฒƒ์œผ๋กœ ์•Œ๋ ค์กŒ๋‹ค.

์‹ค์ œ๋กœ ๋ฐ์ดํ„ฐ์„ผํ„ฐ ์ „๋ ฅ ์†Œ๋น„๊ฐ€ ๊ธ‰๊ฒฉํžˆ ๋Š˜์–ด๋‚  ๊ฒƒ์ด๋ผ๋Š” ์ „๋ง์€ ๊พธ์ค€ํžˆ ์ œ๊ธฐ๋˜๊ณ  ์žˆ๋‹ค. ์ง€๋‚œํ•ด 12์›” ๋กœ๋ Œ์Šค๋ฒ„ํด๋ฆฌ๊ตญ๋ฆฝ์—ฐ๊ตฌ์†Œ๊ฐ€ ๋ฐœํ‘œํ•œ ์—ฐ๊ตฌ์— ๋”ฐ๋ฅด๋ฉด, ๋ฐ์ดํ„ฐ์„ผํ„ฐ ์ „๋ ฅ ์‚ฌ์šฉ๋Ÿ‰์€ 2023๋…„ 176ํ…Œ๋ผ์™€ํŠธ์‹œ(TWh)์—์„œ 2028๋…„ 325~580TWh ์ˆ˜์ค€์œผ๋กœ ์ฆ๊ฐ€ํ•  ์ „๋ง์ด๋‹ค. ํ•˜์ง€๋งŒ ์ „๋ ฅ๋ง์€ ์ด์ฒ˜๋Ÿผ ์ฆ๊ฐ€ํ•˜๋Š” ์ˆ˜์š”๋ฅผ ๊ฐ๋‹นํ•˜์ง€ ๋ชปํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ๋ฏธ๊ตญ์—์„œ๋งŒ ๋ฐœ์ „ ๋ฐ ์—๋„ˆ์ง€ ์ €์žฅ ์„ค๋น„ ๋“ฑ ์•ฝ 2,300๊ธฐ๊ฐ€์™€ํŠธ(GW) ๊ทœ๋ชจ์˜ ์ „๋ ฅ ์šฉ๋Ÿ‰์ด ์•„์ง๋„ ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ์„ ๊ธฐ๋‹ค๋ฆฌ๊ณ  ์žˆ๋Š” ์ƒํ™ฉ์ด๋‹ค.

๊ธฐ์—…์˜ IT ์ธํ”„๋ผ๋ฅผ ์ด๊ด„ํ•˜๋Š” CIO์—๊ฒŒ ์ „๋ ฅ๋ง ๋ฌธ์ œ๋Š” ํด๋ผ์šฐ๋“œ ์šฉ๋Ÿ‰ ํ™•์žฅ์— ๋Œ€ํ•œ ๊ธฐ์กด ์ธ์‹์„ ๋ฐ”๊ฟ”๋†“๊ณ  ์žˆ๋‹ค. ๊ทธ๋ ˆ์ดํ•˜์šด๋“œ ๋ฆฌ์„œ์น˜์˜ ์ˆ˜์„ ์• ๋„๋ฆฌ์ŠคํŠธ ์‚ฐ์นซ ๋น„๋ฅด ๊ณ ๊ธฐ์•„๋Š” โ€œ์ด์ œ ํด๋ผ์šฐ๋“œ๊ฐ€ ๋ฌดํ•œํ•˜๋‹ค๋Š” ์ „์ œ๋ฅผ ๋ฒ„๋ ค์•ผ ํ•œ๋‹ค. ์ง€๊ธˆ์˜ ๋ณ‘๋ชฉ์€ ๋ฐ˜๋„์ฒด๋‚˜ ์†Œํ”„ํŠธ์›จ์–ด๊ฐ€ ์•„๋‹ˆ๋ผ ์ „๊ธฐ ๊ณต๊ธ‰โ€์ด๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค.

์ „๋ ฅ๋ง ์ „๋ฐ˜์œผ๋กœ ํ™•์‚ฐ๋˜๋Š” ๊ตฌ์กฐ์  ์ง€์—ฐ

๊ตฌ๊ธ€์ด ๊ฒช๊ณ  ์žˆ๋Š” ์—ฐ๊ฒฐ ์ง€์—ฐ์€ ํŠน์ • ๊ธฐ์—…์ด ์•„๋‹ˆ๋ผ ์ „๋ ฅ๋ง ์ „๋ฐ˜์— ๋ˆ„์ ๋œ ๋ฌธ์ œ๋‹ค. ๋ฒ„ํด๋ฆฌ๊ตญ๋ฆฝ์—ฐ๊ตฌ์†Œ ์ž๋ฃŒ์— ๋”ฐ๋ฅด๋ฉด ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ ๋Œ€๊ธฐ ๊ธฐ๊ฐ„์€ 2000~2007๋…„์— ์ถ”์ง„๋œ ํ”„๋กœ์ ํŠธ์˜ ๊ฒฝ์šฐ 2๋…„์ด ์ฑ„ ๊ฑธ๋ฆฌ์ง€ ์•Š์•˜์ง€๋งŒ, 2018~2024๋…„์—๋Š” ํ‰๊ท  4๋…„ ์ด์ƒ์œผ๋กœ 2๋ฐฐ ๋„˜๊ฒŒ ๋Š˜์–ด๋‚ฌ๋‹ค. 2000~2019๋…„ ์‚ฌ์ด ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ์„ ์‹ ์ฒญํ•œ ์ „์ฒด ์„ค๋น„ ๊ฐ€์šด๋ฐ 2024๋…„ ๋ง๊นŒ์ง€ ์‹ค์ œ ์ƒ์—… ์šด์ „์— ๋„๋‹ฌํ•œ ๋น„์œจ์€ 13%์— ๋ถˆ๊ณผํ–ˆ๋‹ค.

์ด ๊ฐ™์€ ์ง€์—ฐ์˜ ๊ทผ๋ณธ ์›์ธ์œผ๋กœ๋Š” ๊ธ‰์ฆํ•˜๋Š” ์ „๋ ฅ ์ˆ˜์š”๋ฅผ ๊ฐ๋‹นํ•˜์ง€ ๋ชปํ•˜๋Š” ๋…ธํ›„ ์†ก์ „ ์ธํ”„๋ผ๊ฐ€ ์ง€๋ชฉ๋๋‹ค. ํ•œ๋‚˜๋Š” โ€œ์ง€์—ญ ๊ฐ„ ์†ก์ „์„ ๋กœ๋ฅผ ์ƒˆ๋กœ ๊ตฌ์ถ•ํ•˜๋ ค๋ฉด ํ—ˆ๊ฐ€ ์ ˆ์ฐจ๋งŒ 7~11๋…„์ด ์†Œ์š”๋œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œSPP(Southwest Power Pool) ์˜ˆ์ธก์— ๋”ฐ๋ฅด๋ฉด ์†ก์ „ ์ธํ”„๋ผ ํ™•์ถฉ ์†๋„๊ฐ€ ์ง€๊ธˆ์ฒ˜๋Ÿผ ์ˆ˜์š”๋ฅผ ๋”ฐ๋ผ๊ฐ€์ง€ ๋ชปํ•  ๊ฒฝ์šฐ, ์ „๋ ฅ์„ ๊ณต๊ธ‰ํ•˜์ง€ ๋ชปํ•˜๋Š” โ€˜๋ถ€ํ•˜ ์†์‹คโ€™ ์ƒํ™ฉ์ด ์ตœ๋Œ€ 115์ผ์— ์ด๋ฅผ ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

์ „๋ ฅ๋ง ์ „๋ฐ˜์— ๊ฑธ์นœ ์ง€์—ฐ์€ ๊ธฐ์—…์œผ๋กœ ํ•˜์—ฌ๊ธˆ ํด๋ผ์šฐ๋“œ ์šฉ๋Ÿ‰์— ๋Œ€ํ•œ ๊ธฐ๋ณธ ๊ฐ€์ •์„ ์žฌ๊ณ ํ•˜๊ฒŒ ํ•˜๊ณ  ์žˆ๋‹ค. ๋ถ๋ถ€ ๋ฒ„์ง€๋‹ˆ์•„์™€ ์‚ฐํƒ€ํด๋ผ๋ผ์ฒ˜๋Ÿผ ์ดˆ๋Œ€ํ˜• ๋ฐ์ดํ„ฐ์„ผํ„ฐ ๊ตฌ์ถ•์˜ ์ตœ์ ์ง€๋กœ ๊ผฝํžˆ๋˜ ์ง€์—ญ์กฐ์ฐจ ๊ณต๊ธ‰ํ•  ์ „๋ ฅ์ด ๋น ๋ฅด๊ฒŒ ์†Œ์ง„๋˜๊ณ  ์žˆ๋Š” ์ƒํ™ฉ์ด๋‹ค.

์ธํ”„๋ผ ์ œ์•ฝ์€ ํด๋ผ์šฐ๋“œ ์‹œ์žฅ์˜ ๊ฒฝ์Ÿ ๊ตฌ๋„๋„ ๋ฐ”๊พธ๊ณ  ์žˆ์œผ๋ฉฐ, ๊ธฐ์ˆ  ์—ญ๋Ÿ‰์ด ์•„๋‹ˆ๋ผ ์ „๋ ฅ์—์˜ ์ ‘๊ทผ์„ฑ์ด ๊ฒฝ์Ÿ์˜ ํ•ต์‹ฌ ์š”์†Œ๋กœ ๋– ์˜ค๋ฅด๊ณ  ์žˆ๋‹ค. ๊ณ ๊ธฐ์•„๋Š” โ€œ์ด์ œ ๊ฒฝ์Ÿ์€ ๋ˆ„๊ฐ€ ๋” ๋งŽ์€ GPU ์ธ์Šคํ„ด์Šค๋ฅผ ๋จผ์ € ๋‚ด๋†“๋А๋ƒ์˜ ๋ฌธ์ œ๊ฐ€ ์•„๋‹ˆ๋‹ค. ๋ˆ„๊ฐ€ ๋จผ์ € ์ „๋ ฅ๋ง์— ์ ‘๊ทผํ•˜๋А๋ƒ์˜ ์‹ธ์›€์œผ๋กœ ๋ฐ”๋€Œ๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋Œ€์•ˆ์œผ๋กœ ๋– ์˜ค๋ฅธ โ€˜์ฝ”๋กœ์ผ€์ด์…˜โ€™

๊ธฐ์กด ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ์„ ์ˆ˜๋…„์”ฉ ๊ธฐ๋‹ค๋ฆด ์ˆ˜ ์—†๊ฒŒ ๋˜๋ฉด์„œ, ์ฃผ์š” ํด๋ผ์šฐ๋“œ ์—…์ฒด๋Š” ๋ฐœ์ „์†Œ ๋ฐ”๋กœ ์˜†์— ๋ฐ์ดํ„ฐ์„ผํ„ฐ๋ฅผ ์ง“๋Š” ์ฝ”๋กœ์ผ€์ด์…˜ ๋ฐฉ์‹์„ ๋Œ€์•ˆ์œผ๋กœ ์„ ํƒํ•˜๊ณ  ์žˆ๋‹ค. ์ด๋Š” ์†ก์ „๋ง์„ ๊ฑฐ์น˜์ง€ ์•Š๊ณ  ๋ฐœ์ „์›๊ณผ ์ง์ ‘ ์—ฐ๊ฒฐํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค. ๊ณ ๊ธฐ์•„๋Š” ์ „๋ ฅ ๊ณต๊ธ‰์ด ์ œํ•œ๋œ ์ง€์—ญ์˜ ์ฝ”๋กœ์ผ€์ด์…˜ ์ˆ˜์š”๊ฐ€ ๊ธ‰์ฆํ•˜๋ฉด์„œ ๊ณ„์•ฝ ๊ฐ€๊ฒฉ์ด ์•ฝ 20% ์ƒ์Šนํ–ˆ๋‹ค๊ณ  ๋ฐํžˆ๋ฉฐ, ์ด ๋น„์šฉ์ด ์ง€์—ญ๋ณ„ ๊ฐ€๊ฒฉ ์ฐจ์ด๋ฅผ ํ†ตํ•ด ํด๋ผ์šฐ๋“œ ๊ณ ๊ฐ์—๊ฒŒ ์ „๊ฐ€๋˜๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ตฌ๊ธ€๋„ ์ด ๊ฐ™์€ ๋ฐฉ์•ˆ์„ ๊ฒ€ํ† ํ•˜๊ณ ๋Š” ์žˆ์ง€๋งŒ, ํ•œ๋‚˜๋Š” ํšŒ์‚ฌ๊ฐ€ ์—ฌ์ „ํžˆ ์ „๋ ฅ๋ง์— ์—ฐ๊ฒฐ๋œ ๋ฐ์ดํ„ฐ์„ผํ„ฐ๋ฅผ ์ตœ์šฐ์„  ๋ฐฉ์‹์œผ๋กœ ์‚ผ๊ณ  ์žˆ๋‹ค๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ์ „๋ ฅ์„ ๋” ๋นจ๋ฆฌ ํ™•๋ณดํ•˜๊ธฐ ์œ„ํ•œ ์„ ํƒโ€์ด๋ผ๋ฉด์„œ๋„, ๋ฐ์ดํ„ฐ์„ผํ„ฐ๊ฐ€ ๋ฐœ์ „์†Œ ์ „๋ ฅ์„ ๋‹จ๋…์œผ๋กœ ์‚ฌ์šฉํ•˜๋Š” ๊ณ ๋ฆฝํ˜• ์‹œ์„ค์ด ์•„๋‹ˆ๋ผ ์ „๋ ฅ๋ง์— ์—ฐ๊ฒฐ๋ผ ๋‹ค๋ฅธ ์ˆ˜์š”์ฒ˜์™€ ์ „๋ ฅ์„ ํ•จ๊ป˜ ์‚ฌ์šฉํ•˜๋Š” ํ˜•ํƒœ๋ฅผ ์œ ์ง€ํ•˜๊ธธ ์›ํ•œ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

๊ทธ์— ๋”ฐ๋ฅด๋ฉด ๋‹ค๋ฅธ ํด๋ผ์šฐ๋“œ ์—…์ฒด ์—ญ์‹œ ์ „๋ ฅ ํšŒ์‚ฌ์™€ ์ง์ ‘ ํ˜‘์ƒ์— ๋‚˜์„œ๊ฑฐ๋‚˜ ๋ฐœ์ „์†Œ ์ธ๊ทผ ํ† ์ง€๋ฅผ ํ™•๋ณดํ•˜๊ณ , ๋ฐฐํ„ฐ๋ฆฌ๋ถ€ํ„ฐ ์†Œํ˜• ๋ชจ๋“ˆ ์›์ž๋กœ(SMR)์— ์ด๋ฅด๊ธฐ๊นŒ์ง€ ๋‹ค์–‘ํ•œ ์ „๋ ฅ ์ธํ”„๋ผ์— ๋Œ€ํ•œ ํˆฌ์ž ๊ฐ€๋Šฅ์„ฑ์„ ๊ฒ€ํ† ํ•˜๊ณ  ์žˆ๋‹ค.

๋‹ค๋งŒ ์ฝ”๋กœ์ผ€์ด์…˜์€ ์ƒˆ๋กœ์šด ์•ˆ์ •์„ฑ ๋ฌธ์ œ๋ฅผ ๋™๋ฐ˜ํ•œ๋‹ค. ๊ณ ๊ธฐ์•„๋Š” ๋ฐ์ดํ„ฐ์„ผํ„ฐ๊ฐ€ ์ „๋ ฅ๋ง๊ณผ ๋ถ„๋ฆฌ๋œ ์ƒํƒœ์—์„œ ๋‹จ์ผ ๋ฐœ์ „์›์— ์˜์กดํ•  ๊ฒฝ์šฐ ๊ธฐ์กด์˜ ์ด์ค‘ํ™” ์ „๋žต์„ ์ ์šฉํ•˜๊ธฐ ์–ด๋ ค์›Œ์งˆ ์ˆ˜ ์žˆ๋‹ค๊ณ  ์ง€์ ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ์ „ํ†ต์ ์ธ ํด๋ผ์šฐ๋“œ ๋ฆฌ์ „์€ ์ด์ค‘ ์ „๋ ฅ ๊ณต๊ธ‰๋ง๊ณผ ์ค‘๋ณต ๋ณ€์ „์†Œ, ๋„คํŠธ์›Œํฌ ์ „๋ฐ˜์—์„œ ์ „๋ ฅ์„ ๋Œ์–ด์˜ฌ ์ˆ˜ ์žˆ๋Š” ๊ตฌ์กฐ๋ฅผ ๊ฐ–์ถฐ ์™”๋‹ค. ๋ฐ˜๋ฉด ์ฝ”๋กœ์ผ€์ด์…˜ ๋ฐ์ดํ„ฐ์„ผํ„ฐ๋Š” ์ข…์ข… ๊ณ ๋ฆฝ๋œ ์„ฌ์ฒ˜๋Ÿผ ์šด์˜๋œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์ž„์‹œ ๋Œ€์‘์„ ๋„˜์–ด ๊ตฌ์กฐ์  ํ•ด๋ฒ• ํ•„์š”

์ฝ”๋กœ์ผ€์ด์…˜์ด ๋‹จ๊ธฐ์ ์ธ ๋Œ€์‘์ฑ…์ด ๋  ์ˆ˜๋Š” ์žˆ์ง€๋งŒ, ์†ก์ „ ์šฉ๋Ÿ‰ ๋ถ€์กฑ์ด๋ผ๋Š” ๊ทผ๋ณธ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋ ค๋ฉด ๋ณด๋‹ค ํญ๋„“์€ ์ •์ฑ…์  ๊ฐœ์ž…์ด ํ•„์š”ํ•˜๋‹ค๋Š” ์ง€์ ๋„ ๋‚˜์˜จ๋‹ค. ์ด์™€ ๊ด€๋ จํ•ด ๋ฏธ๊ตญ ์—ฐ๋ฐฉ ๊ทœ์ œ ๋‹น๊ตญ์€ ์ง€๋‚œ 2023๋…„ ์ „๋ ฅ๋ง ์—ฐ๊ฒฐ ์ ˆ์ฐจ๋ฅผ ๊ฐœ์„ ํ•˜๊ธฐ ์œ„ํ•ด ๋‚˜์„  ๋ฐ” ์žˆ๋‹ค. ๊ธฐ์กด์— ์‹ ์ฒญ ์ˆœ์„œ๋Œ€๋กœ ์ฒ˜๋ฆฌํ•˜๋˜ ๋ฐฉ์‹ ๋Œ€์‹ , ์ค€๋น„๊ฐ€ ๊ฐ€์žฅ ์ž˜ ๋œ ์‚ฌ์—…์„ ์šฐ์„  ์—ฐ๊ฒฐํ•˜๋Š” โ€˜์ค€๋น„ ์šฐ์„ โ€™ ํด๋Ÿฌ์Šคํ„ฐ ๋ชจ๋ธ์„ ๋„์ž…ํ•˜๊ณ , ๋ถ€์ง€ ํ™•๋ณด ์š”๊ฑด์„ ๊ฐ•ํ™”ํ•˜๋Š” ํ•œํŽธ ๋ณด์ฆ๊ธˆ ๋ถ€๋‹ด๋„ ๋†’์˜€๋‹ค. ๋‹ค๋งŒ ์‹ค์ œ ์ดํ–‰์€ ์—ฌ์ „ํžˆ ๋”๋””๊ณ  ์ „๋ ฅ๋ง ์ž์ฒด์˜ ์šฉ๋Ÿ‰ ๋ถ€์กฑ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ์—๋Š” ์—ญ๋ถ€์กฑ์ธ ์ƒํ™ฉ์ด๋‹ค.

๋ฏธ๊ตญ ์—๋„ˆ์ง€๋ถ€๋Š” ์ฒจ๋‹จ ์†ก์ „ ๊ธฐ์ˆ ์„ ํ™œ์šฉํ•  ๊ฒฝ์šฐ ์ตœ๋Œ€ 100GW์˜ ์ถ”๊ฐ€ ์šฉ๋Ÿ‰์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ์„ ๊ฒƒ์œผ๋กœ ๋ณด๊ณ  ์žˆ๋‹ค. ๋‹ค๋งŒ ์ด๋ฅผ ํ˜„์‹คํ™”ํ•˜๋ ค๋ฉด ์ „๋ ฅ ๊ธฐ์—…์ด ํ•ด๋‹น ๊ธฐ์ˆ ์„ ๋„์ž…ํ•˜๋„๋ก ์œ ๋„ํ•  ์ˆ˜ ์žˆ๋Š” ๊ทœ์ œ ํ™˜๊ฒฝ ๋ณ€ํ™”๊ฐ€ ํ•„์š”ํ•˜๋‹ค. ๊ณจ๋“œ๋งŒ์‚ญ์Šค ๋ฆฌ์„œ์น˜๋Š” ๋ฐ์ดํ„ฐ์„ผํ„ฐ ํ™•์žฅ์„ ์ง€์›ํ•˜๊ธฐ ์œ„ํ•ด 2030๋…„๊นŒ์ง€ ์•ฝ 7,200์–ต ๋‹ฌ๋Ÿฌ์— ๋‹ฌํ•˜๋Š” ์ „๋ ฅ๋ง ํˆฌ์ž๊ฐ€ ํ•„์š”ํ•  ๊ฒƒ์œผ๋กœ ์ถ”์‚ฐํ–ˆ๋‹ค.

๊ณ ๊ธฐ์•„๋Š” โ€œํด๋ผ์šฐ๋“œ ์—…์ฒด๋‚˜ ๋ฆฌ์ „์„ ์„ ํƒํ•  ๋•Œ ์ด์ œ๋Š” ๊ธฐ์ˆ ์  ์ ํ•ฉ์„ฑ๋ฟ๋งŒ ์•„๋‹ˆ๋ผ, ํ•ด๋‹น ์‚ฌ์—…์ž๊ฐ€ ํ–ฅํ›„ 5๋…„ ๋™์•ˆ ์›Œํฌ๋กœ๋“œ๋ฅผ ๊ฐ๋‹นํ•  ์ˆ˜ ์žˆ๋Š” ์ „๋ ฅ ์‚ฌ์šฉ ์—ญ๋Ÿ‰์„ ํ™•๋ณดํ•˜๊ณ  ์žˆ๋Š”์ง€ ์—ฌ๋ถ€๋„ ํ™•์ธํ•ด์•ผ ํ•œ๋‹ค. ์ง€๊ธˆ ๊ฐ™์€ ์‹œ์žฅ์—์„œ๋Š” ํ›จ์”ฌ ๋” ์‹ ์ค‘ํ•˜๊ฒŒ ์„ ํƒํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

IBM pushes sovereign computing with a software stack that works across cloud platforms

IBM has launched Sovereign Core, a software stack that aims to offer enterprises and governments full operational control over sovereign cloud deployments without relying on hyperscaler-managed regions.

Sovereign deployments, typically, try to combine cloud benefits with strategic autonomy. They are IT infrastructures that have been set up locally, ideally in isolated cloud environments, to ensure complete national or organizational control over data, operations, and security, while ensuring compliance with local laws, such as data residency regulations.

Unlike traditional sovereign clouds from Microsoft or Google that hinge on dedicated data center locations, IBMโ€™s Sovereign Core, expected to be available in tech preview in February, is trying to make sovereignty an inherent property of any software or application that an enterprise or government wants to deploy, enabling customers to run workloads on their own hardware, local providers, or even other clouds.

โ€œItโ€™s less a sovereign cloud and more ofย a software stack to build your own sovereign cloud,โ€ Dion Hinchcliffe, lead of the CIO practice at the Futurum Group, said, adding that Core can be used across environments, such as on-premises data centers, supported in-region cloud infrastructure, or through IT service providers.

Avoiding vendor lock-in

That shift in approach, according to analysts, could redefine how CIOs manage sovereign deployments and help them avoid vendor lock-in.

In traditional sovereign cloud deployments, hyperscalers retain control over critical operations like updates and access, creating regulatory risk and locking customers into provider-specific architectures, APIs, and compliance tools, Hinchcliffe said.

When workloads move, identity management, encryption keys, and audit trails tied to the old provider donโ€™t transfer seamlessly, forcing CIOs to rebuild governance frameworks to meet regulatory requirements in the new environment, Hinchcliffe added.

In contrast, Sovereign Core is trying to offer more control to CIOs by allowing them to keep encryption keys, identity management, and operational authority within their jurisdiction, which should enable them to switch providers without rebuilding governance frameworks, Hinchcliffe pointed out.

Seconding Hinchcliffe, HyperFRAME Researchโ€™s leader of AI stack Stephanie Walter noted that the frequency and stringency of regulator-driven audits were increasing, specifically the EU: Regulators are no longer satisfied with promises of compliance but are seeking more evidence, audit trails, and continuous compliance reporting.

Sovereign Core, according to Hinchcliffe, could also help CIOs tackle these demands with automated evidence collection and continuous monitoring, reducing overhead for banks, government agencies, and defense-adjacent industries.

Boost for moving sovereign AI pilots to production

Analysts say Sovereign Core could help CIOs and their enterprises push their AI pilots into production, especially the ones that require strict data residency and compliance controls.

Most enterprises and organizations are hesitant to send proprietary data to a public AI model, and at the same canโ€™t run GPU-backed inference completely inside their own sovereign boundary, said Phil Fersht, CEO of HFS Research.

Sovereign Coreโ€™s functionalities and capabilities, in contrast, will allow enterprises to run local AI inference inside their own four walls, ensuring the AI model is as โ€œsovereignโ€ as the data itโ€™s processing, in turn providing CIOs with a credible landing zone to move AI from pilots into production under sovereign conditions, Fersht added.

Changing market dynamics

Sovereign Core could be a strategic move by IBM to double down on the sovereignty market ahead of broader AI regulation and surge ahead of hyperscalers such as Microsoft, AWS, and Google.

โ€œWith Europe tightening controls and APAC following, IBM is betting that sovereignty will be a major gating factor for enterprise AI adoption. For some companies, much more even than cost or performance,โ€ Hinchcliffe said.

More so in Europe because regulations restrict foreign entities, such as the hyperscalers, which are all headquartered in the US, from having access to data or control over critical IT systems.

To comply with European regulations, hyperscalers typically work with local integrators and managed service providers, but retain operational control of the underlying platform while partners build and manage services on top, Hinchcliffe said.

IBMโ€™s Sovereign Core takes a different approach: partners can operate the entire environment on behalf of the customer, with IBM stepping out of the operational loop altogether, ensuring more compliance with regulations, Hinchcliffe added.

To that extent, IBM said that it is planning to collaborate with IT service providers globally, starting with an initial rollout in Europe with Computacenter in Germany.

IBM plans to make Sovereign Core generally available around the middle of 2026 with additional capabilities, which are likely to be disclosed soon.

์นผ๋Ÿผ | AI๊ฐ€ ๋ณ€ํ™”์‹œํ‚ค๋Š” SaaS, ํŠธ๋ Œ๋“œ์™€ ๊ธฐํšŒ, ๋„˜์–ด์•ผ ํ•  ๊ณผ์ œ๋Š”?

10์—ฌ ๋…„ ์ „ ์ฒ˜์Œ SaaS ์˜์—ญ์— ๋ฐœ์„ ๋“ค์˜€์„ ๋‹น์‹œ, ์ด ๊ฐœ๋…์€ ๋งค์šฐ ํ˜์‹ ์ ์œผ๋กœ ๋А๊ปด์กŒ๋‹ค. ํด๋ผ์šฐ๋“œ ๊ธฐ๋ฐ˜ ์†”๋ฃจ์…˜์œผ๋กœ์˜ ์ „ํ™˜์€ ๊ธฐ์—…์ด ์†Œํ”„ํŠธ์›จ์–ด๋ฅผ ํ™œ์šฉํ•˜๋Š” ๋ฐฉ์‹๊ณผ ์‚ฌ์—… ํ™•์žฅ ๊ตฌ์กฐ, ๋น„์šฉ ๊ด€๋ฆฌ ๋ฐฉ์‹๊นŒ์ง€ ๋ณ€ํ™”์‹œ์ผฐ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์ตœ๊ทผ ๋“ค์–ด ๊ทธ๋ณด๋‹ค ๋” ํฐ ๋ณ€ํ™”๋ฅผ ๋ชฉ๊ฒฉํ•˜๊ณ  ์žˆ๋‹ค. ๋ฐ”๋กœ SaaS์™€ AI์˜ ๊ฒฐํ•ฉ์ด๋‹ค.

AI๋Š” ๋” ์ด์ƒ ๋ถ€๊ฐ€ ๊ธฐ๋Šฅ์ด๊ฑฐ๋‚˜ ๋ฐœํ‘œ ์ž๋ฃŒ์—๋‚˜ ๋“ฑ์žฅํ•˜๋Š” ์œ ํ–‰์–ด๊ฐ€ ์•„๋‹ˆ๋‹ค. ํ˜„๋Œ€์ ์ธ SaaS ํ”Œ๋žซํผ์ด ์ž‘๋™ํ•˜๊ณ  ์„ฑ์žฅํ•˜๋Š” ํ•ต์‹ฌ ์ถ•์œผ๋กœ ์ž๋ฆฌ ์žก๊ณ  ์žˆ๋‹ค. ์ฑ… โ€˜์ˆ˜๋ฐฑ๋งŒ ๋‹ฌ๋Ÿฌ๋ฅผ ํˆฌ์žํ•˜๊ธฐ ์ „์— ์•Œ์•„์•ผ ํ•  SaaS ์ธ์‚ฌ์ดํŠธ(Get SaaS Insights Before You Invest Millions)โ€™์˜ ์ €์ž์ด์ž SaaS ์‹œ์Šคํ…œ ๋ฐ AI ์ „ํ™˜์„ ๊ฒฝํ—˜ํ•ด ์˜จ ํ•„์ž๋Š”, ์ด ๊ฐ™์€ ์œตํ•ฉ์ด ์ œํ’ˆ๊ณผ ๋น„์ฆˆ๋‹ˆ์Šค ๋ชจ๋ธ, ๊ณ ๊ฐ ๊ธฐ๋Œ€์น˜๋ฅผ ์–ด๋–ป๊ฒŒ ์žฌํŽธํ•˜๊ณ  ์žˆ๋Š”์ง€๋ฅผ ํ˜„์žฅ์—์„œ ์ง์ ‘ ํ™•์ธํ•ด ์™”๋‹ค.

SaaS์™€ AI๊ฐ€ ๋น ๋ฅด๊ฒŒ ๊ฒฐํ•ฉํ•˜๋ฉฐ ์ฐจ์„ธ๋Œ€ ๋””์ง€ํ„ธ ํ”Œ๋žซํผ์œผ๋กœ ์ง„ํ™”ํ•˜๋Š” ๊ณผ์ •์—์„œ, IT ๋ฆฌ๋”๊ฐ€ ๋ฐ˜๋“œ์‹œ ์•Œ์•„์•ผ ํ•  ์ฃผ์š” ํŠธ๋ Œ๋“œ์™€ ์‹ค์งˆ์ ์ธ ๊ธฐํšŒ, ๊ทธ๋ฆฌ๊ณ  ๊ฒ‰์œผ๋กœ ๋“œ๋Ÿฌ๋‚˜์ง€ ์•Š๋Š” ๊ณผ์ œ๋ฅผ ์งš์–ด๋ณธ๋‹ค.

SaaS์˜ ์ƒˆ๋กœ์šด ๊ธฐ๋ฐ˜์œผ๋กœ ์ž๋ฆฌ ์žก์€ AI

์ง€๋‚œ 3๋…„๊ฐ„ ์ฒด๊ฐํ•œ ๊ฐ€์žฅ ํฐ ๋ณ€ํ™”๋Š” AI๊ฐ€ ๋” ์ด์ƒ ํ•˜๋‚˜์˜ ๊ตฌ์„ฑ ์š”์†Œ๋‚˜ ๊ธฐ๋Šฅ์œผ๋กœ๋งŒ ์ธ์‹๋˜์ง€ ์•Š๋Š”๋‹ค๋Š” ์ ์ด๋‹ค. AI๋Š” ํ•ต์‹ฌ ์—ญ๋Ÿ‰์œผ๋กœ ์ง„ํ™”ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, SaaS ์—…์ฒด๋Š” ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ AI๋ฅผ ์ตœ์šฐ์„ ์— ๋‘๋Š” ๋ฐฉํ–ฅ์œผ๋กœ ํ”Œ๋žซํผ์„ ์žฌ์„ค๊ณ„ํ•˜๊ณ  ์žˆ๋‹ค.

์›Œํฌํ”Œ๋กœ์šฐ ์ž๋™ํ™”์—์„œ ์ง€๋Šฅํ˜• ์ž๋™ํ™”๋กœ

์ดˆ๊ธฐ SaaS ์‹œ์Šคํ…œ์ด ์—…๋ฌด๋ฅผ ์ž๋™ํ™”ํ•˜๋Š” ๋ฐ ์ดˆ์ ์„ ๋งž์ท„๋‹ค๋ฉด, AI ๊ธฐ๋ฐ˜ SaaS ์‹œ์Šคํ…œ์€ ์˜์‚ฌ๊ฒฐ์ • ์ž์ฒด๋ฅผ ์ž๋™ํ™”ํ•œ๋‹ค.

์˜ˆ์ธก ๋ถ„์„, ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ, ํ–‰๋™ ๊ธฐ๋ฐ˜ ํŠธ๋ฆฌ๊ฑฐ, ์ž๊ฐ€ ๋ณต๊ตฌ ์‹œ์Šคํ…œ, ์ƒํ™ฉ ์ธ์ง€ํ˜• ์ถ”์ฒœ ๊ธฐ๋Šฅ ๋“ฑ์€ ์ด์ œ ์ฐจ๋ณ„ํ™” ์š”์†Œ๊ฐ€ ์•„๋‹ˆ๋ผ ๊ธฐ๋ณธ ๊ธฐ๋Šฅ์œผ๋กœ ์ž๋ฆฌ ์žก๊ณ  ์žˆ๋‹ค.

ํ•„์ž๊ฐ€ ๊ตฌ์ถ•์— ์ฐธ์—ฌํ–ˆ๋˜ ํ•œ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ํ”Œ๋žซํผ์€ ๊ทœ์น™ ๊ธฐ๋ฐ˜ ์ž๋™ํ™”์—์„œ AI ๊ธฐ๋ฐ˜ ์˜ˆ์ธก ๋ชจ๋ธ๋กœ ์ „ํ™˜ํ–ˆ๋‹ค. ๊ทธ ๊ฒฐ๊ณผ ๊ณ ๊ฐ์—๊ฒŒ ์˜ํ–ฅ์„ ๋ฏธ์น˜๊ธฐ ์ˆ˜์‹œ๊ฐ„ ์ „ ์‹œ์Šคํ…œ ๋ฌธ์ œ๋ฅผ ๊ฐ์ง€ํ•  ์ˆ˜ ์žˆ์—ˆ๊ณ , ๊ฐ€๋™ ์ค‘๋‹จ ์‹œ๊ฐ„์€ ์ค„์–ด๋“ค์—ˆ์œผ๋ฉฐ ๊ณ ๊ฐ ๋งŒ์กฑ๋„๋Š” ๋†’์•„์กŒ๋‹ค. ์•„์šธ๋Ÿฌ ๊ธด๊ธ‰ ๋Œ€์‘ ํ•„์š”์„ฑ๋„ ๋‚ฎ์•„์กŒ๋‹ค.

AI๋Š” SaaS๋ฅผ ๋‹จ์ˆœํžˆ ๊ฐœ์„ ํ•˜๋Š” ์ˆ˜์ค€์„ ๋„˜์–ด, ํšจ์œจ์„ฑ์ด๋ผ๋Š” ๊ฐœ๋… ์ž์ฒด๋ฅผ ๊ทผ๋ณธ์ ์œผ๋กœ ๋ฐ”๊พธ๊ณ  ์žˆ๋‹ค. ์ด์™€ ๊ด€๋ จํ•œ ์ฃผ์š” ํŠธ๋ Œ๋“œ๋ฅผ ์†Œ๊ฐœํ•œ๋‹ค.

ํŠธ๋ Œ๋“œ 1: ๋งž์ถคํ™”๋Š” ์„ ํƒ์ด ์•„๋‹Œ ํ•„์ˆ˜

์œ ํ†ต๋ถ€ํ„ฐ ์˜๋ฃŒ๊นŒ์ง€ ์ „ ์‚ฐ์—…์—์„œ ๊ณ ๊ฐ์˜ ๊ด€์‹ฌ์„ ์œ ์ง€ํ•˜๊ธฐ ์œ„ํ•ด ๊ฐœ์ธ ๋งž์ถคํ™”๋ฅผ ํ•ต์‹ฌ ์š”์†Œ๋กœ ์‚ผ๊ณ  ์žˆ๋‹ค. SaaS ์ œํ’ˆ ์—ญ์‹œ ์ด๋Ÿฐ ํ๋ฆ„์„ ๊ทธ๋Œ€๋กœ ๋”ฐ๋ฅด๊ณ  ์žˆ๋‹ค. ์˜ค๋Š˜๋‚  ์‚ฌ์šฉ์ž๋Š” SaaS ํ”Œ๋žซํผ์ด ๋„ทํ”Œ๋ฆญ์Šค๋‚˜ ์Šคํฌํ‹ฐํŒŒ์ด์ฒ˜๋Ÿผ ์ž‘๋™ํ•˜๊ธธ ๊ธฐ๋Œ€ํ•œ๋‹ค. ๊ฐœ์ธ ๋งž์ถคํ™” ์š”์†Œ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

  • ์‚ฌ์šฉ์ž ๋งž์ถคํ˜• ๋Œ€์‹œ๋ณด๋“œ
  • ๊ฐœ์ธ๋ณ„๋กœ ์„ค์ •๋œ ์›Œํฌํ”Œ๋กœ์šฐ
  • ์ง€๋Šฅํ˜• ์ถ”์ฒœ ๊ธฐ๋Šฅ
  • ์‚ฌ์šฉ ํŒจํ„ด์— ๋”ฐ๋ผ ๋‹ฌ๋ผ์ง€๋Š” ์ ์‘ํ˜• ์ธํ„ฐํŽ˜์ด์Šค

ํ•„์ž๋Š” ๊ฐœ์ธ ๋งž์ถคํ™”๊ฐ€ SaaS ๋„์ž…์— ์–ผ๋งˆ๋‚˜ ํฐ ์˜ํ–ฅ์„ ๋ฏธ์น˜๋Š”์ง€๋ฅผ ํ˜„์žฅ์—์„œ ์ง์ ‘ ํ™•์ธํ•ด ์™”๋‹ค. ์ž๋ฌธ์— ์ฐธ์—ฌํ–ˆ๋˜ ํ•œ ํ•™์Šต ํ”Œ๋žซํผ์—์„œ๋Š” AI ๊ธฐ๋ฐ˜ ํ•™์Šต ๊ฒฝ๋กœ๋ฅผ ๋„์ž…ํ•œ ์ดํ›„, ์‚ฌ์šฉ์ž๊ฐ€ โ€˜์ œํ’ˆ์ด ์ž์‹ ์„ ์ดํ•ดํ•˜๊ณ  ์žˆ๋‹คโ€™๊ณ  ๋А๋ผ๋ฉด์„œ ์‚ฌ์šฉ์ž ์ฐธ์—ฌ๋„๊ฐ€ 60% ์ฆ๊ฐ€ํ–ˆ๋‹ค.

๋‹ค๋งŒ ๊ฐœ์ธํ™”๋Š” ์ƒˆ๋กœ์šด ๊ธฐ๋Œ€๋ฅผ ๋™๋ฐ˜ํ•œ๋‹ค. ์‚ฌ์šฉ์ž๋Š” ๋‹จ์ˆœํžˆ ์ •์ƒ์ ์œผ๋กœ ์ž‘๋™ํ•˜๋Š” ์†Œํ”„ํŠธ์›จ์–ด๋ฅผ ๋„˜์–ด, ์ž์‹ ์˜ ์š”๊ตฌ์— ๋งž๊ฒŒ ์ž‘๋™ํ•˜๋Š” ์†Œํ”„ํŠธ์›จ์–ด๋ฅผ ์›ํ•œ๋‹ค.

ํŠธ๋ Œ๋“œ 2: AI๊ฐ€ SaaS์˜ ๋น„์ฆˆ๋‹ˆ์Šค ์—ญํ•™์„ ์žฌํŽธ

SaaS๋Š” ๊ตฌ๋… ๋ชจ๋ธ๊ณผ ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ์ธํ”„๋ผ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์„ฑ์žฅํ•ด ์™”๋‹ค. ์—ฌ๊ธฐ์— AI๊ฐ€ ๊ฒฐํ•ฉ๋˜๋ฉด์„œ ์ƒˆ๋กœ์šด ์ฐจ์›์˜ ๋น„์ฆˆ๋‹ˆ์Šค๊ฐ€ ์ฐฝ์ถœ๋˜๊ณ  ์žˆ๋‹ค.

์‚ฌ์šฉ๋Ÿ‰ ๊ธฐ๋ฐ˜ ์š”๊ธˆ ์ฑ…์ •

ํ–‰๋™ ๋ชจ๋‹ˆํ„ฐ๋ง๊ณผ ๋ถ„์„์„ ์œ„ํ•œ AI ๊ธฐ์ˆ ์ด ๋ฐœ์ „ํ•˜๋ฉด์„œ, SaaS ๊ธฐ์—…์€ ๊ณ ๊ฐ์ด ์‹ค์ œ๋กœ ์–ป๋Š” ๊ฐ€์น˜์— ๋”ฐ๋ผ ์š”๊ธˆ์„ ์ฑ…์ •ํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋๋‹ค. ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์ ‘๊ทผ ๋ฐฉ์‹์ด ์ฆ๊ฐ€ํ•˜๊ณ  ์žˆ๋‹ค.

  • ๊ตฌ๋…๊ณผ ์‚ฌ์šฉ๋Ÿ‰์„ ๊ฒฐํ•ฉ
  • ๊ตฌ๋…๊ณผ ์ง€๋Šฅํ˜• ๊ธฐ๋Šฅ ๋“ฑ๊ธ‰์„ ๊ฒฐํ•ฉ
  • AI ๋น„์ค‘์ด ๋†’์€ ๊ธฐ๋Šฅ์— ๋Œ€ํ•œ ์ „์šฉ ์š”๊ธˆ์ œ

์ด๋Ÿฐ ๋ชจ๋ธ์€ ์ƒˆ๋กœ์šด ์ˆ˜์ต ๊ธฐํšŒ๋ฅผ ์ œ๊ณตํ•˜์ง€๋งŒ, ๊ณ ๊ฐ ํ–‰๋™์„ ์ •๋ฐ€ํ•˜๊ฒŒ ์ดํ•ดํ•ด์•ผ ํ•œ๋‹ค๋Š” ์ „์ œ๊ฐ€ ๋”ฐ๋ฅธ๋‹ค. ์ด ์—ญ๋Ÿ‰์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ๊ฒƒ์ด AI๋‹ค.

AI์— ๊ธฐ๋ฐ˜ํ•œ ์ œํ’ˆ ์ค‘์‹ฌ ์„ฑ์žฅ

AI ์ธ์‚ฌ์ดํŠธ๋ฅผ ํ™œ์šฉํ•˜๋ฉด ์˜จ๋ณด๋”ฉ์„ ๊ฐœ์„ ํ•˜๊ณ , ์‚ฌ์šฉ์ž๊ฐ€ ์ œํ’ˆ์˜ ๊ฐ€์น˜๋ฅผ ๋ช…ํ™•ํžˆ ์ฒด๊ฐํ•˜๋Š” ํ•ต์‹ฌ ์ ‘์ ์„ ๊ฐ•ํ™”ํ•˜๋ฉฐ, ์ดํƒˆ๋ฅ ์„ ์ค„์ด๋Š” ๋ฐ ๋„์›€์ด ๋œ๋‹ค. AI๊ฐ€ ์–ธ์ œ ์‚ฌ์šฉ์ž๋ฅผ ์•ˆ๋‚ดํ• ์ง€, ์–ด๋–ค ๋„์›€์„ ์ œ๊ณตํ• ์ง€, ์–ธ์ œ ํ•ต์‹ฌ ๊ธฐ๋Šฅ์œผ๋กœ ์œ ๋„ํ• ์ง€๋ฅผ ํŒ๋‹จํ•  ์ˆ˜ ์žˆ๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค. ์ด๋Š” ์„ฑ์žฅ ์ˆ˜์ต์„ ๋†’์ด๋Š” ๋ฐ๋„ ๊ธฐ์—ฌํ•œ๋‹ค.

ํŠธ๋ Œ๋“œ 3: AI ๋„ค์ดํ‹ฐ๋ธŒ SaaS ์ œํ’ˆ์˜ ๋ถ€์ƒ

์ด์ œ๋Š” AI๊ฐ€ ์ œํ’ˆ์˜ ํ•ต์‹ฌ ๊ฐ€์น˜ ์ž์ฒด๋ฅผ ํ˜•์„ฑํ•˜๋Š” โ€˜AI ๋„ค์ดํ‹ฐ๋ธŒ SaaSโ€™ ์‹œ๋Œ€๋กœ ์ ‘์–ด๋“ค๊ณ  ์žˆ๋‹ค. ์ด๋“ค ์†”๋ฃจ์…˜์€ ์ฒ˜์Œ๋ถ€ํ„ฐ ์ง€๋Šฅํ™”์™€ ์˜ˆ์ธก, ์ž์œจ์„ฑ์„ ํ•ต์‹ฌ์— ๋‘๊ณ  ์„ค๊ณ„๋œ ๊ฒƒ์ด ํŠน์ง•์ด๋‹ค.

๋Œ€ํ‘œ์ ์ธ ์‚ฌ๋ก€๋กœ๋Š” AI ๊ธฐ๋ฐ˜ CRM, ์ž์œจํ˜• ๋ณด์•ˆ ํ”Œ๋žซํผ, ์˜ˆ์ธก ์œ ์ง€๋ณด์ˆ˜ ์‹œ์Šคํ…œ, AI ๊ธฐ๋ฐ˜ ์žฌ๋ฌด ์˜ˆ์ธก ๋„๊ตฌ, ์ž๋™ํ™”๋œ ์ปดํ”Œ๋ผ์ด์–ธ์Šค ์—”์ง„ ๋“ฑ์ด ์žˆ๋‹ค.

ํ˜„์žฌ SaaS ์Šคํƒ€ํŠธ์—…์„ ํ‰๊ฐ€ํ•˜๋‹ค ๋ณด๋ฉด ๋‘ ์œ ํ˜•์„ ๋น„๊ต์  ๋ถ„๋ช…ํ•˜๊ฒŒ ๊ตฌ๋ถ„ํ•  ์ˆ˜ ์žˆ๋‹ค. ํ•˜๋‚˜๋Š” ๊ธฐ์กด SaaS์— AI๋ฅผ ๊ฒฐํ•ฉํ•œ ํ˜•ํƒœ์ด๊ณ , ๋‹ค๋ฅธ ํ•˜๋‚˜๋Š” AI ์—†์ด๋Š” ์ œํ’ˆ์ด ์ž‘๋™ํ•  ์ˆ˜ ์—†๋Š” ๊ตฌ์กฐ์˜ SaaS๋‹ค. ์•ž์œผ๋กœ ๋“ฑ์žฅํ•  ์ฐจ์„ธ๋Œ€ ์œ ๋‹ˆ์ฝ˜ ๊ธฐ์—…์€ ํ›„์ž์—์„œ ๋‚˜์˜ฌ ๊ฐ€๋Šฅ์„ฑ์ด ํฌ๋‹ค.

AI์™€ SaaS์˜ ์œตํ•ฉ์„ ํ†ตํ•ด ์–ป์„ ์ˆ˜ ์žˆ๋Š” ๊ธฐํšŒ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

๊ธฐํšŒ 1: AI๋กœ ์žฌ๊ตฌ์„ฑ๋˜๋Š” ๊ณ ๊ฐ ์ง€์›

๊ณ ๊ฐ ์ง€์›์€ ์˜ค๋žซ๋™์•ˆ SaaS ์ œํ’ˆ์„ ์šด์˜ํ•˜๋Š” ๋ฐ ์žˆ์–ด ๋น„์šฉ ๋ถ€๋‹ด์ด ํฐ ์˜์—ญ์œผ๋กœ ๊ผฝํ˜€ ์™”๋‹ค.

๊ทธ๋Ÿฌ๋‚˜ AI๊ฐ€ ๋„์ž…๋˜๋ฉด์„œ ๊ณ ๊ฐ ์ง€์› ๋ฐฉ์‹๋„ ๋น ๋ฅด๊ฒŒ ๋ณ€ํ™”ํ•˜๊ณ  ์žˆ๋‹ค. ๊ฐ์ • ์ธ์‹ ๊ธฐ๋ฐ˜ ์ฑ—๋ด‡, ์ด์Šˆ ์ž๋™ ๋ถ„๋ฅ˜, ์˜ˆ์ธกํ˜• ํ‹ฐ์ผ“ ๋ผ์šฐํŒ…, ์ž๋™ ์ƒ์„ฑ๋˜๋Š” ๋ฌธ์ œ ํ•ด๊ฒฐ ๊ฐ€์ด๋“œ, ์Œ์„ฑ ํ…์ŠคํŠธ ๋ณ€ํ™˜ ๋ฐ ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ ๊ธฐ๋ฐ˜ ์ง€์› ๊ธฐ๋Šฅ์ด ๋Œ€ํ‘œ์ ์ด๋‹ค.

ํ•„์ž๊ฐ€ ์ฐธ์—ฌํ–ˆ๋˜ ํ•œ SaaS ์ œํ’ˆ์—์„œ๋Š” AI ์ง€์› ์–ด์‹œ์Šคํ„ดํŠธ๋ฅผ ๋„์ž…ํ•œ ์ง€ ํ•œ ๋‹ฌ ๋งŒ์— ๋ฏธ์ฒ˜๋ฆฌ ํ‹ฐ์ผ“์ด 40% ์ค„์–ด๋“ค์—ˆ๋‹ค. ๊ณ ๊ฐ์€ ๋” ๋น ๋ฅด๊ฒŒ ์‘๋‹ต์„ ๋ฐ›์„ ์ˆ˜ ์žˆ์—ˆ๊ณ , ์ง€์› ์ธ๋ ฅ์€ ๋ฐ˜๋ณต์ ์ธ ๋ฌธ์˜ ๋Œ€์‹  ๋ณต์žกํ•œ ๋ฌธ์ œ ํ•ด๊ฒฐ์— ์ง‘์ค‘ํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋๋‹ค.

๊ธฐํšŒ 2: ์ง€๋Šฅํ˜• ์ œํ’ˆ ๋ถ„์„

AI๋Š” SaaS ํŒ€์— ์ด์ „๊ณผ๋Š” ๋‹ค๋ฅธ ์ˆ˜์ค€์˜ ์ธ์‚ฌ์ดํŠธ๋ฅผ ์ œ๊ณตํ•˜๊ณ  ์žˆ๋‹ค. ๊ตฌ์ฒด์ ์ธ ๋‚ด์šฉ์€ ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

  • ๋„์ž…๊ณผ ํ™•์‚ฐ์— ์˜ํ–ฅ์„ ๋ฏธ์น˜๋Š” ์ œํ’ˆ ํŠน์„ฑ
  • ๊ณ ๊ฐ ์ดํƒˆ๋กœ ์ด์–ด์ง€๋Š” ํ–‰๋™
  • ์‚ฌ์šฉ์ž๊ฐ€ ์ œํ’ˆ์„ ์ด์šฉํ•˜๋Š” ๊ฒฝ๋กœ
  • ํšจ๊ณผ์ ์ธ ๊ฐ€๊ฒฉ ์ฑ…์ • ์ „๋žต
  • ์ดํƒˆ ์œ„ํ—˜์— ๋†“์—ฌ์žˆ๋Š” ์กฐ์ง

๊ธฐ์กด์˜ ๋ถ„์„ ๋„๊ตฌ๊ฐ€ ๊ณผ๊ฑฐ์— ์–ด๋–ค ์ผ์ด ์žˆ์—ˆ๋Š”์ง€๋ฅผ ๋ณด์—ฌ์ฃผ๋Š” ๋ฐ ๊ทธ์ณค๋‹ค๋ฉด, AI๋Š” ๊ทธ ์›์ธ์„ ์„ค๋ช…ํ•˜๊ณ  ์•ž์œผ๋กœ ์ผ์–ด๋‚  ์ผ์„ ์˜ˆ์ธกํ•  ์ˆ˜ ์žˆ๋‹ค.

์˜ˆ์ธก ๋ถ„์„์„ ํ™œ์šฉํ•˜๋ฉด SaaS ํŒ€ ๋ฆฌ๋”๊ฐ€ ์ดํƒˆ ๊ฐ€๋Šฅ์„ฑ์„ ์‚ฌ์ „์— ํŒŒ์•…ํ•˜๊ณ , ๊ธฐ๋Šฅ ๋ณ‘๋ชฉ ๊ตฌ๊ฐ„์„ ์ฐพ์•„๋‚ด๋ฉฐ, ์ถ”๊ฐ€ ๊ตฌ๋งค๋ฅผ ์œ ๋„ํ•  ์ˆ˜ ์žˆ๋Š” ๊ธฐํšŒ๋ฅผ ์‹๋ณ„ํ•˜๊ณ , ์ œํ’ˆยท์‹œ์žฅ ์ ํ•ฉ์„ฑ์„ ํ›จ์”ฌ ๋†’์€ ์ •ํ™•๋„๋กœ ๊ฐœ์„ ํ•  ์ˆ˜ ์žˆ๋‹ค.

๊ธฐํšŒ 3: ํ™•์žฅ ๊ฐ€๋Šฅํ•˜๊ณ  ์Šค์Šค๋กœ ์ตœ์ ํ™”ํ•˜๋Š” ์ธํ”„๋ผ

AI๋Š” ๋ฐ๋ธŒ์˜ต์Šค(DevOps)์™€ ์ธํ”„๋ผ ๊ด€๋ฆฌ ์ „๋ฐ˜์„ ํ˜์‹ ํ•˜๊ณ  ์žˆ๋‹ค. ์ž๋™ ํ™•์žฅ, ์ด์ƒ ์ง•ํ›„ ํƒ์ง€, ์˜ˆ์ธกํ˜• ๋ถ€ํ•˜ ๋ถ„์‚ฐ, ์ž๋™ํ™”๋œ ๋ฐฐํฌ ๊ฒ€์ฆ, ์ง€๋Šฅํ˜• ์ž์› ํ”„๋กœ๋น„์ €๋‹์ด ๋Œ€ํ‘œ์ ์ด๋‹ค.

ํ•„์ž๊ฐ€ ์ฐธ์—ฌํ–ˆ๋˜ ํ•œ ํ”Œ๋žซํผ์—์„œ๋Š” AI ๊ธฐ๋ฐ˜ ๋ถ€ํ•˜ ์˜ˆ์ธก ๊ธฐ์ˆ ์„ ๋„์ž…ํ•ด ์ธํ”„๋ผ ๋น„์šฉ์„ 20% ์ด์ƒ ์ ˆ๊ฐํ–ˆ๋‹ค. ์‹œ์Šคํ…œ์ด ์ž์› ์ˆ˜์š” ๊ธ‰์ฆ์„ ์‚ฌ์ „์— ์˜ˆ์ธกํ•ด ์„ ์ œ์ ์œผ๋กœ ํ™•์žฅํ•˜๋ฉด์„œ, ๊ณผ๊ฑฐ์—๋Š” ์ˆ˜์ž‘์—… ๊ฐœ์ž…์ด ํ•„์š”ํ–ˆ๋˜ ์„ฑ๋Šฅ ์ €ํ•˜๋ฅผ ๋ฏธ๋ฆฌ ๋ฐฉ์ง€ํ•  ์ˆ˜ ์žˆ์—ˆ๋‹ค.

AI๋Š” ์„ฑ์žฅ ๊ณผ์ •์—์„œ ์Šค์Šค๋กœ ํ•™์Šตํ•˜๊ณ  ์ตœ์ ํ™”๋˜๋Š” SaaS ํ”Œ๋žซํผ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•œ๋‹ค. ํ•˜์ง€๋งŒ ์ด๋Ÿฌํ•œ ์œตํ•ฉ์—๋Š” ๊ณผ์ œ ๋˜ํ•œ ๋‚จ์•„์žˆ๋‹ค.

๊ณผ์ œ 1: ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ๊ณผ ๊ฑฐ๋ฒ„๋„Œ์Šค

AI๋Š” ์ •์ œ๋˜๊ณ  ์ผ๊ด€๋˜๋ฉฐ ์•ˆ์ „ํ•˜๊ฒŒ ๊ด€๋ฆฌ๋˜๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ์ „์ œ๋กœ ํ•œ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ๋งŽ์€ SaaS ๊ธฐ์—…์ด ์ด ์ ์„ ๊ฐ„๊ณผํ•˜๊ณ  ์žˆ๋‹ค.

์œ ๋งํ•ด ๋ณด์ด๋˜ AI ์‹œ์Šคํ…œ์ด ๋ฐ์ดํ„ฐ ๋ฌธ์ œ๋กœ ๋ฌด๋„ˆ์ง€๋Š” ์‚ฌ๋ก€๋ฅผ ์ˆ˜์ฐจ๋ก€ ๋ชฉ๊ฒฉํ•ด ์™”๋‹ค. ๋ฐ์ดํ„ฐ๊ฐ€ ๋ถˆ์™„์ „ํ–ˆ๊ฑฐ๋‚˜, ํ•™์Šต์— ํ•„์š”ํ•œ ๋ผ๋ฒจ์ด ๋ถ€์กฑํ–ˆ๊ฑฐ๋‚˜, ์‹œ์Šคํ…œ ๊ตฌ์กฐ ์ž์ฒด๊ฐ€ AI ํ™œ์šฉ์„ ๊ณ ๋ คํ•ด ์„ค๊ณ„๋˜์ง€ ์•Š์•˜๊ฑฐ๋‚˜, ์ ‘๊ทผ ์ œ์–ด๋กœ ์ธํ•ด ๋ชจ๋ธ ํ•™์Šต์ด ์ฐจ๋‹จ๋œ ๊ฒฝ์šฐ๊ฐ€ ๋Œ€ํ‘œ์ ์ด๋‹ค.

์กฐ์ง์€ ๋ฐ์ดํ„ฐ๋ฅผ ์‚ฌํ›„์ ์œผ๋กœ ์ฒ˜๋ฆฌํ•ด์•ผ ํ•  ์š”์†Œ๊ฐ€ ์•„๋‹ˆ๋ผ, ํ•˜๋‚˜์˜ ์ œํ’ˆ์œผ๋กœ ์ธ์‹ํ•˜๊ณ  ๊ด€๋ฆฌํ•ด์•ผ ํ•œ๋‹ค.

๊ณผ์ œ 2: ํŽธํ–ฅ, ๊ฐœ์ธ์ •๋ณด ๋ณดํ˜ธ, ์œค๋ฆฌ์  AI

๊ณ ๊ฐ์€ ์ ์  ๋” AI๋ฅผ ๊ฒฝ๊ณ„ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ๊ตฌ์ฒด์ ์ธ ์˜์—ญ์€ ๋‹ค์Œ๊ณผ ๊ฐ™๋‹ค.

  • ์ž์‹ ์˜ ๋ฐ์ดํ„ฐ๊ฐ€ ํ™œ์šฉ๋˜๋Š” ๋ฐฉ์‹
  • AI ๋ชจ๋ธ์ด ์˜์‚ฌ๊ฒฐ์ •์„ ๋‚ด๋ฆฌ๋Š” ๋ฐฉ์‹
  • ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ž‘๋™์˜ ๊ณต์ •์„ฑ
  • ํ”„๋ผ์ด๋ฒ„์‹œ ๋ณดํ˜ธ ์—ฌ๋ถ€

GDPR์„ ๋น„๋กฏํ•ด ๊ฐ์ข… AI ๊ทœ์ œ๊ฐ€ ๋“ฑ์žฅํ•˜๋ฉด์„œ, SaaS ๊ธฐ์—…์€ ์œค๋ฆฌ์™€ ์ปดํ”Œ๋ผ์ด์–ธ์Šค๋ฅผ ์šฐ์„  ๊ณผ์ œ๋กœ ์‚ผ์„ ์ˆ˜๋ฐ–์— ์—†๋Š” ์ƒํ™ฉ์ด๋‹ค.

ํ•„์ž ์—ญ์‹œ ํˆฌ๋ช…์„ฑ์„ ํ™•๋ณดํ•˜๊ณ  ํŽธํ–ฅ์„ ์ค„์ด๋ฉฐ, ๊ฐ์‚ฌ ์ถ”์ ์„ ์œ ์ง€ํ•˜๊ธฐ ์œ„ํ•ด AI ์›Œํฌํ”Œ๋กœ์šฐ๋ฅผ ์ „๋ฉด์ ์œผ๋กœ ์žฌ์„ค๊ณ„ํ•ด์•ผ ํ–ˆ๋˜ ๊ฒฝํ—˜์ด ์žˆ๋‹ค. ์ด๋Ÿฐ ์กฐ์น˜๋Š” ์‹œ๊ฐ„๊ณผ ์ž์›์„ ์š”๊ตฌํ•˜์ง€๋งŒ, AI ์‹œ์Šคํ…œ์˜ ์„ฑ๊ณต์— ํ•„์ˆ˜์ ์ธ ์š”์†Œ์ธ ์‹ ๋ขฐ๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ๊ธฐ๋ฐ˜์ด ๋œ๋‹ค.

๊ณผ์ œ 3: ์—ญ๋Ÿ‰ ๊ฒฉ์ฐจ

SaaS์™€ AI๋ฅผ ์œตํ•ฉํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์ „๋ฌธ๊ฐ€๊ฐ€ ํ•„์š”ํ•˜๋‹ค.

  • ๋จธ์‹ ๋Ÿฌ๋‹์— ๋Œ€ํ•œ ์ดํ•ด๋ฅผ ๊ฐ–์ถ˜ ์—”์ง€๋‹ˆ์–ด
  • ์ œํ’ˆ ๊ด€์ ์—์„œ ์‚ฌ๊ณ ํ•  ์ˆ˜ ์žˆ๋Š” ML ์ „๋ฌธ๊ฐ€
  • ๋ฐ์ดํ„ฐ๋ฅผ ์ดํ•ดํ•˜๋Š” ํ”„๋กœ๋•ํŠธ ๋งค๋‹ˆ์ €
  • AI ์ค‘์‹ฌ ์•„ํ‚คํ…์ฒ˜๋ฅผ ์„ค๊ณ„ํ•  ์ˆ˜ ์žˆ๋Š” ๋””์ž์ด๋„ˆ

์ด๋Ÿฐ ์—ญ๋Ÿ‰์„ ๋ชจ๋‘ ๊ฐ–์ถ˜ ์ธ์žฌ ์กฐํ•ฉ์„ ํ™•๋ณดํ•˜๋Š” ์ผ์€ ์‰ฝ์ง€ ์•Š๋‹ค. ๋”ฐ๋ผ์„œ ๋‚ด๋ถ€ ์ธ๋ ฅ์— ๋Œ€ํ•œ ๊ต์œก๊ณผ ์—ญ๋Ÿ‰ ๊ฐ•ํ™”๋Š” ํ•„์ˆ˜์ ์ด๋‹ค.

๋‹ค๋งŒ ๊ฐ€์žฅ ๋น ๋ฅธ ์„ฑ๊ณผ๋Š” ๊ฐ ์กฐ์ง์ด ๊ฐœ๋ณ„์ ์œผ๋กœ ์›€์ง์ด๊ธฐ๋ณด๋‹ค, ๊ธฐ๋Šฅ ๊ฐ„ ๊ฒฝ๊ณ„๋ฅผ ํ—ˆ๋ฌผ๊ณ  ๊ธด๋ฐ€ํ•˜๊ฒŒ ํ˜‘์—…ํ•˜๋Š” โ€˜AI ์Šค์ฟผ๋“œโ€™๋ฅผ ๊ตฌ์„ฑํ•  ๋•Œ ๋‚˜ํƒ€๋‚˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค.

AI-SaaS ์œตํ•ฉ์„ ์ด๋„๋Š” ๋ฆฌ๋”๋ฅผ ์œ„ํ•œ ์ „๋žต

๊ธฐ์—…์˜ SaaS ํ”Œ๋žซํผ ๊ณ ๋„ํ™”๋ฅผ ์ง€์›ํ•ด ์˜จ ๊ฒฝํ—˜์„ ๋ฐ”ํƒ•์œผ๋กœ, ํ•„์ž๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์ „๋žต ๋กœ๋“œ๋งต์„ ์ œ์•ˆํ•œ๋‹ค.

1. ๋ช…ํ™•ํ•˜๊ณ  ๋ถ€๊ฐ€๊ฐ€์น˜๊ฐ€ ๋†’์€ ์‚ฌ์šฉ๋ก€๋กœ ์‹œ์ž‘

๊ณ ๊ฐ ์ดํƒˆ์„ ์ค„์ด๊ฑฐ๋‚˜ ์˜จ๋ณด๋”ฉ์„ ๊ฐœ์„ ํ•˜๊ณ , ๊ณ ๊ฐ ์ง€์› ๋น„์šฉ์„ ์ ˆ๊ฐํ•˜๊ฑฐ๋‚˜ ์ธํ”„๋ผ๋ฅผ ์ตœ์ ํ™”ํ•˜๋Š” ๋“ฑ ์‹ค์งˆ์ ์ธ ํšจ๊ณผ๊ฐ€ ๊ธฐ๋Œ€๋˜๋Š” ์˜์—ญ์„ ์„ ํƒํ•˜๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•˜๋‹ค. ๋ชฉ์ ์ด ์—†๋Š”, โ€˜AI๋ฅผ ์œ„ํ•œ AIโ€™ ๋„์ž…์€ ํ”ผํ•ด์•ผ ํ•œ๋‹ค.

2. ๊ฐ•๋ ฅํ•œ ๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜ ๊ตฌ์ถ•

์ด ๋‹จ๊ณ„๋Š” ์ƒ๋žตํ•  ์ˆ˜ ์—†๋‹ค. ๋‹ค์Œ์— ํˆฌ์žํ•ด์•ผ ํ•œ๋‹ค.

  • ๋ฐ์ดํ„ฐ ํŒŒ์ดํ”„๋ผ์ธ
  • ๊ฑฐ๋ฒ„๋„Œ์Šค ๊ธฐ์ค€
  • ๋ณด์•ˆ ํ†ต์ œ ์ฒด๊ณ„
  • ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ์— ๋Œ€ํ•œ ์ฑ…์ž„ ๊ตฌ์กฐ

3. ์ž‘๊ณ  ์ธก์ • ๊ฐ€๋Šฅํ•œ ํŒŒ์ผ๋Ÿฟ ํ”„๋กœ๊ทธ๋žจ์„ ์‹œ์ž‘

์„ฑ๊ณต์€ ์ถ”์ง„๋ ฅ์„ ๋งŒ๋“ค๋ฉฐ, ์ž‘์€ ์‹คํŒจ๋Š” ๊ฒฝํ—˜๊ณผ ๊ตํ›ˆ์„ ๋‚ณ๋Š”๋‹ค.

4. ์œค๋ฆฌ์™€ ์ปดํ”Œ๋ผ์ด์–ธ์Šค๋ฅผ ์กฐ๊ธฐ์— ํ†ตํ•ฉ

์‹ ๋ขฐ๊ฐ€ ๋’ท๋ฐ›์นจ๋˜์ง€ ์•Š๋Š” AI๋Š” ์‹ค์งˆ์ ์œผ๋กœ ํ™œ์šฉ๋  ์ˆ˜ ์—†๋‹ค.

5. AI๋ฅผ ์ „์ œ๋กœ ์•„ํ‚คํ…์ฒ˜ ์žฌ์„ค๊ณ„

์ด ์ง€์ ์—์„œ ๋งŽ์€ ์กฐ์ง์ด ์–ด๋ ค์›€์„ ๊ฒช๋Š”๋‹ค. ์•ž์œผ๋กœ์˜ SaaS๋Š” ์‹ค์‹œ๊ฐ„ ๋ฐ์ดํ„ฐ ์ŠคํŠธ๋ฆผ, ๋ชจ๋ธ ๋ฐฐํฌ, ์ง€์†์ ์ธ ํ•™์Šต, ์ด๋ฒคํŠธ ๊ธฐ๋ฐ˜ ์ฒ˜๋ฆฌ๋ฅผ ์ˆ˜์šฉํ•  ์ˆ˜ ์žˆ์–ด์•ผ ํ•œ๋‹ค.

6. ์กฐ์ง ๊ฐ„ ํ˜‘์—…์— ์ง‘์ค‘

AI๋Š” ์—”์ง€๋‹ˆ์–ด๋ง ์กฐ์ง๋งŒ์˜ ๊ณผ์ œ๊ฐ€ ์•„๋‹ˆ๋‹ค. ์ œํ’ˆ, ๋””์ž์ธ, ๋ณด์•ˆ, ์ปดํ”Œ๋ผ์ด์–ธ์Šค, ๊ณ ๊ฐ ์„ฑ๊ณต ์กฐ์ง์˜ ์ฐธ์—ฌ๊ฐ€ ํ•จ๊ป˜ ์ด๋ค„์ ธ์•ผ ํ•œ๋‹ค.

7. AI ์šฐ์„  ์‚ฌ๊ณ ๋ฐฉ์‹์œผ๋กœ ์ „ํ™˜

์•ž์œผ๋กœ ์„ฑ๊ณตํ•˜๋Š” SaaS ์ œํ’ˆ์€ AI๋ฅผ ํ•˜๋‚˜์˜ ๊ธฐ๋Šฅ์ด ์•„๋‹ˆ๋ผ, ํ•ต์‹ฌ ์—ญ๋Ÿ‰์œผ๋กœ ๋‹ค๋ฃจ๊ฒŒ ๋  ๊ฒƒ์ด๋‹ค.

SaaS์˜ ๋ฏธ๋ž˜๋Š” ํด๋ผ์šฐ๋“œ ์ค‘์‹ฌ์ด ์•„๋‹ˆ๋‹ค

SaaS์™€ AI์˜ ์œตํ•ฉ์€ ๋จผ ๋ฏธ๋ž˜์˜ ์ด์•ผ๊ธฐ๊ฐ€ ์•„๋‹ˆ๋ผ ์ง€๊ธˆ ์ด ์ˆœ๊ฐ„์—๋„ ์ง„ํ–‰๋˜๊ณ  ์žˆ๋‹ค. ์ง€๋Šฅํ˜• ์ž๋™ํ™”, ์˜ˆ์ธก ์ธ์‚ฌ์ดํŠธ, ๊ฐœ์ธํ™”๊ฐ€ ์†Œํ”„ํŠธ์›จ์–ด ์ œ๊ณต ๋ฐฉ์‹์˜ ํ•ต์‹ฌ ์ถ•์œผ๋กœ ์ž๋ฆฌ ์žก๊ณ  ์žˆ๋‹ค.

AI๋Š” ๊ณ ๊ฐ ๊ฒฝํ—˜์„ ์žฌํŽธํ•˜๊ณ  ์šด์˜ ํšจ์œจ์„ฑ์„ ๋Œ์–ด์˜ฌ๋ฆฌ๋Š” ๋™์‹œ์—, ์ ์‘ํ˜• ์•„ํ‚คํ…์ฒ˜๋ฅผ ๊ตฌํ˜„ํ•˜๋ฉฐ SaaS ํ”Œ๋žซํผ์˜ ๋ฒ”์œ„๋ฅผ ํฌ๊ฒŒ ํ™•์žฅํ•˜๊ณ  ์žˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ๊ทธ ๊ณผ์ •์—์„œ ๋ฐ์ดํ„ฐ ์ค€๋น„๋„, ์œค๋ฆฌ์  ์ด์Šˆ, ์ƒˆ๋กœ์šด ์ธ์žฌ์ƒ์— ๋Œ€ํ•œ ์š”๊ตฌ๋ผ๋Š” ๊ณผ์ œ๋„ ํ•จ๊ป˜ ์ œ๊ธฐ๋œ๋‹ค.

ํ•„์ž๋Š” ์กฐ๊ธฐ์— AI์™€ SaaS๋ฅผ ์œตํ•ฉํ•˜๊ธฐ ์œ„ํ•ด ๋‚˜์„  ๊ธฐ์—…์ด ์ง€์†์ ์ธ ๊ฒฝ์Ÿ ์šฐ์œ„๋ฅผ ํ™•๋ณดํ•˜๋Š” ๋ชจ์Šต์„ ํ™•์ธํ•ด ์™”๋‹ค. ์ด๋“ค ๊ธฐ์—…์€ ๋” ๋น ๋ฅด๊ฒŒ ํ˜์‹ ํ•˜๊ณ , ๋” ๋งŽ์€ ๊ฐ€์น˜๋ฅผ ์ œ๊ณตํ•˜๋ฉฐ, ์‚ฌ์šฉ์ž์™€ ํ•จ๊ป˜ ์ง„ํ™”ํ•˜๋Š” ์ œํ’ˆ์„ ๋งŒ๋“ค๊ณ  ์žˆ๋‹ค.

SaaS์˜ ๋ฏธ๋ž˜๋Š” ํด๋ผ์šฐ๋“œ ์ค‘์‹ฌ์ด ์•„๋‹ˆ๋ผ AI ์ค‘์‹ฌ์ด๋‹ค. ์ด ๋ณ€ํ™”๋ฅผ ์ดํ•ดํ•˜๋Š” ๋ฆฌ๋”๊ฐ€ ํ–ฅํ›„ 10๋…„๊ฐ„์˜ ๋””์ง€ํ„ธ ์ „ํ™˜ ๋ฐฉํ–ฅ์„ ์ฃผ๋„ํ•˜๊ฒŒ ๋  ๊ฒƒ์ด๋‹ค.
dl-ciokorea@foundryco.com

The convergence of SaaS and AI: Trends, opportunities and challenges

More than 10 years ago, when I started stepping into the realm of software-as-a-service, the concept of SaaS seemed groundbreaking. The move towards cloud-based solutions altered the way companies utilize software, expanded their activities and controlled expenses. Yet recently I have observed a development thatโ€™s even more game-changing: the merging of SaaS with artificial intelligence.

AI is no longer an add-on feature or a buzzword sprinkled into slide decks. Itโ€™s becoming the backbone of how modern SaaS platforms operate, differentiate and grow. As the author of Get SaaS Insights Before You Invest Millions and as someone who has worked extensively with SaaS systems and AI-led transformations, Iโ€™ve seen firsthand how this convergence is reshaping products, business models and customer expectations.

In this article, Iโ€™ll break down the top trends, real opportunities and hidden challenges leaders need to understand as SaaS and AI rapidly fuse into the next generation of digital platforms.

AI is becoming the new foundation of SaaS

The significant change Iโ€™ve witnessed over the past three years is that AI is no longer viewed merely as a component or feature. It is evolving into a core capability. SaaS providers are reengineering their platforms to prioritize AI first in the cloud.

From workflow automation to intelligent automation

Earlier SaaS systems automated tasks. AI-powered SaaS systems automate decisions.

Capabilities such as:

  • Predictive analytics
  • Natural language processing
  • Behavior-based triggers
  • Self-healing systems
  • Context-aware recommendations

โ€ฆare becoming table stakes.

At an enterprise platform I contributed to, we transitioned from rule-based automations to AI-powered forecasts that detected system problems hours ahead of customer impact. This change decreased downtime, enhanced customer satisfaction and lowered emergency interventions.

AI is not merely enhancing SaaS; itโ€™s transforming the concept of efficiency itself.

Trend 1: Customization is turning into a requirement

All sectors โ€” from retail to healthcare โ€” personalization is becoming essential to maintain customer interest. SaaS products follow this trend well. Users today anticipate platforms to function similarly to Netflix or Spotify:

  • Tailored dashboards
  • Customized workflows
  • Intelligent suggestions
  • Adaptive interfaces based on usage patterns

I have directly observed how profoundly personalization influences SaaS adoption. In a learning platform, I consulted for implementing AI-powered learning paths boosted user engagement by 60% since users felt the product โ€œgot them.โ€

However, personalization introduces anticipations. Users expect more than software that functions โ€” They need software that suits their needs.

Trend 2: AI is reshaping the dynamics of SaaS

SaaS was appealing due to its subscription models and scalable infrastructure. AI introduces a level of value by facilitating:

Usage-based pricing

With advancements in AI for behavior monitoring and analytics, SaaS firms are able to set prices according to customer value. There is an increase in approaches:

  • Subscription + usage
  • Subscription + intelligence tier
  • Usage-only for AI-heavy features

This generates income possibilities but demands accuracy in comprehending customer actions โ€” a skill provided by AI.

AI-driven product-led growth

I have assisted teams in leveraging AI insights to enhance onboarding, emphasize โ€œaha moments,โ€ and decrease drop-offs. AI precisely determines when to prompt a user, what assistance to offer and when to direct them toward valuable features.

This significantly boosts growth income. Lowers attrition.

Trend 3: The rise of AI-native SaaS products

We are stepping into the age of AI-SaaS, where AI is fundamentally embedded in the value offered. These solutions are built from scratch with a focus on intelligence, forecasting and self-direction.

Instances consist of:

  • AI-powered CRMs
  • Autonomous security platforms
  • Predictive maintenance systems
  • AI-driven financial forecasting tools
  • Automated compliance engines

At present, when I assess a SaaS startup, I can readily distinguish among:

  • Software-as-a-Service integrated with AI.
  • Dependent SaaS on AI to operate
  • The upcoming generation of unicorns will originate from the latter.

Opportunity 1: Reimagining customer support with AI

Customer support was once among the expensive aspects of managing a SaaS product.

AI is currently revolutionizing it by:

  • Sentiment-aware chatbots
  • Automated issue classification
  • Predictive ticket routing
  • Auto-generated troubleshooting steps
  • Voice-to-text and NLP-based assistance

At a SaaS product I was involved with, incorporating an AI support assistant cut down the ticket backlog by 40% within the month. Clients got replies, and support staff were able to concentrate on complicated problems instead of routine questions.

Opportunity 2: Intelligent product analytics

AI is providing SaaS teams with insight into:

  • What characteristics influence uptake
  • Which behaviors result in customer attrition
  • The way users navigate the product
  • Which pricing approaches connect
  • Which teams are at risk of attrition

Conventional analytics described what occurred.

AI clarifies the reasons behind the event. Even predicts what will occur next.

With predictive analytics, SaaS leaders can forecast churn, spot feature bottlenecks, identify upsell opportunities and improve product-market fit with far greater accuracy.

Opportunity 3: Scalable, self-optimizing infrastructure

AI is transforming DevOps and infrastructure management through:

  • Auto-scaling
  • Anomaly detection
  • Predictive load-balancing
  • Automated deployment validation
  • Intelligent resource provisioning

In one platform, we implemented AI-based load forecasting that reduced infrastructure costs by over 20%. The system predicted resource spikes, scaled ahead of time and prevented performance drops that previously required manual intervention.

AI enables SaaS platforms that learn as they grow.

Challenge 1: Data quality and governance

AI requires data thatโ€™s clean, uniform and secure. SaaS companies frequently overlook this.

Iโ€™ve witnessed promising AI concepts collapse due to:

  • The data was incomplete
  • The information lacked labels
  • The systemโ€™s structure was not created with AI in mind
  • Access controls blocked model training

Organizations must treat data as a product โ€” not an afterthought.

Challenge 2: Bias, privacy and ethical AI

Customers are becoming increasingly cautious about:

  • How their data is used
  • How models make decisions
  • Whether algorithms are fair
  • How their privacy is protected

Rules such as GDPR and emerging AI-focused legislation require SaaS companies to prioritize ethics and compliance.

I have been required to revamp AI workflows to guarantee transparency, lessen bias and maintain audit trails. These measures demand time and resources. They foster trust โ€” an essential element for the success of any AI system.

Challenge 3: The skills gap

The merging of SaaS and AI requires a kind of expertise:

  • Engineers knowledgeable about ML
  • Product-savvy ML specialists
  • Product managers, with a grasp of data
  • Designers capable of creating AI-centric architectures

Finding this blend is difficult.

Training internally is critical.

However, the fastest successes usually arise from forming functional โ€œAI squadsโ€ that work closely together instead of functioning independently.

A strategic roadmap for leaders navigating AI-SaaS convergence

Drawing from my experience assisting organizations in updating their SaaS platforms here is the roadmap I suggest:

1. Start with a clear, high-value use case

Pick something meaningful:

  • Reduce churn
  • Improve onboarding
  • Cut support costs
  • Optimize infrastructure

Avoid โ€œAI for the sake of AI.โ€

2. Build a strong data foundation

This step cannot be bypassed. Invest in:

  • Data pipelines
  • Governance standards
  • Security controls
  • Data quality ownership

3. Launch small, measurable pilots

Success builds momentum.

Failures, when small, build learning.

4. Integrate ethics and compliance early

AI without trust is unusable.

5. Redesign your architecture for AI

This is the point at which numerous teams become halted. Upcoming SaaS needs to accommodate:

  • Live data streams
  • Model deployment
  • Continuous training
  • Event-driven processing

6. Focus on cross-functional collaboration

AI is not an engineering-only initiative. Participation is needed from:

  • Product
  • Design
  • Security
  • Compliance
  • Customer success

7. Shift your mindset to AI-first

In the future, successful SaaS products will treat AI as a core capability, not a feature.

The future of SaaS is not cloud-first

The convergence of SaaS and AI is not a distant future โ€” itโ€™s happening right now. We are entering a new era where intelligent automation, predictive insights and personalization are becoming fundamental pillars of software delivery.

From reshaping customer experiences to transforming operational efficiency and enabling adaptive architectures, AI is expanding what SaaS platforms can achieve. But it also brings challenges: data readiness, ethical concerns and new talent expectations.

In my work and research โ€” as the author of Get SaaS Insights Before You Invest Millions and through my published contributions to IEEE โ€” Iโ€™ve seen how organizations that embrace AI-SaaS convergence early gain a lasting competitive advantage. They innovate faster, deliver more value and build products that truly evolve with their users.

The future of SaaS is not cloud-first. Itโ€™s AI-first โ€” and the leaders who understand this shift will shape the next decade of digital transformation.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

โ€œ๋ฏธ๊ตญ ํด๋ผ์šฐ๋“œ ์˜์กด๋„ ๋‚ฎ์ถœ ๋ชฉ์ โ€ EU, ์˜คํ”ˆ์†Œ์Šค ๋ถ€๋ฌธ ๊ฐ•ํ™” ์ „๋žต ๋ชจ์ƒ‰

์œ ๋Ÿฝ์—ฐํ•ฉ ์ง‘ํ–‰์œ„์›ํšŒ(EC)๋Š” ์œ ๋Ÿฝ ์™ธ ์ง€์—ญ์˜ ๊ธฐ์ˆ  ๋ฒค๋”์— ๋Œ€ํ•œ ์˜์กด๋„๋ฅผ ์ค„์ด๊ธฐ ์œ„ํ•ด ์ž์ฒด ์˜คํ”ˆ์†Œ์Šค ์ƒํƒœ๊ณ„๋ฅผ ๊ฐ•ํ™”ํ•  ์˜๊ฒฌ์„ ๋ฐ›๊ณ  ์žˆ๋‹ค.

EC๋Š” โ€˜์œ ๋Ÿฝํ˜• ๊ฐœ๋ฐฉํ˜• ๋””์ง€ํ„ธ ์ƒํƒœ๊ณ„๋ฅผ ํ–ฅํ•˜์—ฌ(Towards European open digital ecosystems)โ€™๋ผ๋Š” ์ƒˆ๋กœ์šด ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ๋ฅผ ํ†ตํ•ด EU ์˜คํ”ˆ์†Œ์Šค ๋ถ€๋ฌธ์˜ ์„ฑ์žฅ๊ณผ ์ง€์† ๊ฐ€๋Šฅ์„ฑ์„ ์ง€์›ํ•  ๊ณ„ํš์ด๋‹ค. EC๋Š” ์˜คํ”ˆ์†Œ์Šค๋ฅผ ๋””์ง€ํ„ธ ์ฃผ๊ถŒ ์ „๋žต์˜ ํ•ต์‹ฌ ์ถ•์œผ๋กœ ๋ณด๊ณ  ์žˆ๋‹ค.

์ด๋ฒˆ ์ „๋žต์€ ๊ณต๊ณต๊ณผ ๋ฏผ๊ฐ„ ๋ถ€๋ฌธ ์ „๋ฐ˜์—์„œ์˜ ํ™œ์šฉ ํ™•๋Œ€์™€ ๊ฒฝ์Ÿ๋ ฅ ์ œ๊ณ , ์Šคํƒ€ํŠธ์—…์— ๋Œ€ํ•œ ์ง€์› ๊ฐ•ํ™”๋ฅผ ์ฃผ์š” ๋ชฉํ‘œ๋กœ ํ•œ๋‹ค. ํด๋ผ์šฐ๋“œ, AI, ๋ณด์•ˆ, ์˜คํ”ˆ ํ•˜๋“œ์›จ์–ด ๋“ฑ ๋‹ค์–‘ํ•œ ๋ถ„์•ผ๋ฅผ ํฌ๊ด„ํ•˜๋ฉฐ, EU์˜ ํ•ต์‹ฌ ์‚ฐ์—…์ธ ์ž๋™์ฐจ์™€ ์ œ์กฐ์—…์„ ์ค‘์‹ฌ์œผ๋กœ ์ถ”์ง„๋  ์˜ˆ์ •์ด๋‹ค.

์˜๊ฒฌ ์ˆ˜๋ ด์€ 2์›” 3์ผ๊นŒ์ง€ ์ง„ํ–‰๋˜๋ฉฐ, ๊ฐœ๋ฐœ์ž์™€ ๋ฒค๋”, ์—ฐ๊ตฌ์ž ๋“ฑ ์˜คํ”ˆ์†Œ์Šค ์ปค๋ฎค๋‹ˆํ‹ฐ ์ดํ•ด๊ด€๊ณ„์ž๋“ค์ด ์ฐธ์—ฌํ•ด ์˜๊ฒฌ์„ ์ œ์ถœํ•˜๋„๋ก ์ดˆ์ฒญ๋ฐ›์•˜๋‹ค.

EC๋Š” โ€œ๋””์ง€ํ„ธ ๋ถ„์•ผ์—์„œ ๋น„์œ ๋Ÿฝ ๊ตญ๊ฐ€์— ๋Œ€ํ•œ ์˜์กด๋„๊ฐ€ ๋†’๋‹ค๋Š” ์ค‘๋Œ€ํ•œ ๋ฌธ์ œ์— ์ง๋ฉดํ•ด ์žˆ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค. ์ด๋Š” ์œ ๋Ÿฝ ๋‚ด์—์„œ ๋ฏธ๊ตญ ํด๋ผ์šฐ๋“œ ์—…์ฒด๊ฐ€ ๊ด‘๋ฒ”์œ„ํ•˜๊ฒŒ ์‚ฌ์šฉ๋˜๊ณ  ์žˆ๋Š” ํ˜„์‹ค์„ ์ง€์นญํ•œ ๊ฒƒ์ด๋‹ค. ๋˜ํ•œ EC๋Š” โ€œ์˜คํ”ˆ์†Œ์Šค ๊ธฐ์ˆ ์ด ๋””์ง€ํ„ธ ์ธํ”„๋ผ์— ๋Œ€ํ•œ ํ†ต์ œ๋ ฅ์„ ํšŒ๋ณตํ•˜๊ณ  ํšŒ๋ณตํƒ„๋ ฅ์„ฑ์„ ๊ฐ•ํ™”ํ•˜๋Š” ๋ฐ ๊ธฐ์—ฌํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

EC๋Š” ์˜คํ”ˆ์†Œ์Šค๊ฐ€ ์ด๋ฏธ ๋Œ€๋ถ€๋ถ„์˜ ์†Œํ”„ํŠธ์›จ์–ด ์‹œ์Šคํ…œ์— ๊นŠ์ˆ™์ด ๋‚ด์žฌ๋ผ ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. EC๊ฐ€ ๊ณต๊ฐœํ•œ ํ•œ ์ถ”์ •์— ๋”ฐ๋ฅด๋ฉด ์ „์ฒด ์ฝ”๋“œ ๋ผ์ธ์˜ 70~90%๊ฐ€ ์˜คํ”ˆ์†Œ์Šค๋กœ ๊ตฌ์„ฑ๋ผ ์žˆ์œผ๋ฉฐ, ๊ฐœ๋ฐฉํ˜• AI ๋ชจ๋ธ์˜ ํ™œ์šฉ๋„๋„ ์ฆ๊ฐ€ํ•˜๊ณ  ์žˆ๋‹ค. ์‹ค์ œ๋กœ EU๋Š” ์„ธ๊ณ„ ์ตœ๋Œ€ ์ˆ˜์ค€์˜ ์˜คํ”ˆ์†Œ์Šค ๊ฐœ๋ฐœ์ž ์ปค๋ฎค๋‹ˆํ‹ฐ ์ƒํƒœ๊ณ„๋ฅผ ๋ณด์œ ํ•˜๊ณ  ์žˆ์ง€๋งŒ, ์‹ค์ œ๋กœ ์ฐฝ์ถœ๋˜๋Š” ๊ฐ€์น˜๋Š” ์ฃผ๋กœ ์œ ๋Ÿฝ ์™ธ๋ถ€์˜ ๊ธฐ์ˆ  ๋Œ€๊ธฐ์—…์ด ๊ฐ€์ ธ๊ฐ€๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค.

EC๋Š” โ€œEU ์ดํ•ด๊ด€๊ณ„์ž๋“ค์ด ๊ณต๊ณต ์กฐ๋‹ฌ ์‹œ์žฅ๊ณผ ๋ฏผ๊ฐ„ ์‹œ์žฅ ๋ชจ๋‘์—์„œ ์ง€๋ฐฐ์  ์‚ฌ์—…์ž์˜ ๋†’์€ ์ง„์ž… ์žฅ๋ฒฝ๊ณผ ๋„คํŠธ์›Œํฌ ํšจ๊ณผ๋กœ ์ธํ•ด ์ „๋ฐ˜์ ์œผ๋กœ ์–ด๋ ค์›€์„ ๊ฒช๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ํ‰๊ฐ€ํ–ˆ๋‹ค.

EU๋Š” ๊ทธ๋™์•ˆ ์ฐจ์„ธ๋Œ€ ์ธํ„ฐ๋„ท๊ณผ โ€˜GenAI4EUโ€™ ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ ๋“ฑ์„ ํ†ตํ•ด ์˜คํ”ˆ์†Œ์Šค ์ง€์›์— ํˆฌ์žํ•ด ์™”๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ EC๋Š” ๊ธฐ์กด ์—ฐ๊ตฌยทํ˜์‹  ํ”„๋กœ๊ทธ๋žจ๋งŒ์œผ๋กœ๋Š” ์ถฉ๋ถ„ํ•˜์ง€ ์•Š๋‹ค๋ฉฐ, ์˜คํ”ˆ์†Œ์Šค ์ปค๋ฎค๋‹ˆํ‹ฐ๋ฅผ ํ™•์žฅํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์ง€์† ๊ฐ€๋Šฅํ•œ ์ง€์›๊ณผ ๊ฑฐ๋ฒ„๋„Œ์Šค ํ”„๋ ˆ์ž„์›Œํฌ๊ฐ€ ํ•„์š”ํ•˜๋‹ค๊ณ  ๋ณด๊ณ  ์žˆ๋‹ค.

EC๋Š” ์ƒˆ๋กœ์šด ๊ณ„ํš์ด ์ง€์—ญ ์„ฑ์žฅ์„ ์ด‰์ง„ํ•˜๊ธฐ ์œ„ํ•œ ์žฌ์ • ์ง€์›๊ณผ ์ •์ฑ… ์ˆ˜๋‹จ์„ ๊ฒฐํ•ฉํ•˜๋Š” ํ˜•ํƒœ๊ฐ€ ๋  ๊ฒƒ์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด์™€ ํ•จ๊ป˜ 2020~2023๋…„ ์˜คํ”ˆ์†Œ์Šค ์ „๋žต์— ๋Œ€ํ•œ ์žฌ๊ฒ€ํ† ๋„ ์ง„ํ–‰ํ•  ์˜ˆ์ •์ด๋‹ค.

๋ฏธ๊ตญ ์ค‘์‹ฌ์˜ ๊ธฐ์ˆ  ํŒจ๊ถŒ ๊ฒฝ์Ÿ๊ณผ ๋ฐ์ดํ„ฐ ์ฃผ๊ถŒ์— ๋Œ€ํ•œ ์šฐ๋ ค ๋“ฑ ์ง€์ •ํ•™์  ๊ธด์žฅ์ด ๋†’์•„์ง€๋ฉด์„œ ์˜คํ”ˆ์†Œ์Šค ๊ธฐ์ˆ ์€ ์œ ๋Ÿฝ ๋‚ด ๊ธฐ๊ด€ ์‚ฌ์ด์—์„œ ๋ฏธ๊ตญ ํด๋ผ์šฐ๋“œ ์—…์ฒด์˜ ๋Œ€์•ˆ์œผ๋กœ ์ ์  ๋” ์ฃผ๋ชฉ๋ฐ›๊ณ  ์žˆ๋‹ค. ๊ฐ€ํŠธ๋„ˆ์˜ ์ตœ๊ทผ ์„ค๋ฌธ์กฐ์‚ฌ์— ๋”ฐ๋ฅด๋ฉด CIO์™€ IT ๋ฆฌ๋”์˜ 55%๊ฐ€ ์ง€์ •ํ•™์  ์š”์ธ์œผ๋กœ ์ธํ•ด ์ƒˆ๋กœ์šด ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์„ ํƒํ•  ๋•Œ ์˜คํ”ˆ์†Œ์Šค๊ฐ€ ๋” ์ค‘์š”ํ•œ ๊ธฐ์ค€์ด ๋๋‹ค๊ณ  ์‘๋‹ตํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

11 perfiles profesionales relacionados con la nube muy demandados por las empresas

Es una evidencia: las organizaciones no cejan de invertir con fuerza en soluciones en la nube de cara a 2026. El objetivo es claro segรบn el informe 2025 CIO Cloud Computing Study de Foundry: mejorar la productividad de los empleados en el 31% de los casos, mejorar la seguridad y la gobernanza en la organizaciรณn en el 30%, y acelerar la adopciรณn de la IA y el machine learning en el 13%.

La encuesta revela que el 70% de los responsables de la toma de decisiones de TI estรก de acuerdo en que su organizaciรณn ha acelerado la migraciรณn a la nube en los รบltimos 12 meses, en comparaciรณn con el 62% en 2024 y el 57% en 2023. Ademรกs, el 70% reconoce que su organizaciรณn opta por defecto por los servicios basados en la nube cuando actualiza o adquiere nueva tecnologรญa, y el 71% afirma que las inversiones en la nube han ayudado a su organizaciรณn a mantener un aumento de los ingresos en los รบltimos 12 meses.

Este crecimiento en la adopciรณn de la nube tambiรฉn ha provocado un aumento de la demanda de determinados puestos relacionados con la nube. A continuaciรณn, y sirviรฉndonos de la investigaciรณn de Foundry, le detallamos algunos de los que las empresas probablemente hayan aรฑadido para respaldar sus inversiones en la nube.

Arquitecto de seguridad

Los arquitectos de seguridad se encargan de crear, diseรฑar e implementar soluciones de seguridad en la organizaciรณn para mantener la seguridad de la infraestructura de TI. Para los que trabajan en un entorno de nube, el objetivo es diseรฑar e implementar soluciones de seguridad que protejan la infraestructura, los datos y las aplicaciones basados en dicho entorno.

Habilidades: diseรฑo de arquitectura de seguridad, seguridad de redes, cumplimiento y gobernanza de la seguridad, respuesta a incidentes y anรกlisis forense, cifrado de datos, gestiรณn de identidades y accesos (IAM), automatizaciรณn y DevSecOps.

Crecimiento del puesto: el 22% de las empresas ha aรฑadido puestos de arquitecto de seguridad como parte de sus inversiones en la nube.

Administrador de sistemas en la nube

Los administradores de sistemas en la nube se encargan de supervisar el mantenimiento y la gestiรณn general de la infraestructura en la nube. Estos profesionales de TI son expertos en navegar por entornos virtualizados ya sea implementando polรญticas basadas en la nube, desplegando parches y actualizaciones o analizando el rendimiento de la red.

Habilidades: conocimiento de la implementaciรณn e integraciรณn, la seguridad y la configuraciรณn, asรญ como de las herramientas de software en la nube mรกs populares, como Azure, AWS, GCP, Exchange y Office 365.

Crecimiento del puesto: el 22% de las empresas cuenta con nuevos puestos de administrador de sistemas en la nube como parte de sus inversiones en la nube.

Arquitecto de datos

La funciรณn de un arquitecto de datos es asegurarse de que los datos de una organizaciรณn estรฉn estructurados de manera que se pueda acceder a ellos fรกcilmente, estรฉn protegidos y se almacenen de manera eficiente, y satisfagan las necesidades empresariales. Los datos se han convertido en la principal forma en que las empresas realizan anรกlisis y ayudan en la toma de decisiones empresariales, y la mayor parte de esos datos se almacenan ahora en la nube.

Habilidades: almacenamiento de datos, escalabilidad y optimizaciรณn del rendimiento, automatizaciรณn y virtualizaciรณn, gobernanza de datos y seguridad en la nube, migraciรณn de datos y conocimiento de soluciones de nube hรญbrida.

Crecimiento del puesto: el 22% de las empresas ha incorporado puestos de arquitecto de datos como parte de sus inversiones en la nube.

Gerente de gobernanza y cumplimiento normativo en la nube

Los gestores de gobernanza y cumplimiento normativo en la nube ayudan a las empresas a navegar por las complejidades de la seguridad, la gobernanza, la normativa internacional y las polรญticas internas. Se encargan de identificar los riesgos potenciales, implementan herramientas automatizadas para supervisar la seguridad y el cumplimiento normativo, y ayudan a las empresas a mantener operaciones seguras en la nube.

Habilidades: sรณlidos conocimientos de polรญticas normativas como el RGPD, la HIPAA, el PCI DSS y otras leyes internacionales de protecciรณn de datos. Otras habilidades adicionales incluyen el conocimiento de herramientas como CSPM, Azure, AWS, Microsoft Purview Compliance Manager y otras herramientas de gobernanza de TI.

Crecimiento del puesto: el 20% de las empresas ha incorporado puestos de gestor de cumplimiento normativo y gobernanza en la nube como parte de sus inversiones en la nube.

Ingeniero de seguridad

Los ingenieros de seguridad se encargan de supervisar la seguridad de los sistemas, las redes y los datos de una organizaciรณn para garantizar que estรฉn protegidos contra las amenazas de ciberseguridad. Se trata de perfiles que pueden ayudar a las organizaciones que invierten en la nube a garantizar que los servicios, las aplicaciones y los datos que se ejecutan en plataformas en la nube sean seguros y cumplan con cualquier normativa gubernamental.

Habilidades: seguridad de redes, IAM, cifrado, gestiรณn de vulnerabilidades, arquitectura de seguridad, seguridad en la nube, automatizaciรณn y diseรฑo y optimizaciรณn de infraestructuras.

Crecimiento del puesto: el 19% de las empresas ha aรฑadido puestos de ingeniero de seguridad como parte de sus inversiones en la nube.

Gerente de productos en la nube

La adopciรณn de la nube suele ir acompaรฑada de un aumento del desarrollo interno de servicios basados en la nube. Un gerente de productos en la nube puede ayudar a los equipos de la nube a desarrollar soluciones eficaces destinadas a cumplir los objetivos empresariales. Tambiรฉn se encargan de utilizar su profundo conocimiento de la gestiรณn de productos en el entorno de la nube para trabajar en estrecha colaboraciรณn con las partes interesadas que son clave en el negocio, identificar y definir los requisitos de los usuarios o clientes, desarrollar hojas de ruta de productos y supervisar el proceso de control de calidad para obtener comentarios sobre cรณmo mejorar la oferta de productos.

Habilidades: gestiรณn de productos, diseรฑo de UX, comunicaciรณn y colaboraciรณn, y una sรณlida formaciรณn tรฉcnica.

Crecimiento del puesto: el 19% de las organizaciones dispone de nuevos puestos de gestor de productos en la nube como parte de sus inversiones en la nube.

Consultor de nube

Con la rรกpida adopciรณn y migraciรณn a la nube, las organizaciones buscan profesionales que puedan aprovechar las tecnologรญas de nube para satisfacer las necesidades empresariales, hacer crecer el negocio y mejorar la eficiencia. Estos profesionales son expertos en nube y se mantienen al dรญa de las รบltimas innovaciones en tecnologรญa de nube para asesorar mejor a los lรญderes empresariales.

Habilidades: conocimientos de arquitectura y diseรฑo de soluciones, DevOps, automatizaciรณn, gestiรณn de proyectos, seguridad en la nube, cumplimiento normativo, migraciรณn a la nube y conocimientos de las plataformas de nube mรกs populares.

Crecimiento del puesto: el 18% de las empresas ha incorporado consultores de nube como parte de sus inversiones en nube.

Ingeniero de DevOps

DevOps se centra en combinar las operaciones de TI con el proceso de desarrollo para mejorar los sistemas de TI y actuar como intermediario en el mantenimiento del flujo de comunicaciรณn entre los equipos de codificaciรณn e ingenierรญa. Es un puesto que se centra en la implementaciรณn de aplicaciones automatizadas, el mantenimiento de la infraestructura de TI y nube, y la identificaciรณn de los posibles riesgos y beneficios de los nuevos programas y sistemas.

Habilidades: automatizaciรณn, Linux, pruebas de control de calidad, seguridad, contenedorizaciรณn y conocimientos de lenguajes de programaciรณn como Java y Ruby.

Crecimiento del puesto: el 17% de las empresas cuenta ya con nuevos puestos de ingeniero DevOps como parte de sus inversiones en la nube.

Profesional de FinOps/optimizaciรณn de costes en la nube

Los profesionales de FinOps y optimizaciรณn de costes en la nube combinan conocimientos de finanzas, tecnologรญa y negocios para ayudar a supervisar el panorama cada vez mรกs complejo de las inversiones en la nube. La computaciรณn en la nube es parte integral de la IA, por lo que, a medida que mรกs organizaciones invierten en ella, tambiรฉn estรกn revisando sus inversiones en infraestructura en la nube. Los profesionales de FinOps y optimizaciรณn de costes en la nube pueden ayudar a las organizaciones a tomar las decisiones financieras adecuadas en torno a las inversiones en tecnologรญa que afectarรกn al negocio.

Habilidades: conocimientos de finanzas, negocios y tecnologรญa, junto con habilidades en el uso de herramientas y plataformas como AWS, Azure, GCP y plataformas FinOps nativas de la nube.

Crecimiento del puesto: el 16% de las empresas ha incorporado puestos de profesional de FinOps y optimizaciรณn de costes en la nube como parte de sus inversiones en la nube.

Responsable de FinOps/gerente de FinOps

Los responsables y gerentes de FinOps se encargan de supervisar la intersecciรณn entre la ingenierรญa, las finanzas y los negocios. A medida que mรกs organizaciones crean servicios y herramientas en la nube, buscan profesionales de FinOps con conocimientos tรฉcnicos que les ayuden a salvar la brecha entre las finanzas y la tecnologรญa, que aporten mejores ideas sobre cรณmo reducir costes y ajustarse al presupuesto, al tiempo que se implementan tecnologรญas innovadoras.

Habilidades: sรณlidos conocimientos de ingenierรญa, finanzas y tecnologรญa. Otras habilidades adicionales incluyen conocimientos de plataformas en la nube, habilidades bรกsicas de codificaciรณn y anรกlisis de datos.

Crecimiento del puesto: el 15% de las empresas ha incorporado puestos de responsable de FinOps y director de FinOps como parte de sus inversiones en la nube.

Ingeniero de fiabilidad del sitio (SRE)

Cualquier organizaciรณn que implemente estrategias en la nube presta especial atenciรณn a la fiabilidad y la escalabilidad, con lo que se garantizan el acceso a los datos desde la nube y bajo demanda, segรบn sea necesario. Los ingenieros de fiabilidad del sitio son responsables de supervisar la automatizaciรณn de la infraestructura de TI, la supervisiรณn de aplicaciones y la gestiรณn de sistemas. La infraestructura en la nube requiere actualizaciones frecuentes de software, y los servicios deben poder escalarse con el crecimiento de la organizaciรณn.

Habilidades: gestiรณn del cambio, gestiรณn de la infraestructura de TI, respuesta a incidentes de emergencia, mejora de procesos y supervisiรณn de aplicaciones.

Crecimiento del puesto: el 10% de las empresas ha aรฑadido puestos de ingeniero de fiabilidad del sitio como parte de sus inversiones en la nube.

ํด๋ผ์šฐ๋“œ ์šด์˜์˜ ๋™๋ฐ˜์ž, MCSP์˜ ์žฅ์ ๊ณผ ํ•œ๊ณ„๋Š”?

๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค ์ œ๊ณต์—…์ฒด(Managed Cloud Services Provider, MCSP)๋Š” ๊ธฐ์—…์ด ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์˜ ์ผ๋ถ€ ๋˜๋Š” ์ „๋ฐ˜์„ ์šด์˜ํ•˜๋Š” ๋ฐ ๋„์›€์„ ์ฃผ๋Š” ์—ญํ• ์„ ํ•œ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์‹œ์Šคํ…œ์˜ ํด๋ผ์šฐ๋“œ ์ด์ „, ๋ชจ๋‹ˆํ„ฐ๋ง๊ณผ ์œ ์ง€ ๊ด€๋ฆฌ, ์„ฑ๋Šฅ ๊ฐœ์„ , ๋ณด์•ˆ ๋„๊ตฌ ์šด์˜, ๋น„์šฉ ํ†ต์ œ ์ง€์› ๋“ฑ์ด ํฌํ•จ๋œ๋‹ค. MCSP๋Š” ์ผ๋ฐ˜์ ์œผ๋กœ ํผ๋ธ”๋ฆญ, ํ”„๋ผ์ด๋น—, ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ ์ „๋ฐ˜์—์„œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•œ๋‹ค.

๊ธฐ์—…์€ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ ๊ฐ€์šด๋ฐ ์–ด๋–ค ์˜์—ญ์„ ์ œ๊ณต์—…์ฒด์— ๋งก๊ธฐ๊ณ , ์–ด๋–ค ๋ถ€๋ถ„์„ ๋‚ด๋ถ€์—์„œ ์ง์ ‘ ์šด์˜ํ• ์ง€๋ฅผ ๊ฒฐ์ •ํ•œ๋‹ค. ๋Œ€๋ถ€๋ถ„์˜ ๊ฒฝ์šฐ ๊ธฐ์—…๊ณผ MCSP๋Š” ์ฑ…์ž„์„ ๊ณต์œ ํ•˜๋Š” ๊ตฌ์กฐ๋‹ค. ์ œ๊ณต์—…์ฒด๋Š” ์ผ์ƒ์ ์ธ ์šด์˜๊ณผ ๋„๊ตฌ ๊ด€๋ฆฌ๋ฅผ ๋‹ด๋‹นํ•˜๊ณ , ๊ธฐ์—…์€ ๋น„์ฆˆ๋‹ˆ์Šค ์˜์‚ฌ๊ฒฐ์ •๊ณผ ๋ฐ์ดํ„ฐ, ๊ฑฐ๋ฒ„๋„Œ์Šค์— ๋Œ€ํ•œ ์ฑ…์ž„์„ ์œ ์ง€ํ•œ๋‹ค.

์‚ฌ์ด๋ฒ„๋ณด์•ˆ ์ปจ์„คํŒ… ๊ธฐ์—… ์‚ฌ์ด์—‘์…€(CyXcel)์˜ ๋ถ๋ฏธ ๋””์ง€ํ„ธ ํฌ๋ Œ์‹ ๋ฐ ์‚ฌ๊ณ  ๋Œ€์‘ ๋ถ€๋ฌธ MCSP ๋ถ€์‚ฌ์žฅ์ธ ๋ธŒ๋ ŒํŠธ ๋ผ์ผ๋ฆฌ๋Š” MCSP๋ฅผ ์„ ํƒํ•˜๋Š” ๊ณผ์ •์ด ์–ธ์ œ๋‚˜ ๋ถ€๋‹ด์Šค๋Ÿฝ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ผ์ผ๋ฆฌ๋Š” โ€œ์„œ๋น„์Šค ์ˆ˜์ค€ ๊ณ„์•ฝ(SLA)์— ๋ช…์‹œ๋œ ์ˆ˜์ค€์œผ๋กœ ์„œ๋น„์Šค๋ฅผ ์ˆ˜ํ–‰ํ•  ๊ฒƒ์ด๋ผ๋Š” ์‹ ๋ขฐ์— ํฌ๊ฒŒ ์˜์กดํ•˜์ง€๋งŒ, ์‹ค์ œ๋กœ ์ด๋ฅผ ์ถฉ์กฑํ•˜๊ณ  ์žˆ๋Š”์ง€๋Š” ์žฅ์• ๋‚˜ ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ ์‚ฌ๊ณ ๊ฐ€ ๋ฐœ์ƒํ•ด ๋ฌธ์ œ๊ฐ€ ๋“œ๋Ÿฌ๋‚˜๊ธฐ ์ „๊นŒ์ง€ ๊ฒ€์ฆํ•˜๊ธฐ ์–ด๋ ต๋‹คโ€๋ผ๋ฉฐ โ€œ๊ทธ ์‹œ์ ์—๋Š” ์ด๋ฏธ ํ”ผํ•ด๊ฐ€ ๋ฐœ์ƒํ•œ ๋’ค์ธ ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹คโ€๋ผ๊ณ  ์ „ํ–ˆ๋‹ค. ๊ทธ๋Š” ๋˜ โ€œMCSP๋Š” ์ ๊ฒ€ํ•  ์ˆ˜ ์žˆ๋Š” ๋ฌผ๋ฆฌ์  ์ธํ”„๋ผ๊ฐ€ ์—†๊ณ , ์˜จํ”„๋ ˆ๋ฏธ์Šค ํ™˜๊ฒฝ์ฒ˜๋Ÿผ ๋ˆˆ์— ๋ณด์ด๋Š” ์ž‘์—…๋„ ์—†์–ด ํ‰๊ฐ€์™€ ์„ ํƒ์ด ๋”์šฑ ๊นŒ๋‹ค๋กญ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

MCSP์˜ ์žฅ์ 

์šด์˜ ๋ถ€๋‹ด ๊ฐ์†Œ: MCSP๋Š” ์ผ์ƒ์ ์ธ ํด๋ผ์šฐ๋“œ ๊ด€๋ฆฌ ์—…๋ฌด๋ฅผ ๋Œ€์‹  ์ˆ˜ํ–‰ํ•ด ๋‚ด๋ถ€์— ๋Œ€๊ทœ๋ชจ ํด๋ผ์šฐ๋“œยท์ธํ”„๋ผ ์กฐ์ง์„ ์œ ์ง€ํ•ด์•ผ ํ•˜๋Š” ๋ถ€๋‹ด์„ ์ค„์—ฌ์ค€๋‹ค. ํŠนํžˆ ๋‚ด๋ถ€์— ํด๋ผ์šฐ๋“œ๋‚˜ ํ•€์˜ต์Šค(FinOps) ์ „๋ฌธ์„ฑ์ด ์ถฉ๋ถ„ํ•˜์ง€ ์•Š์€ ์กฐ์ง์— ํšจ๊ณผ์ ์ด๋‹ค.

์‹ ์†ํ•œ ๋ฌธ์ œ ๋Œ€์‘ : ๋Œ€๋ถ€๋ถ„์˜ MCSP๋Š” 24์‹œ๊ฐ„ ๋ชจ๋‹ˆํ„ฐ๋ง๊ณผ ์ง€์› ์ฒด๊ณ„๋ฅผ ์ œ๊ณตํ•œ๋‹ค. ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•˜๋ฉด ์‚ฌ์šฉ์ž๋‚˜ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์— ํฐ ์˜ํ–ฅ์„ ๋ฏธ์น˜๊ธฐ ์ „์— ๋น ๋ฅด๊ฒŒ ๋Œ€์‘ํ•  ์ˆ˜ ์žˆ๋‹ค.

์žฌํ•ด ๋ณต๊ตฌ์™€ ๋ณต์›๋ ฅ ์ง€์› : MCSP๋Š” ๋ฐฑ์—…๊ณผ ์žฌํ•ด ๋ณต๊ตฌ ํ™˜๊ฒฝ์˜ ์„ค๊ณ„, ์šด์˜, ํ…Œ์ŠคํŠธ๋ฅผ ์ง€์›ํ•œ๋‹ค. ๋ณต๊ตฌ ๋ชฉํ‘œ๋Š” ๊ณ ๊ฐ์ด ์ •์˜ํ•˜์ง€๋งŒ, ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ–ˆ์„ ๋•Œ ์‹œ์Šคํ…œ์„ ์‹ ์†ํ•˜๊ฒŒ ๋ณต๊ตฌํ•  ์ˆ˜ ์žˆ๋„๋ก ๋•๋Š” ์—ญํ• ์„ ๋งก๋Š”๋‹ค.

์ง€์†์ ์ธ ํ”Œ๋žซํผ ๊ด€๋ฆฌ : ํด๋ผ์šฐ๋“œ ํ”Œ๋žซํผ์€ ๋ณ€ํ™” ์†๋„๊ฐ€ ๋น ๋ฅด๋‹ค. MCSP๋Š” ์ธํ”„๋ผ ๊ตฌ์„ฑ ์š”์†Œ๋ฅผ ์ตœ์‹  ์ƒํƒœ๋กœ ์œ ์ง€ํ•˜๊ณ  ํ˜ธํ™˜์„ฑ์„ ๊ด€๋ฆฌํ•ด, ์˜ค๋ž˜๋œ ์„ค์ •์œผ๋กœ ์ธํ•œ ์œ„ํ—˜์„ ์ค„์ด๋Š” ๋™์‹œ์— ์ฃผ์š” ๋ณ€๊ฒฝ ์‹œ์ ์— ๋Œ€ํ•œ ํ†ต์ œ๊ถŒ์€ ๊ณ ๊ฐ์ด ์œ ์ง€ํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•œ๋‹ค.

๋ณด์•ˆ ์ „๋ฌธ์„ฑ๊ณผ ๋„๊ตฌ ์ œ๊ณต : ํด๋ผ์šฐ๋“œ ๋ณด์•ˆ์—๋Š” ์ˆ˜์š”๊ฐ€ ๋†’์€ ์ „๋ฌธ ์—ญ๋Ÿ‰์ด ์š”๊ตฌ๋œ๋‹ค. MCSP๋Š” ์•„์ด๋ดํ‹ฐํ‹ฐ ๊ด€๋ฆฌ, ๋ชจ๋‹ˆํ„ฐ๋ง, ๊ทœ์ • ์ค€์ˆ˜ ๋„๊ตฌ, ๋ณด์•ˆ ๋ชจ๋ฒ” ์‚ฌ๋ก€์— ๋Œ€ํ•œ ๊ฒฝํ—˜์„ ๋ฐ”ํƒ•์œผ๋กœ ์ผ์ƒ์ ์ธ ๋ณด์•ˆ ์ˆ˜์ค€ ๊ฐ•ํ™”๋ฅผ ์ง€์›ํ•œ๋‹ค. ๋ณด์•ˆ ์ฑ…์ž„์€ ์—ฌ์ „ํžˆ ๊ธฐ์—…๊ณผ ์ œ๊ณต์—…์ฒด๊ฐ€ ๊ณต์œ ํ•œ๋‹ค.

์‹ ๋ขฐ์„ฑ๊ณผ ์„ฑ๋Šฅ ํ–ฅ์ƒ : ๋Œ€๊ทœ๋ชจ์ด๋ฉด์„œ ๋ณต์žกํ•œ ํ™˜๊ฒฝ์„ ์šด์˜ํ•ด ์˜จ ๊ฒฝํ—˜์„ ๋ฐ”ํƒ•์œผ๋กœ, ๋ณด๋‹ค ์•ˆ์ •์ ์ด๊ณ  ํ™•์žฅ ๊ฐ€๋Šฅํ•˜๋ฉฐ ๋ณต์›๋ ฅ ์žˆ๋Š” ํด๋ผ์šฐ๋“œ ์ธํ”„๋ผ์˜ ์„ค๊ณ„์™€ ์šด์˜์„ ์ง€์›ํ•œ๋‹ค.

๊ธฐ์กด ์‹œ์Šคํ…œ๊ณผ์˜ ํ†ตํ•ฉ : MCSP๋Š” ํด๋ผ์šฐ๋“œ ์ž์›์„ ์˜จํ”„๋ ˆ๋ฏธ์Šค ์‹œ์Šคํ…œ, ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜, ์•„์ด๋ดํ‹ฐํ‹ฐ ํ”Œ๋žซํผ๊ณผ ์—ฐ๊ณ„ํ•ด ์‚ฌ์šฉ์ž์™€ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์ด ์ค‘๋‹จ ์—†์ด ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์ด์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•œ๋‹ค.

๋น„์šฉ ์ ˆ๊ฐ๋ณด๋‹ค๋Š” ์˜ˆ์ธก ๊ฐ€๋Šฅํ•œ ์šด์˜ : MCSP๋ฅผ ํ™œ์šฉํ•˜๋ฉด ๋‚ด๋ถ€ ์ธ๋ ฅ๊ณผ ๋„๊ตฌ ๋น„์šฉ์„ ์ค„์ผ ์ˆ˜ ์žˆ์ง€๋งŒ, ์ „์ฒด ํด๋ผ์šฐ๋“œ ์ง€์ถœ์ด ํ•ญ์ƒ ๊ฐ์†Œํ•˜๋Š” ๊ฒƒ์€ ์•„๋‹ˆ๋‹ค. ํ˜„์žฌ MCSP์˜ ๊ฐ€์น˜๋Š” ์ €๋ ดํ•œ ํด๋ผ์šฐ๋“œ ์š”๊ธˆ๋ณด๋‹ค๋Š” ์šด์˜ ํšจ์œจ์„ฑ, ์ „๋ฌธ์„ฑ, ๋Œ€์‘ ์†๋„์— ๋” ์žˆ๋‹ค.

MCSP ์„ ํƒ ์‹œ ํ•ต์‹ฌ ๊ณ ๋ ค ์‚ฌํ•ญ

IT ๊ด€๋ฆฌ ์†Œํ”„ํŠธ์›จ์–ด ์ œ๊ณต์—…์ฒด ์ปค๋„ฅํŠธ์™€์ด์ฆˆ(ConnectWise)์˜ ์ตœ๊ณ ๊ฒฝ์˜์ž ๋งค๋‹ˆ ๋ฆฌ๋ฒจ๋กœ๋Š” ์กฐ์ง์ด ์ ์  ๋” ์ž์œจ์ ์ด๊ณ  AI ๊ธฐ๋ฐ˜ ์„œ๋น„์Šค๋กœ ์ „ํ™˜ํ•˜๋Š” ๊ณผ์ •์—์„œ, MCSP๊ฐ€ ์ž๋™ํ™”๋ฅผ ์‹ค์ œ ์ผ์ƒ ์šด์˜์—์„œ ์ œ๋Œ€๋กœ ์ž‘๋™ํ•˜๋„๋ก ๋งŒ๋“œ๋Š” ์ค‘์š”ํ•œ ์—ญํ• ์„ ํ•œ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ฆฌ๋ฒจ๋กœ๋Š” ๋งŽ์€ ์กฐ์ง์ด ์˜ˆ์ƒ๋ณด๋‹ค ์ค‘์š”ํ•˜๊ฒŒ ์ธ์‹ํ•˜์ง€ ๋ชปํ•˜๋Š” ์š”์†Œ๋กœ ์šด์˜ ํˆฌ๋ช…์„ฑ์„ ๊ผฝ์•˜๋‹ค. ๊ทธ๋Š” ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์ด ์–ด๋–ป๊ฒŒ ์„ค๊ณ„๋˜๊ณ , ๋ณด์•ˆ์ด ์ ์šฉ๋˜๋ฉฐ, ์šด์˜๋˜๊ณ  ์žˆ๋Š”์ง€์— ๋Œ€ํ•œ ๋ช…ํ™•ํ•œ ๊ฐ€์‹œ์„ฑ์ด ํ•„์š”ํ•˜๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์•„์šธ๋Ÿฌ ์—์ด์ „ํ‹ฑ AI๊ฐ€ ์‹œ์Šคํ…œ์„ ์–ด๋–ป๊ฒŒ ๋ชจ๋‹ˆํ„ฐ๋งํ•˜๊ณ  ์˜์‚ฌ๊ฒฐ์ •์„ ๋‚ด๋ฆฌ๋ฉฐ ์‹ค์ œ ์กฐ์น˜๋ฅผ ์ทจํ•˜๋Š”์ง€๊นŒ์ง€ ์กฐ์ง์ด ์ดํ•ดํ•˜๊ณ  ์žˆ์–ด์•ผ, ์ธ์ง€ํ•˜์ง€ ๋ชปํ•œ ์ƒํƒœ์—์„œ ์ค‘์š”ํ•œ ์ผ์ด ์ง„ํ–‰๋˜๋Š” ์ƒํ™ฉ์„ ๋ง‰์„ ์ˆ˜ ์žˆ๋‹ค๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

๋ฆฌ๋ฒจ๋กœ๋Š” โ€œ์ž์œจ์„ฑ์ด ๋†’์•„์งˆ์ˆ˜๋ก ์šด์˜ ์„ฑ์ˆ™๋„์˜ ์ค‘์š”์„ฑ๋„ ์ปค์ง„๋‹คโ€๋ผ๋ฉฐ โ€œ์—ฌ๊ธฐ์—๋Š” ์ฒด๊ณ„์ ์ธ ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค, ๊ฐ•๋ ฅํ•œ ๋ฌผ๋ฆฌ์ ยท๋…ผ๋ฆฌ์  ๋ณด์•ˆ, ์ž๋™ํ™”์™€ ์ธ๊ฐ„ ๊ฐ๋…์˜ ๊ท ํ˜•์„ ๊ณ ๋ คํ•œ ๋ช…ํ™•ํ•œ ์‚ฌ๊ณ  ๋Œ€์‘ ํ”„๋กœ์„ธ์Šค๊ฐ€ ํฌํ•จ๋œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ์—์ด์ „ํ‹ฑ AI๋Š” ๋ฌธ์ œ๋ฅผ ํƒ์ง€ํ•˜๊ณ  ์‹ ํ˜ธ๋ฅผ ์—ฐ๊ด€ ๋ถ„์„ํ•˜๋ฉฐ ๊ธฐ๊ณ„ ์†๋„๋กœ ๋Œ€์‘ํ•  ์ˆ˜ ์žˆ์ง€๋งŒ, ์ •์ฑ… ์„ค์ •๊ณผ ๊ฒฐ๊ณผ ๊ฒ€์ฆ, ์˜ˆ์ƒ ๋ฒ”์œ„๋ฅผ ๋ฒ—์–ด๋‚œ ์ƒํ™ฉ์—์„œ์˜ ํŒ๋‹จ์€ ์—ฌ์ „ํžˆ ์‚ฌ๋žŒ์ด ๋‹ด๋‹นํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋ฆฌ๋ฒจ๋กœ๋Š” MCSP๊ฐ€ ๊ด€๋ฆฌํ˜• ์„œ๋น„์Šค ๋ชจ๋ธ๊ณผ ๊ทธ ์ฃผ๋ณ€ ์ƒํƒœ๊ณ„์— ์–ผ๋งˆ๋‚˜ ์ž˜ ๋ถ€ํ•ฉํ•˜๋Š”์ง€๋„ ์ค‘์š”ํ•˜๋‹ค๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. ์ ํ•ฉํ•œ ์ œ๊ณต์—…์ฒด๋Š” ์ž๋™ํ™”์™€ AI๋ฅผ ํ™œ์šฉํ•ด ์šด์˜์„ ๋‹จ์ˆœํ™”ํ•ด์•ผ ํ•˜๋ฉฐ, ์ž๋™ํ™”๊ฐ€ ์ œ๋Œ€๋กœ ์ž‘๋™ํ•  ๊ฒฝ์šฐ ํ˜„์—… ์ธ๋ ฅ์„ ๋’ท๋ฐ›์นจํ•˜๊ณ  ์šด์˜์˜ ์ผ๊ด€์„ฑ์„ ๋†’์ด๋ฉฐ, ํŒ€์ด ๋˜ ๋‹ค๋ฅธ ๋„๊ตฌ๋ฅผ ๊ด€๋ฆฌํ•˜๋Š” ๋ฐ ์‹œ๊ฐ„์„ ์“ฐ๋Š” ๋Œ€์‹  ์‹ค์ œ๋กœ ์ค‘์š”ํ•œ ์—…๋ฌด์— ์ง‘์ค‘ํ•  ์ˆ˜ ์žˆ๋„๋ก ๋•๋Š”๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์†Œํ”„ํŠธ์›จ์–ด ๋ผ์ด์„ ์Šค ํ™œ์šฉ ๊ฐ€์น˜๋ฅผ ๋†’์ด๊ณ  MS, ์˜ค๋ผํด, ์‹œ์Šค์ฝ” ๋“ฑ ๋ฒค๋” ๊ฐ์‚ฌ ๋Œ€์‘์„ ์ง€์›ํ•˜๋Š” NPI์˜ ์ตœ๊ณ ๊ฒฝ์˜์ž ์กด ์œˆ์…‹์€ ๊ฐ€๊ฒฉ ๊ตฌ์กฐ์˜ ์œ ์—ฐ์„ฑ์ด MCSP ์„ ํƒ ๊ณผ์ •์—์„œ ์ข…์ข… ๊ฐ„๊ณผ๋œ๋‹ค๊ณ  ์ง€์ ํ–ˆ๋‹ค. ๊ทธ๋Š” MCSP์™€ ๊ด€๋ จ๋œ ์œ„ํ—˜์ด ์ดˆ๊ธฐ ๋น„์šฉ ์ฆ๊ฐ€๋ณด๋‹ค๋Š”, ์‹œ๊ฐ„์ด ์ง€๋‚˜๋ฉด์„œ ์ธ์ง€ํ•˜์ง€ ๋ชปํ•œ ์ฑ„ ํ˜‘์ƒ๋ ฅ์ด ์•ฝํ™”๋˜๋Š” ๋ฐ ์žˆ๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

์œˆ์…‹์€ ์†Œ๊ทœ๋ชจ ํŒ€์ด๋‚˜ ์•„์ง ํด๋ผ์šฐ๋“œ ๊ฒฝํ—˜์„ ๊ตฌ์ถ•ํ•˜๋Š” ๋‹จ๊ณ„์— ์žˆ๋Š” ์กฐ์ง์—๋Š” MCSP๊ฐ€ ํฐ ๋„์›€์ด ๋  ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ง๋ถ™์˜€๋‹ค. ํด๋ผ์šฐ๋“œ ์ง€์ถœ์„ ํ†ตํ•ฉํ•˜๊ณ  ๋งˆ์ด๊ทธ๋ ˆ์ด์…˜ ์ง€์›, ๋ฆฌ์†Œ์Šค ์ตœ์ ํ™”, ๋น„์šฉ ํ†ต์ œ์™€ ๊ฐ™์€ ์„œ๋น„์Šค๋ฅผ ํŒจํ‚ค์ง€๋กœ ์ œ๊ณตํ•จ์œผ๋กœ์จ ๋‚ญ๋น„๋ฅผ ์ค„์ด๊ณ  ํด๋ผ์šฐ๋“œ ์šด์˜์„ ๋ณด๋‹ค ์ˆ˜์›”ํ•˜๊ฒŒ ๋งŒ๋“ค ์ˆ˜ ์žˆ๋‹ค๋Š” ์„ค๋ช…์ด๋‹ค. ๋‚ด๋ถ€์— ํด๋ผ์šฐ๋“œ๋‚˜ ํ•€์˜ต์Šค ์—ญ๋Ÿ‰์ด ์ถฉ๋ถ„ํ•˜์ง€ ์•Š์€ ์กฐ์ง์ด๋ผ๋ฉด ์ด๋Ÿฌํ•œ ์ด์ ์ด ์ผ์ • ๋ถ€๋ถ„์˜ trade-off๋ฅผ ๊ฐ์ˆ˜ํ•  ๋งŒํ•œ ๊ฐ€์น˜๊ฐ€ ์žˆ๋‹ค๊ณ  ์ „ํ–ˆ๋‹ค.

๊ทธ๋Š” โ€œํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์ด ํ™•์žฅ๋ ์ˆ˜๋ก ๊ฐ€๊ฒฉ ๊ตฌ์กฐ๋Š” ์ ์  ๋ถˆํˆฌ๋ช…ํ•ด์ง„๋‹คโ€๋ผ๋ฉฐ โ€œMCSP๋Š” MS๋‚˜ ์•„๋งˆ์กด์›น์„œ๋น„์Šค(AWS) ์š”๊ธˆ ์œ„์— ์ž์ฒด ๋งˆ์ง„์„ ๋”ํ•˜๋Š”๋ฐ, ๊ธฐ๋ณธ ์‚ฌ์šฉ๋ฃŒ ๊ธฐ์ค€ ์ตœ๋Œ€ 8% ์ˆ˜์ค€์ด๋ฉฐ ์„œ๋น„์Šค๊ฐ€ ๋ฌถ์ผ ๊ฒฝ์šฐ ๊ทธ ์ด์ƒ์ด ๋  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด์–ด โ€œ์ด ๊ฐ™์€ ๊ด€๋ฆฌ ๊ณ„์ธต์„ ํ†ตํ•ด MCSP๋Š” ์•ฝ 30~40% ์ˆ˜์ค€์˜ ์ˆ˜์ต๋ฅ ์„ ํ™•๋ณดํ•œ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

MCSP์˜ ๋‹จ์ 

๊ธฐ์ˆ  ์ปจ์„คํŒ… ๊ธฐ์—… ํ•˜์ด๋ผ์ธ์˜ ๊ธฐ์ˆ  ๋ถ€์‚ฌ์žฅ ๋ผ์ด์–ธ ๋งฅ์—˜๋กœ์ด๋Š” MCSP๋ฅผ ํ™œ์šฉํ•  ๋•Œ ๊ฐ€์žฅ ํฐ ๋‹จ์ ์œผ๋กœ ํ†ต์ œ๋ ฅ ์ƒ์‹ค์„ ๊ผฝ์•˜๋‹ค.

๋งฅ์—˜๋กœ์ด๋Š” โ€œ๊ฐ์ข… ๋ผ์ด์„ ์Šค ํ• ์ธ ํ˜œํƒ์„ ๋ฐ›๋”๋ผ๋„ ๊ณ„์•ฝ์— ๋ฌถ์—ฌ ํ•„์š” ์ด์ƒ์œผ๋กœ ๊ตฌ๋งคํ•ด์•ผ ํ•˜๋Š” ๊ตฌ์กฐ๋ผ๋ฉด ์‹ค์ œ๋กœ๋Š” ๋น„์šฉ์„ ์ ˆ๊ฐํ•˜์ง€ ๋ชปํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œMCSP๋ฅผ ํ™œ์šฉํ•˜๋ฉด ์กฐ์ง์˜ ๊ณต๊ฒฉ ํ‘œ๋ฉด์ด ํ™•๋Œ€๋  ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ๋„ ๊ณ ๋ คํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” ๋˜ โ€œMS์™€ ๊ฐ™์€ ๋Œ€ํ˜• ํด๋ผ์šฐ๋“œ ๋ฒค๋”๊ฐ€ MCSP๋ฅผ ๊ต์œกํ•˜๊ณ  ๊ฐ€์ด๋“œ๋ฅผ ์ œ๊ณตํ•˜๋”๋ผ๋„, ๋Œ€๊ทœ๋ชจ ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ ์‚ฌ๊ณ  ์ดํ›„ ์ž‘์„ฑ๋˜๋Š” ๊ทผ๋ณธ ์›์ธ ๋ถ„์„ ๋ณด๊ณ ์„œ๋ฅผ ์‚ดํŽด๋ณด๋ฉด MCSP๊ฐ€ ๊ณต๊ฒฉ ๊ฒฝ๋กœ๋กœ ์ž‘์šฉํ•œ ์‚ฌ๋ก€๊ฐ€ ์šฐ๋ ค์Šค๋Ÿฌ์šธ ์ •๋„๋กœ ์ž์ฃผ ๋“ฑ์žฅํ•œ๋‹คโ€๋ผ๊ณ  ์ „ํ–ˆ๋‹ค.

๋ฆฌ์„œ์น˜ ๊ธฐ์—… ISG์˜ ๋””๋ ‰ํ„ฐ ์•„๋„ค์ด ๋‚˜์™€ํ…Œ๋Š” MCSP ํ˜‘์—…์ด ๋งŽ์€ ์ด์ ์„ ์ œ๊ณตํ•˜๋Š” ๋™์‹œ์— ๋ถ„๋ช…ํ•œ ์œ„ํ—˜๋„ ๋™๋ฐ˜ํ•œ๋‹ค๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

๋‚˜์™€ํ…Œ๋Š” โ€œMCSP๊ฐ€ ์กฐ์ง ๋‚ด ์•„ํ‚คํ…์ฒ˜ ๋…ผ์˜์˜ ์ค‘์‹ฌ์ ์ธ ๋ชฉ์†Œ๋ฆฌ๊ฐ€ ๋˜์–ด์„œ๋Š” ์•ˆ ๋œ๋‹คโ€๋ผ๋ฉฐ โ€œํ•ต์‹ฌ ์‹œ์Šคํ…œ์— ๋Œ€ํ•œ ์ง€์‹์„ ๋‚ด๋ถ€์— ์œ ์ง€ํ•˜๊ณ , ๋ฒค๋” ์ข…์†์„ ์ค„์ด๋ฉฐ, ์‹œ์žฅ ๋ชจ๋ฒ” ์‚ฌ๋ก€์™€ ๋น„๊ตํ–ˆ์„ ๋•Œ ์ œ๊ณต์—…์ฒด๋กœ ์ธํ•œ ์•„ํ‚คํ…์ฒ˜ ํŽธํ–ฅ์„ ์™„ํ™”ํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์•„ํ‚คํ…์ฒ˜ ์˜์‚ฌ๊ฒฐ์ •์„ ๋‚ด๋ถ€์—์„œ ์†Œ์œ ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ๋Š” ๋˜ MCSP๊ฐ€ ํด๋ผ์šฐ๋“œ ๋น„์šฉ ๊ด€๋ฆฌ์— ๋Œ€ํ•ด ์‹ค์ œ๋กœ ํด๋ผ์šฐ๋“œ๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๊ธฐ์—…๋งŒํผ์˜ ์••๋ฐ•์„ ๋А๋ผ์ง€ ์•Š๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค๊ณ  ๋ง๋ถ™์˜€๋‹ค. ๊ฒฐ๊ตญ ๊ณผ๋„ํ•œ ์ง€์ถœ์˜ ์˜ํ–ฅ์€ ๊ธฐ์—…์ด ์ง์ ‘ ๊ฐ๋‚ดํ•˜๊ฒŒ ๋˜๋ฉฐ, ์ด๋Ÿฌํ•œ ์ด์œ ๋กœ ๋งŽ์€ ๊ธฐ์—…์ด ํด๋ผ์šฐ๋“œ ๋น„์šฉ์— ๋Œ€ํ•œ ํ†ต์ œ๋ ฅ์„ ํ™•๋ณดํ•˜๊ธฐ ์œ„ํ•ด ํ•€์˜ต์Šค ์—ญํ• ์„ ๋‹ค์‹œ ๋‚ด๋ถ€๋กœ ๊ฐ€์ ธ์˜จ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ธ€๋กœ๋ฒŒ ์‹œ์žฅ์—์„œ ์ฃผ๋ชฉ๋ฐ›๋Š” MCSP 6๊ณณ

๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค ์ œ๊ณต์—…์ฒด๋Š” ์ˆ˜์‹ญ ๊ณณ์— ์ด๋ฅธ๋‹ค. ์กฐ์‚ฌ ๋ถ€๋‹ด์„ ์ค„์ด๊ธฐ ์œ„ํ•ด ๋…๋ฆฝ์ ์ธ ๋ฆฌ์„œ์น˜์™€ ์• ๋„๋ฆฌ์ŠคํŠธ์™€์˜ ๋…ผ์˜๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ, ์•ŒํŒŒ๋ฒณ์ˆœ์œผ๋กœ ์ฃผ์š” MCSP 6๊ณณ์„ ์ •๋ฆฌํ–ˆ๋‹ค. ๊ฐ€๊ฒฉ ์ •๋ณด๋Š” ๊ฐ ์ œ๊ณต์—…์ฒด์— ์ง์ ‘ ๋ฌธ์˜ํ•ด์•ผ ํ•œ๋‹ค.

์•ก์„ผ์ถ”์–ด

์•ก์„ผ์ถ”์–ด(Accenture)๋Š” ์ „ ์„ธ๊ณ„ ์ฃผ์š” ์ง€์—ญ๊ณผ ์‹œ์žฅ์— ๋ถ„ํฌํ•œ ํŒ€๊ณผ ์„ผํ„ฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•œ๋‹ค. ๊ธฐ์—…์˜ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ ์„ค๊ณ„, ์šด์˜, ์œ ์ง€ ๊ด€๋ฆฌ๋ฅผ ์ง€์›ํ•˜๋ฉฐ, ์ดˆ๊ธฐ ํด๋ผ์šฐ๋“œ ๊ตฌ์ถ•๋ถ€ํ„ฐ ๋ชจ๋‹ˆํ„ฐ๋ง, ์œ ์ง€ ๋ณด์ˆ˜, ๋ณด์•ˆ์„ ํฌํ•จํ•œ ์ง€์†์ ์ธ ์šด์˜๊นŒ์ง€ ํญ๋„“๊ฒŒ ๋‹ค๋ฃฌ๋‹ค. MS ์• ์ €, ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ, AWS ๋“ฑ ์ฃผ์š” ํด๋ผ์šฐ๋“œ ํ”Œ๋žซํผ ์ „๋ฐ˜์—์„œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๋Š” ๊ฒƒ๋„ ํŠน์ง•์ด๋‹ค. ๊ธฐ์—…์€ ๋ณต์žกํ•œ ํด๋ผ์šฐ๋“œ ์‹œ์Šคํ…œ์„ ์ „๋ถ€ ๋‚ด๋ถ€์—์„œ ๊ด€๋ฆฌํ•˜๋Š” ๋Œ€์‹ , ์•ก์„ผ์ถ”์–ด๋ฅผ ํ†ตํ•ด ์ผ์ƒ์ ์ธ ์šด์˜๊ณผ ๊ธฐ์ˆ ์  ๊ด€๋ฆฌ ์—…๋ฌด๋ฅผ ๋งก๊ธธ ์ˆ˜ ์žˆ๋‹ค. ์‹œ์Šคํ…œ ๋ชจ๋‹ˆํ„ฐ๋ง๊ณผ ์ด์Šˆ ๋Œ€์‘, ํ™˜๊ฒฝ ์—…๋ฐ์ดํŠธ ๋“ฑ ์ผ์ƒ์ ์ธ ์ธํ”„๋ผ ์šด์˜์„ ์•ก์„ผ์ถ”์–ด๊ฐ€ ๋‹ด๋‹นํ•จ์œผ๋กœ์จ, ๋‚ด๋ถ€ ์ธ๋ ฅ์€ ํ•ต์‹ฌ ๋น„์ฆˆ๋‹ˆ์Šค ๊ณผ์ œ์— ์ง‘์ค‘ํ•  ์ˆ˜ ์žˆ๋‹ค.

์บก์ œ๋ฏธ๋‹ˆ

์บก์ œ๋ฏธ๋‹ˆ(Capgemini)๋Š” ์ „ ์„ธ๊ณ„๋ฅผ ๋Œ€์ƒ์œผ๋กœ ๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๋ฉฐ, ํŠนํžˆ ์œ ๋Ÿฝ๊ณผ ๋ถ๋ฏธ๋ฅผ ์ค‘์‹ฌ์œผ๋กœ ๋ฉ€ํ‹ฐํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์„ ์ง€์›ํ•œ๋‹ค. ์ œ์กฐ, ๋ฆฌํ…Œ์ผ, ๊ธˆ์œต ์„œ๋น„์Šค, ๋ณดํ—˜ ์‚ฐ์—…๊ณผ์˜ ํ˜‘์—… ๊ฒฝํ—˜์ด ํ’๋ถ€ํ•˜๋‹ค. AWS, MS ์• ์ €, ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ ๋“ฑ ์ฃผ์š” ํด๋ผ์šฐ๋“œ ํ”Œ๋žซํผ๊ณผ ์ผ๋ถ€ ํŠนํ™”๋œ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ํด๋ผ์šฐ๋“œ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜๊ณผ ์ธํ”„๋ผ ์šด์˜์„ ์ง€์›ํ•œ๋‹ค. ๋ชจ๋‹ˆํ„ฐ๋ง, ๋ฐฑ์—…, ๊ธฐ์ˆ  ์ง€์›์„ ํฌํ•จํ•œ ๊ด€๋ฆฌํ˜• ์„œ๋น„์Šค์™€ ํ•จ๊ป˜, ํด๋ผ์šฐ๋“œ ์ด์ „์ด ์ ํ•ฉํ•œ ์›Œํฌ๋กœ๋“œ๋ฅผ ์‹๋ณ„ํ•˜๊ณ  ํ•ด๋‹น ์‹œ์Šคํ…œ์„ ์ด์ „ยท์šด์˜ํ•˜๋Š” ๊ณผ์ •๊นŒ์ง€ ํฌ๊ด„์ ์œผ๋กœ ์ง€์›ํ•œ๋‹ค. ์ค‘๊ฒฌ๊ธฐ์—…๋ณด๋‹ค๋Š” ๋Œ€๊ทœ๋ชจ์ด๋ฉด์„œ ๋ณต์žกํ•œ ํ™˜๊ฒฝ์„ ๊ฐ€์ง„ ๋Œ€๊ธฐ์—…์— ์ ํ•ฉํ•œ ์„œ๋น„์Šค ์„ฑ๊ฒฉ์„ ๊ฐ–๊ณ  ์žˆ๋‹ค.

๋”œ๋กœ์ดํŠธ

๋”œ๋กœ์ดํŠธ(Deloitte)๋Š” ์ „ ์„ธ๊ณ„ ๊ณ ๊ฐ์„ ๋Œ€์ƒ์œผ๋กœ ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๋ฉฐ, ๋ถ๋ฏธ์™€ ์œ ๋Ÿฝ ์ง€์—ญ์˜ ๋น„์ค‘์ด ํฌ๋‹ค. ๊ธˆ์œตยท๋ณดํ—˜, ๊ณต๊ณต, ํ—ฌ์Šค์ผ€์–ด ์‚ฐ์—…์—์„œ ํŠนํžˆ ๊ฐ•์ ์„ ๋ณด์ธ๋‹ค. AWS, MS ์• ์ €, ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ, VM์›จ์–ด ํด๋ผ์šฐ๋“œ, ์˜ค๋ผํด ํด๋ผ์šฐ๋“œ๋ฅผ ํฌํ•จํ•œ ๋ฉ€ํ‹ฐํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์„ ์ง€์›ํ•œ๋‹ค. ๊ธฐ์—…์˜ ๋น„์ฆˆ๋‹ˆ์Šค ๋ชฉํ‘œ์— ๋งž์ถฐ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์„ ๊ธฐํšยท๊ตฌ์ถ•ยท์šด์˜ํ•˜๋Š” ๋ฐ ์ดˆ์ ์„ ๋งž์ถ”๊ณ  ์žˆ์œผ๋ฉฐ, ํ”„๋กœ์„ธ์Šค์™€ ์šด์˜ ๊ฐœ์„ ์„ ํฌํ•จํ•œ ํด๋ผ์šฐ๋“œ ์ „ํ™˜์ด ํ•ต์‹ฌ ์˜์—ญ์ด๋‹ค. ์ปจ์„คํŒ…์ด ์ฃผ๋ ฅ ์‚ฌ์—…์ด์ง€๋งŒ, ๋””์ง€ํ„ธ ์ „ํ™˜์„ ์ถ”์ง„ํ•˜๋Š” ๋Œ€๊ธฐ์—…์„ ์ค‘์‹ฌ์œผ๋กœ ๊ด€๋ฆฌํ˜• ์„œ๋น„์Šค ์˜์—ญ๋„ ์ง€์†์ ์œผ๋กœ ํ™•๋Œ€ํ•˜๊ณ  ์žˆ๋‹ค.

HCLํ…Œํฌ๋†€๋กœ์ง€์Šค

HCLํ…Œํฌ๋†€๋กœ์ง€์Šค(HCL Technologies)๋Š” ์ „ ์„ธ๊ณ„์— ๋ถ„ํฌํ•œ ํŒ€๊ณผ ์„ผํ„ฐ๋ฅผ ํ†ตํ•ด ๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•œ๋‹ค. AWS, MS ์• ์ €, ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ ๋“ฑ ์ฃผ์š” ํด๋ผ์šฐ๋“œ ์ œ๊ณต์—…์ฒด์™€ ํ˜‘๋ ฅํ•ด ๊ฐ ๊ธฐ์—…์˜ ์š”๊ตฌ์— ๋งž๋Š” ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์„ ์„ค๊ณ„ยท๊ตฌ์ถ•ํ•˜๊ณ , ์ดํ›„ ์•ˆ์ •์ ์ธ ์šด์˜์„ ์ง€์›ํ•œ๋‹ค. ๊ตฌ์ถ• ์ดํ›„์—๋Š” 24์‹œ๊ฐ„ ๋ชจ๋‹ˆํ„ฐ๋ง, ์„ฑ๋Šฅ ๊ด€๋ฆฌ, ์žฅ์•  ๋Œ€์‘ ๋“ฑ ์ผ์ƒ์ ์ธ ์šด์˜์„ ๋‹ด๋‹นํ•˜๋ฉฐ, ๋ฐ˜๋ณต์ ์ธ IT ์ž‘์—…์—๋Š” ์ž๋™ํ™”์™€ AI ๋„๊ตฌ๋ฅผ ํ™œ์šฉํ•œ๋‹ค. ๊ธˆ์œต, ์ œ์กฐ, ํ—ฌ์Šค์ผ€์–ด ๋“ฑ ๋‹ค์–‘ํ•œ ์‚ฐ์—…์—์„œ ์•ˆ์ •์ ์ธ ํด๋ผ์šฐ๋“œ ์‹œ์Šคํ…œ ์šด์˜์„ ์ง€์›ํ•˜๋Š” ๊ฒƒ์ด ํŠน์ง•์ด๋‹ค.

NTT๋ฐ์ดํ„ฐ

NTT๋ฐ์ดํ„ฐ(NTT Data)๋Š” ์ „ ์„ธ๊ณ„ ๊ณ ๊ฐ์„ ๋Œ€์ƒ์œผ๋กœ ๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๋ฉฐ, ์ œ์กฐ, ํ—ฌ์Šค์ผ€์–ด, ๊ธˆ์œต ์„œ๋น„์Šค, ๋ณดํ—˜ ๋“ฑ ํญ๋„“์€ ์‚ฐ์—…์„ ์ง€์›ํ•œ๋‹ค. MS ์• ์ €, ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ, IBM ํด๋ผ์šฐ๋“œ, AWS๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ ๋ฉ€ํ‹ฐํด๋ผ์šฐ๋“œ ์ „๋žต์„ ์ฑ„ํƒํ•˜๊ณ  ์žˆ๋‹ค. ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์˜ ํด๋ผ์šฐ๋“œ ์ด์ „, ๋…ธํ›„ ์‹œ์Šคํ…œ ํ˜„๋Œ€ํ™”, ๋ ˆ๊ฑฐ์‹œ ๊ธฐ์ˆ  ์ „ํ™˜์„ ์ง€์›ํ•˜๋Š” ํ•œํŽธ, NTT ๊ทธ๋ฃน ์ „๋ฐ˜์˜ ์—ญ๋Ÿ‰์„ ํ™œ์šฉํ•ด ์•„์ด๋ดํ‹ฐํ‹ฐ ๋ฐ ์ ‘๊ทผ ๊ด€๋ฆฌ, ๋„คํŠธ์›Œํ‚น, ๊ด€๋ฆฌํ˜• ๋ณด์•ˆ ์„œ๋น„์Šค๋„ ํ•จ๊ป˜ ์ œ๊ณตํ•œ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ๊ณ ๊ฐ์ด ๋น„์ฆˆ๋‹ˆ์Šค๋ฅผ ๋ณด๋‹ค ํšจ๊ณผ์ ์œผ๋กœ ์ง€์›ํ•˜๋Š” ํด๋ผ์šฐ๋“œ ๊ธฐ๋ฐ˜ ์‹œ์Šคํ…œ์„ ๊ตฌ์ถ•ํ•˜๋„๋ก ๋•๋Š”๋‹ค.

ํƒ€ํƒ€์ปจ์„คํ„ด์‹œ์„œ๋น„์Šค

ํƒ€ํƒ€์ปจ์„คํ„ด์‹œ์„œ๋น„์Šค(Tata Consultancy Services, TCS)๋Š” ์ „ ์„ธ๊ณ„ ๊ธฐ์—…๊ณผ ํ˜‘๋ ฅํ•˜๊ณ  ์žˆ์ง€๋งŒ, ๊ด€๋ฆฌํ˜• ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค ๊ณ ๊ฐ์€ ์ฃผ๋กœ ๋ถ๋ฏธ์™€ ์œ ๋Ÿฝ์— ์ง‘์ค‘๋ผ ์žˆ๋‹ค. ๊ธˆ์œต ์„œ๋น„์Šค, ์ƒ๋ช…๊ณผํ•™ยท์ œ์•ฝ, ๋ฆฌํ…Œ์ผ ์‚ฐ์—…์—์„œ ๊ฐ•ํ•œ ๊ฒฝํ—˜์„ ๋ณด์œ ํ•˜๊ณ  ์žˆ๋‹ค. MS ์• ์ €, ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ, ์˜ค๋ผํด ํด๋ผ์šฐ๋“œ, AWS๋ฅผ ์ค‘์‹ฌ์œผ๋กœ ๋ฉ€ํ‹ฐํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์„ ์ง€์›ํ•˜๋ฉฐ, ์ผ๋ถ€ IBM ํด๋ผ์šฐ๋“œ๋„ ์ œ๊ณตํ•œ๋‹ค. ์ฃผ์š” ํด๋ผ์šฐ๋“œ ํŒŒํŠธ๋„ˆ๋ณ„ ์ „๋‹ด ํŒ€์„ ์šด์˜ํ•˜๋ฉฐ, ๋Œ€๊ธฐ์—…์„ ๋Œ€์ƒ์œผ๋กœ ํด๋ผ์šฐ๋“œ ์ด์ „ ์ „๋žต ์ˆ˜๋ฆฝ, ๊ธฐ์กด ์‹œ์Šคํ…œ ์ด์ „, ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ํ˜„๋Œ€ํ™”๋ฅผ ์ง€์›ํ•œ๋‹ค. ์„œ๋น„์Šค์˜ ์ค‘์‹ฌ์€ ๋Œ€๊ธฐ์—…์— ๋งž์ถฐ์ ธ ์žˆ์œผ๋ฉฐ, ์ค‘๊ฒฌ๊ธฐ์—… ๋Œ€์ƒ ๋น„์ค‘์€ ์ƒ๋Œ€์ ์œผ๋กœ ์ œํ•œ์ ์ด๋‹ค.
dl-ciokorea@foundryco.com

5 essential skills every project manager needs during a data center transformation to the cloud

As organizations accelerate their shift from traditional data center environments to hybrid and multi-cloud architectures, the scale and complexity of these initiatives demand a new caliber of project leadership. Having recently led a multi-year enterprise-wide data center transformation with global stakeholders, Iโ€™ve seen firsthand that technology alone is not what ensures success. Leadership is the key.

Even the most advanced platforms and tools can fall short without a project manager who brings the right mindset, adaptability and technical fluency. These programs are simultaneously technical undertakings and organizational-change journeys.

Based on lessons learned from managing one of the most ambitious transformations in my organization, here are the five skills essential for any project manager responsible for navigating cloud and data center modernization.

1. Systems thinking & architectural awareness

Data center transformations operate at an enterprise scale, where no system exists in isolation. Every application, integration point and data flow is part of a wider ecosystem and understanding that ecosystem is critical from day one. Systems thinking means looking beyond servers and environments to examine business processes, downstream dependencies, data protection needs and operational realities.

This requires asking targeted questions such as:

  • What is the business impact if this application is down for four hours or more?
  • How many teams, processes or users rely on it?
  • What are its recovery objectives and how does it interact with upstream and downstream systems?

With these insights, project managers can make informed decisions about cutover sequencing and avoid grouping applications solely by physical infrastructure โ€” an approach that often leads to outages or misplaced dependencies. Indeed, a recent empirical study of migrating legacy systems to cloud platforms identified a lack of architectural mapping and understanding of interdependencies as a key risk factor in migration failures.

Takeaway

Architectural awareness isnโ€™t memorizing components; itโ€™s understanding how a single change reverberates across the entire enterprise system.

2. Elastic governance & proactive risk anticipation

Large-scale migrations rarely follow a predictable or linear path. They unfold in iterative phases, each introducing new variables, technical constraints and lessons learned. Because of this, a traditional waterfall approach quickly becomes a liability. What teams need instead is an elastic governance framework that provides structure while adapting to shifting realities.

Elastic governance means adjusting processes, decision models and approval flows as new insights surface. Each application and business unit often carries its own architecture, dependencies and constraints, so a one-size-fits-all model simply doesnโ€™t work. During our migration, daily interactions with implementation teams, developers and product owners gave me real-time visibility into emerging issues and allowed us to refine our approach continuously.

This approach mirrors trends highlighted in the ISACA Journalโ€™s 2023 article, โ€œRedefining Enterprise Cloud Technology Governance.โ€ ISACA argues that traditional governance frameworks are far too rigid for modern cloud environments. Instead, they advocate for adaptive, decentralized models that empower teams to respond quickly as new constraints and dependencies emerge.

Vendor-related challenges were especially common with aging legacy systems. Proactive engagement โ€” rather than reactive firefighting โ€” helped us avoid failures and maintain momentum.

Takeaway

Governance should guide, not grind. Flexibility is essential for managing uncertainty and sustaining progress in complex transformations.

3. Stakeholder coordination and strategic communication

In enterprise-wide transformation programs, stakeholder alignment is often the difference between controlled progress and project derailment. Every migration window, firewall rule adjustment, environment change or sequence shift requires close coordination across security, networking, infrastructure, operations, product teams and business leadership โ€” all operating with their own priorities and pressures.

Research shows that stakeholders often have different โ€œframesโ€ of a digital transformation and successful programs actively manage these perspectives to create shared understanding and alignment. Similarly, a 2023 KPMG report highlights that building trust among stakeholders โ€” particularly around risk, security and compliance โ€” is essential for successful cloud adoption.

A critical part of this role is translation. The project manager must convert technical constraints into clear, business-friendly updates while also translating business expectations into actionable direction for engineering teams. This dual fluency reduces misunderstanding and accelerates decision-making.

To maintain alignment, structured communication becomes essential. I established predictable rhythms โ€” daily standups, weekly product syncs, monthly executive briefings and shared dashboards โ€” to ensure transparency, quick escalation and consistent visibility into progress and risks.

Takeaway

The stronger and more structured the communication, the smoother and more predictable the migration.

4. Technical fluency & decision facilitation

Modernization initiatives involve ongoing decisions about whether to re-host, re-platform or re-architect applications. While a project manager doesnโ€™t need to be the most technical person in the room, they must understand the implications of each option well enough to facilitate informed decision-making.

Technical fluency builds credibility with developers, architects, vendors and deployment teams. It also enables the project manager to ask the right questions, challenge assumptions and guide discussions toward solutions. This is especially important given the โ€œ6 Rsโ€ of cloud migration โ€” re-host, re-platform, refactor (re-architect) and others โ€” which are commonly used to rationalize workloads based on business goals and technical fit.

Takeaway

Technical fluency enables clarity, connection and better decisions.

5. Resilience & change leadership

Data center transformations are long, complex and filled with uncertainties. Unexpected technical issues, compliance demands and shifting business priorities can slow down momentum and strain teams. According to the KPMG report mentioned earlier, many organizations struggle with operational resilience โ€” more than half experienced outages or compliance issues in their cloud operations over the past year. This reinforces the importance of proactive governance and risk management. In such environments, a resilient project manager provides clarity, maintains stability and ensures the team keeps moving forward.

During our project, an unexpected compliance mandate required rapid reprioritization and additional resources. With leadership support, we realigned the plan and still met the migration deadline. Maintaining team morale during such periods is just as important as technical delivery.

Takeaway

Resilient teams donโ€™t resist change; they stay confident through it.

Integrating the 5 skills: The project manager as transformation leader

A data center transformation is more than a technical project โ€” it reshapes processes, roles and behaviors across the organization. When these five skills come together, the project manager transitions from a delivery role into a true transformation leader.

  • Systems thinking eliminates hidden dependencies.
  • Elastic governance adapts to evolving needs.
  • Stakeholder coordination maintains across-the-board alignment.
  • Technical fluency builds trust and accelerates decision-making.
  • Resilience keeps teams focused during disruption.

The most effective transformation leaders balance discipline with flexibility.

Measuring success beyond migration

Traditional success metrics such as reduced downtime, regulatory compliance and cost optimization are important. But true success becomes clear only when the organization demonstrates improved adaptability and stronger collaboration between IT and the business.

When a project manager embeds adaptability deep into the organization, the transformation continues long after the final cutover.

The future-ready project manager

Looking ahead, managing a data center transformation a decade from now will be fundamentally different. The next generation of migrations will involve greater complexity, including advanced automation, AI-driven orchestration, multi-cloud environments and more sophisticated compliance and security requirements. Without continuous upskilling, project managers will struggle to lead confidently in this evolving landscape.

Future-ready leaders must be both technologically fluent and human-centered. They need to leverage data effectively, make decisions at the pace of AI and automation and understand emerging tools and methodologies. At the same time, they must maintain essential human leadership qualities โ€” trust, accountability, resilience and the ability to inspire teams under pressure.

By balancing these technical and human skills, project managers remain indispensable. They not only ensure that migrations succeed technically but also guide teams and organizations with purpose, clarity and adaptability, enabling sustainable transformation that goes beyond the immediate project and strengthens the organizationโ€™s long-term capabilities.

Closing thoughts

Data center transformation was not an easy migration, as it was a complicated and most ambitious undertaking by the organization. Orchestrating more than a hundred stakeholders was not an easy feat and we accomplished it with meticulous planning and risk management. Hence, a project manager with those five skills doesnโ€™t just lead, they become the transformation agents for the organization. As the saying goes: Real transformation happens when leadership turns complexity into clarity and uncertainty into forward motion.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

MCSP buyerโ€™s guide: 6 top managed cloud services providers โ€” and how to choose

A managed cloud services provider (MCSP) helps organizations run some or all of their cloud environments. This can include moving systems to the cloud, monitoring and maintaining them, improving performance, managing security tools, and helping control costs. MCSPs typically work across public, private, and hybrid cloud environments.

Organizations decide which parts of their cloud environments they want the provider to handle and which parts they want to keep in-house. In most cases, the company and the MCSP share responsibility. The provider manages day-to-day operations and tooling, while the organization stays accountable for business decisions, data, and governance.

Choosing an MCSP is always an unnerving experience, says Brent Riley, MCSP VP of digital forensics and incident response for North America at cybersecurity consultancy CyXcel.

โ€œSo much trust is placed in their ability to perform to the level promised in their SLA, but it can be tough to validate whether theyโ€™re being met until thereโ€™s an outage or cybersecurity incident that reveals issues,โ€ he says. โ€œAt that point, the damage is done. MCSPs are even more challenging to evaluate and select as thereโ€™s no physical infrastructure to inspect, and no visible work being done within an on-premise infrastructure.โ€

Benefits using an MCSP

Reduced operational burden: MCSPs can take on day-to-day cloud management tasks, reducing the need for large internal cloud and infrastructure teams. This is especially helpful for organizations that donโ€™t have deep cloud or FinOps expertise in-house.

Faster problem response: Most MCSPs provide 24/7 monitoring and support. When issues arise, their teams can respond quickly, often before problems significantly impact users or applications.

Support for disaster recovery and resilience: MCSPs help design, manage, and test backup and disaster recovery setups. While customers still define recovery goals, providers help ensure systems can be restored quickly if something goes wrong.

Ongoing platform management: Cloud platforms change frequently. MCSPs help keep infrastructure components current and compatible, reducing the risk of outdated configurations while allowing customers to control when major changes are introduced.

Security expertise and tooling: Cloud security requires specialized skills in high demand. MCSPs bring experience with identity management, monitoring, compliance tools, and security best practices. Security remains a shared responsibility, but providers help strengthen day-to-day protection.

Improved reliability and performance: With experience running large and complex environments, MCSPs can help design and operate cloud infrastructure thatโ€™s more stable, scalable, and resilient.

Integration with existing systems: MCSPs help connect cloud resources with on-prem systems, applications, and identity platforms. This makes it easier for users and applications to access cloud services without disruption.

More predictable operations, not always lower costs: While MCSPs can reduce internal staffing and tooling costs, they donโ€™t always lower overall cloud spend. Their value today is more about operational efficiency, expertise, and speed than cheaper cloud pricing.

Key considerations when choosing an MCSP

As organizations move toward more autonomous, AI-driven services, MCSPs play an important role turning automation into something that actually works every day, says Manny Rivelo, CEO at ConnectWise, a provider of IT management software.

Rivelo says one thing matters more than many teams realize: operational transparency. Organizations need a clear view into how their cloud environments are designed, secured, and managed, as well as how agentic AI monitors systems, makes decisions, and takes action so nothing important happens behind the scenes without their knowledge.

โ€œOperational maturity matters more as autonomy increases,โ€ Rivelo says. โ€œThis includes disciplined data governance, strong physical and logical security, and well-defined incident response processes that balance automation with human oversight. While agentic AI can detect issues, correlate signals, and respond at machine speed, humans remain essential to set policy, validate outcomes, and make judgment calls when conditions fall outside expected patterns.โ€

Itโ€™s also important that the MCSP fits well with the managed services model and the broader ecosystem around it, according to Rivelo. The right provider should use automation and AI to make things simpler. After all, when automation is done right, it backs up the people doing the work, brings more consistency to operations, and gives teams more time to focus on what actually matters, not manage another set of tools.

One factor that often gets missed when choosing an MCSP is how flexible pricing really is, says Jon Winsett, CEO at NPI, which helps enterprises get more value from their software licenses and navigate audits from vendors such as Microsoft, Oracle, and Cisco. The risk with an MCSP is usually not paying more at the start but losing negotiating power over time without noticing it.

MCSPs can be a big help for smaller teams or organizations still building cloud experiences, he adds. By combining cloud spend and packaging services, such as migration support, rightsizing, and cost controls, they can cut down on waste and make the cloud easier to run. For organizations without strong cloud or FinOps skills in-house, those benefits can be worth the tradeoffs.

โ€œAs cloud environments grow, pricing often becomes less clear,โ€ says Winsett. โ€œMCSPs add their own markup on top of Microsoft or AWS pricing, up to 8% for basic spend and more when services are bundled. That managed layer is how MCSPs reach profit margins of roughly 30 to 40%.โ€

Disadvantages to working with an MCSP

The biggest disadvantage using an MCSP is loss of control, according to Ryan McElroy, VP of technology at tech consulting firm Hylaine.ย 

โ€œIf you get discounts for various licenses, but youโ€™re locked into contracts and have to overbuy, then you may not be saving money,โ€ he says. โ€œAnd an MCSP adds to your organizationโ€™s attack surface area. While Microsoft and other large cloud vendors train their MCSPs and provide guidance, if you read the root cause analysis reports produced after major cybersecurity incidents, youโ€™ll find itโ€™s a worryingly common vector.โ€

Anay Nawathe, director at research and advisory firm ISG, says that while working with MCSPs has many benefits, there are also risks.

โ€œYour MCSP shouldnโ€™t be the main voice of architecture in your organization,โ€ he says. โ€œArchitectural decisions should be owned internally to maintain key systems knowledge in-house, reduce vendor lock-in, and mitigate architectural bias from a provider compared to market best practices.โ€

Additionally, he adds that MCSPs donโ€™t always feel the same pressure to manage costs as the companies using the cloud. In the end, enterprises are the ones who feel the impact of overspending, which is why many bring FinOps roles back in-house to take direct control of cloud costs, he says.

6 top MCSPs

There are dozens, so to help streamline the research, we highlight the following products, arranged alphabetically, based on independent research and discussions with analysts. Organizations should contact providers directly for pricing information.

Accenture

Accenture offers its managed cloud services to customers worldwide, backed by teams and centers in most major regions and markets. It helps organizations design, run, and maintain their cloud environments, and supports everything from initial cloud setup to ongoing operations, including monitoring, maintenance, and security. Accenture also works across major cloud platforms, such as Microsoft Azure, Google Cloud, and AWS. Instead of managing complex cloud systems entirely in-house, companies can use Accentureโ€™s services to handle routine operations and technical oversight. This includes monitoring systems, addressing issues as they come up, and keeping cloud environments updated. Overall, Accenture manages the day-to-day cloud infrastructure so organizational in-house staff can focus on key business priorities.

Capgemini

Capgemini provides managed cloud services worldwide and supports multicloud environments across all major regions, with much of its work centered in Europe and North America. The company works closely with industries such as manufacturing, retail, financial services, and insurance. Capgemini helps organizations run and manage applications on major cloud platforms, including AWS, Microsoft Azure, and Google Cloud, as well as specialized enterprise clouds. Its managed services cover both infrastructure and applications, including monitoring, backups, and technical support. Capgemini also helps companies decide which workloads make sense to move to the cloud, migrate those systems, and manage them over time. The firm is best suited for large enterprises and complex environments rather than midsize organizations.

Deloitte

Deloitte provides cloud services to customers around the world, with much of its work focused on organizations in North America and Europe. It works heavily with industries in financial services and insurance, government, and healthcare. Deloitte supports multicloud environments and works with platforms including AWS, Microsoft Azure, Google Cloud, VMware Cloud, and Oracle Cloud. The firm helps companies plan, build, and operate cloud environments tailored to business goals. A key focus is cloud transformation, including identifying where cloud tech can improve processes and operations. Deloitte is best suited for large enterprises pursuing digital transformation, and while consulting remains its core business, the firm continues to expand its managed services offerings.

HCL Technologies

Managed cloud services from HCL Technologies are offered globally, and supported by teams and centers around the world. HCL helps organizations move their systems to the cloud and keep them running smoothly over time. It works with major cloud providers, such as AWS, Microsoft Azure, and Google Cloud to design and set up cloud environments that match each businessโ€™s needs. Once everythingโ€™s in place, HCL handles the daily operations, including around-the-clock monitoring, performance management, and fixing issues as they arise, and also uses automation and AI tools for routine IT tasks. Overall, HCL helps organizations maintain reliable cloud systems across industries like banking, manufacturing, and healthcare.

NTT Data

NTT Data delivers managed cloud services to customers globally. It supports a wide range of industries, including manufacturing, healthcare, financial services, and insurance. NTT Data takes a multicloud approach, with managed services customers running on Microsoft Azure, Google Cloud, IBM Cloud, and AWS. NTT Data also helps companies move applications to the cloud, modernize aging systems, and move away from legacy tech, as well as draws on expertise from across the NTT Group to offer services like identity and access management, networking, and managed security, helping customers build cloud-based systems that better support their businesses.

Tata Consultancy Services

TCS works with organizations worldwide, but most of its cloud and managed services customers are in North America and Europe. The company has strong experience in industries such as financial services, life sciences and pharmaceuticals, and retail. TCS supports multicloud environments and works with leading cloud platforms like Microsoft Azure, Google Cloud, Oracle Cloud, and AWS, with some support for IBM Cloud. TCS has dedicated teams for its largest cloud partners and helps large enterprises plan cloud migrations, move existing systems, and modernize applications for the cloud. The majority of this work is focused on large enterprises, with limited emphasis on midsize organizations.


Epicor sets timeline to sunset on-prem ERP as cloud becomes the only path forward

Reflecting the continued push by ERP vendors to the cloud, Epicor has announced its schedule to sunset several of its legacy on-premises tools.

The company will roll out final releases for Epicor Kinetic, Epicor Prophet 21, and Epicor BisTrack, and will offer tiered support levels in a phased schedule beginning later this year.

Epicor says this will allow enterprises to take advantage of tools exclusive to Epicor Cloud without having to maintain infrastructure. But the move will present challenges for some organizations, particularly those in highly regulated and data-sensitive industries, analysts point out.ย 

โ€œThese organizations shouldnโ€™t just see this change as a hosting decision shift; it signals a long-term operating model change,โ€ noted Manish Jain, a principal research director at Info-Tech Research Group. Itโ€™s not customers choosing the cloud, he said, โ€œItโ€™s about vendors taking alternatives off the table.โ€

Not an overnight shift, but a fundamental one

With this move, customers will have quicker access to new features and AI-powered capabilities, such as the first ERP AI agent with outcomes-based pricing, as well as a โ€œa modern, resilient platformโ€ that reduces IT burden and operational risk, Epicor said.

Customers using on-premises versions of Kinetic, Prophet 21, and BisTrack will continue to receive support, the company noted, but final releases will roll out between 2026 and 2028, based on platform. Enterprises will then transition into โ€˜active supportโ€™ until 2029 at the latest, and โ€˜sustaining supportโ€™ will begin as early as 2027.

More than 20,000 businesses run on Epicor Cloud. Generally, Epicor Kinetic is used by mid-market and upper mid-market manufacturers, such as discrete manufacturers with complex production, supply chain, and shop-floor requirements, explained Robert Kramer, VP and principal analyst at Moor Insights & Strategy.

Wholesale and industrial distributors who require strong inventory management, pricing, and order fulfillment steer toward Prophet 21, while BisTrack is popular among building materials, lumber, and construction supply distribution sectors, he explained.

โ€œEpicor is not turning off on-premises systems overnight,โ€ Kramer emphasized, but all new capabilities, platform improvements, and long-term roadmap investments will be cloud-only.

The Epicor sunset timeline is as follows:

Kinetic

  • Final on-premises release tentatively scheduled for January 2028
  • Active support, which provides full access to Epicor phone support, security updates, new issue investigation, and more, will be offered through December 31, 2029
  • Sustaining support, which offers limited phone support, access to the latest release (but not to new modules), and an online knowledge base, begins January 1, 2030

Prophet 21

  • Final on-premises release tentatively scheduled for May 2028
  • Active support through June 30, 2029
  • Sustaining support beginning July 1, 2029

BisTrack

  • Final on-premises BisTrack Web Browser & API release tentatively scheduled for July 2028
  • Active support for on-premises BisTrack Web Browser & API through June 30, 2029
  • Sustaining support for on-premises BisTrack Web Browser & API release beginning July 1, 2029

BisTrack Desktop

  • Final on-premises release tentatively scheduled for December 2026
  • Active support through December 31, 2028
  • Sustaining support beginning January 1, 2029

BisTrack UK 3.9 (2017)

  • Active support through December 31, 2026
  • Sustaining support beginning January 1, 2027

New possibilities, different risks

This move will benefit customers looking to modernize and take advantage of the ERP systems of tomorrow: agentic AI and event-driven, noted Moorโ€™s Kramer. Benefits will include simpler infrastructure, more predictable upgrades, and access to new capabilities without the need to manage servers, databases, or patches, or to cycle through time and resource-intensive upgrades.

โ€œStaying on-prem becomes a supportable maintenance decision, not a growth one,โ€ said Kramer.

Organizations will gain the freedom to innovate and โ€œdynamically match costsโ€ with revenue through unit economics, noted Info-Techโ€™s Jain. โ€œThis move will be projected as one that favors organizations prioritizing speed, scalability, and reduced infrastructure management, especially those with limited IT capacity to maintain ERP environments at production-grade reliability.โ€

For businesses with continuous operations, tight schedules, or requiring limited downtime, operational risk moves from internal IT to the vendorโ€™s architecture, SLAs, and incident response, he explained.

Going forward, enterprises must plan for vendor-led upgrade cycles, tighter dependency on release roadmaps, and reduced control over infrastructure, he said. Cloud ERPs (whether Epicor, Microsoft, or SAP) donโ€™t eliminate risk; they reshape it. Companies trade on-prem localized failures for platform-wide dependencies that can halt entire value chains if resilience and governance arenโ€™t engineered deliberately.

โ€œFor organizations that rely on these solutions, the strategic shift is from deployment flexibility to dependency management, and many CIOs arenโ€™t fully resourcing that transition,โ€ said Jain.

Highly-regulated sectors wonโ€™t be excluded, he noted. Rather, they will be forced to adopt and combine cloud ERP cores with stricter data controls, residency requirements, and compensating governance mechanisms. Or, if they require strict data sovereignty, they may need to shift to sovereign clouds.

โ€œGoing forward, regulated industries arenโ€™t cloud-blocked; theyโ€™re architecture blocked,โ€ Jain emphasized. โ€œAs on-prem options disappear from the ERP space, compliance becomes an engineering challenge.โ€

The cloud is the future, and all enterprises must adapt

Across the board, ERP vendors, and most SaaS providers, have been converging on cloud-first models.

This helps them โ€œaccelerate innovation, standardize platforms, embed AI capabilities, and, most importantly, sustain recurring revenue,โ€ Jain pointed out. ERP companies have considered on-premises architectures as roadblocks in achieving these objectives.

Concentrating development in the cloud has become the primary way vendors deliver continuous updates, embed AI integrated analytics, and provide security at scale without โ€œforcing disruptive upgrade projects every few years,โ€ said Kramer.

โ€œMaintaining parallel on-prem and cloud platforms slows innovation and increases cost, which is why vendors are trying to draw a clearer line now,โ€ he said.

The move will allow Epicor to focus engineering, security, and innovation on a single deployment model instead of โ€œfragmenting effort across cloud and on-prem versions,โ€ he said.

Customers do give up some control and accept dependencies on a centralized service. Cloud platforms are resilient, Kramer emphasized, but outages are no longer local events that customers can mitigate with internal failover or workarounds.

For regulated or sensitive industries that canโ€™t fully pivot to public cloud, โ€œthis does not mean an immediate cliff, but it does narrow long-term options,โ€ he pointed out. Hybrid, private cloud, and sovereign deployments will become the middle ground, but they come with their own challenges, requiring more deliberate planning, stronger governance, and clearer accountability.

โ€œOver time, even highly regulated organizations will be pushed to modernize how they consume ERP,โ€ Kramer noted. โ€œNot because on-prem stops working, but because it gradually stops evolving in ways that support new business and regulatory demands.โ€

GPU ์ˆ˜์š” ๊ธ‰์ฆ์—โ€ฆAWS, EC2 ์šฉ๋Ÿ‰ ์˜ˆ์•ฝ ์š”๊ธˆ ์ธ์ƒ

AWS๊ฐ€ ๋จธ์‹ ๋Ÿฌ๋‹(ML)์šฉ ์—˜๋ผ์Šคํ‹ฑ ์ปดํ“จํŠธ ํด๋ผ์šฐ๋“œ(EC2) ์šฉ๋Ÿ‰ ์˜ˆ์•ฝ ์„œ๋น„์Šค ์ผ๋ถ€์— ๋Œ€ํ•œ ์š”๊ธˆ ๊ตฌ์กฐ๋ฅผ ์กฐ์ •ํ–ˆ๋‹ค. ์š”๊ธˆ์ด ์•ฝ 15% ์ธ์ƒ๋œ ์ด๋ฒˆ ์กฐ์น˜๋Š” ๋Œ€๊ทœ๋ชจ ML ์›Œํฌ๋กœ๋“œ๋ฅผ ๊ณ„ํšํ•˜๋Š” ๊ธฐ์—…์— ์˜ํ–ฅ์„ ๋ฏธ์น  ์ˆ˜ ์žˆ๋‹ค.

์•„๋งˆ์กด ์šฉ๋Ÿ‰ ๋ธ”๋ก(Amazon Capacity Blocks)์€ ๊ธฐ์—…์ด ํ–ฅํ›„ ํŠน์ • ์‹œ์ ๋ถ€ํ„ฐ ๊ฐ€์† ์ปดํ“จํŒ… ์ž์›์— ์ ‘๊ทผํ•  ์ˆ˜ ์žˆ๋Š” ๊ถŒํ•œ์„ ์‚ฌ์ „์— ํ™•๋ณดํ•˜๋„๋ก ํ•˜๋Š” ์˜ˆ์•ฝ ์„œ๋น„์Šค๋‹ค. AWS๋Š” ๊ธฐ์—…์ด 1๊ฐœ๋ถ€ํ„ฐ 64๊ฐœ ์ธ์Šคํ„ด์Šค ๊ทœ๋ชจ์˜ ํด๋Ÿฌ์Šคํ„ฐ๋ฅผ ์ตœ๋Œ€ 6๊ฐœ์›”๊นŒ์ง€ ์˜ˆ์•ฝํ•ด ๋‹ค์–‘ํ•œ ML ์›Œํฌ๋กœ๋“œ๋ฅผ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ง€์›ํ•œ๋‹ค. ์ด๋Š” GPU ๊ธฐ์ค€ ์ตœ๋Œ€ 512๊ฐœ, ํŠธ๋ ˆ์ด๋‹ˆ์—„(Trainium) ์นฉ ๊ธฐ์ค€์œผ๋กœ๋Š” ์ตœ๋Œ€ 1,024๊ฐœ์— ํ•ด๋‹นํ•œ๋‹ค. ๋‹ค๋งŒ EC2 ์šฉ๋Ÿ‰ ๋ธ”๋ก ์˜ˆ์•ฝ์€ ์ตœ๋Œ€ 8์ฃผ ์ „๊นŒ์ง€๋งŒ ์‚ฌ์ „ ์˜ˆ์•ฝ์ด ๊ฐ€๋Šฅํ•˜๋‹ค.

ํ•œํŽธ AWS๋Š” ์ง€๋‚œํ•ด 6์›” P4์™€ P5 ์ธ์Šคํ„ด์Šค๋ฅผ ํฌํ•จํ•œ ์—”๋น„๋””์•„ GPU ๊ฐ€์† EC2 ์ธ์Šคํ„ด์Šค์˜ ์š”๊ธˆ์„ ์ตœ๋Œ€ 45% ์ธํ•˜ํ•œ ๋ฐ” ์žˆ๋‹ค. ๋ถˆ๊ณผ ๋ช‡ ๋‹ฌ ์ „๊นŒ์ง€ GPU ์ธ์Šคํ„ด์Šค ์ „๋ฐ˜์˜ ์š”๊ธˆ์„ ๋‚ฎ์ถ”๋Š” ์ „๋žต์„ ์ทจํ–ˆ๋˜ ๋งŒํผ, ์ด๋ฒˆ ์กฐ์ •์€ ์ผ๋ฐ˜ ์ธ์Šคํ„ด์Šค๊ฐ€ ์•„๋‹Œ ๋ณด์žฅํ˜• ์˜ˆ์•ฝ ์„œ๋น„์Šค์—๋งŒ ๋ณ„๋„์˜ ๊ฐ€๊ฒฉ ์ •์ฑ…์„ ์ ์šฉํ•œ๋‹ค๋Š” ์ทจ์ง€๋กœ ํ•ด์„๋œ๋‹ค.

P5 ๊ณ„์—ด ์˜ˆ์•ฝ ์„œ๋น„์Šค ์ „๋ฐ˜์—์„œ ์š”๊ธˆ ์ƒ์Šน

์ด๋ฒˆ ์š”๊ธˆ ์กฐ์ •์€ ๊ณ ์„ฑ๋Šฅ GPU๋ฅผ ์ถ”ํ›„์— ํ™•์ •์ ์œผ๋กœ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋Š” EC2 ์šฉ๋Ÿ‰ ์˜ˆ์•ฝ ์„œ๋น„์Šค ์ „๋ฐ˜์— ์ ์šฉ๋œ๋‹ค. ํ•ด๋‹น ์„œ๋น„์Šค์—๋Š” ์ตœ์‹  ์—”๋น„๋””์•„ ๋ธ”๋ž™์›ฐ GPU๋ฅผ ํƒ‘์žฌํ•œ EC2 P6 ์ธ์Šคํ„ด์Šค, ์—”๋น„๋””์•„ H100 ๋ฐ H200 ํ…์„œ ์ฝ”์–ด GPU ๊ธฐ๋ฐ˜์˜ P5 ์ธ์Šคํ„ด์Šค, ์—”๋น„๋””์•„ A100 ํ…์„œ ์ฝ”์–ด GPU๋กœ ๊ตฌ๋™๋˜๋Š” P4 ์ธ์Šคํ„ด์Šค๊ฐ€ ํฌํ•จ๋œ๋‹ค.

๋ฏธ๊ตญ ๋™๋ถ€(์˜คํ•˜์ด์˜ค) ๋ฆฌ์ „์—์„œ ์—”๋น„๋””์•„ H200 ๊ฐ€์†๊ธฐ 8๊ฐœ๋ฅผ ์žฅ์ฐฉํ•œ p5e.48xlarge ์ธ์Šคํ„ด์Šค์˜ ๊ฒฝ์šฐ, ์œ ํšจ ์‹œ๊ฐ„๋‹น ์š”๊ธˆ์ด ๊ฐ 34.608๋‹ฌ๋Ÿฌ์—์„œ 39.799๋‹ฌ๋Ÿฌ๋กœ ์ธ์ƒ๋๋‹ค.

๊ฐ™์€ ๋ฆฌ์ „์˜ p5en.48xlarge ์ธ์Šคํ„ด์Šค ์—ญ์‹œ 36.184๋‹ฌ๋Ÿฌ์—์„œ 41.612๋‹ฌ๋Ÿฌ๋กœ ๊ฐ€๊ฒฉ์ด ์˜ฌ๋ž๋‹ค. ์ด ์š”๊ธˆ์€ ์œ ๋Ÿฝ์˜ ์Šคํ†กํ™€๋ฆ„, ๋Ÿฐ๋˜, ์ŠคํŽ˜์ธ๊ณผ ์•„์‹œ์•„ํƒœํ‰์–‘ ์ง€์—ญ์˜ ์ž์นด๋ฅดํƒ€, ๋ญ„๋ฐ”์ด, ๋„์ฟ„, ์„œ์šธ ๋“ฑ์—์„œ๋„ ๋™์ผํ•˜๊ฒŒ ์ ์šฉ๋œ๋‹ค. ๋‹ค๋งŒ ๋ฏธ๊ตญ ์„œ๋ถ€(๋ถ๋ถ€ ์บ˜๋ฆฌํฌ๋‹ˆ์•„) ๋ฆฌ์ „์—์„œ๋Š” p5e.48xlarge๊ฐ€ 43.26๋‹ฌ๋Ÿฌ์—์„œ 49.749๋‹ฌ๋Ÿฌ๋กœ, p5en.48xlarge๋Š” 45.23๋‹ฌ๋Ÿฌ์—์„œ 52.015๋‹ฌ๋Ÿฌ๋กœ ๊ฐ๊ฐ ๋” ๋†’์€ ์š”๊ธˆ์ด ์ฑ…์ •๋๋‹ค.

๋ฐ˜๋ฉด P6e ๊ณ„์—ด์˜ ๊ฐ€๊ฒฉ์€ ๊ฑฐ์˜ ๋ณ€๋™์ด ์—†๋‹ค. ๋‹ฌ๋ผ์Šค ๋กœ์ปฌ ์กด์—์„œ p4d.24xlarge ๊ธฐ์ค€์œผ๋กœ 72๊ฐœ์˜ B200 ๊ฐ€์†๊ธฐ๋ฅผ ์ œ๊ณตํ•˜๋Š” P6e ์ธ์Šคํ„ด์Šค๋Š” ์‹œ๊ฐ„๋‹น 761.904๋‹ฌ๋Ÿฌ๋กœ ๊ธฐ์กด ๊ฐ€๊ฒฉ์ด ์œ ์ง€๋๋‹ค.

EIIRํŠธ๋ Œ๋“œ ๋ฐ ํŒŒ๋ฆฌํฌ์ปจ์„คํŒ…์˜ CEO ํŒŒ๋ฆฌํฌ ์ž์ธ์€ ์ด๋ฒˆ ์š”๊ธˆ ์กฐ์ •์˜ ๋ฐฐ๊ฒฝ์— ๋Œ€ํ•ด โ€œ์ˆ˜์š”์™€ ๊ณต๊ธ‰์— ๋”ฐ๋ฅธ ์‹œ์žฅ ๊ฐ€๊ฒฉ ์ฑ…์ •์œผ๋กœ ๋ณด๋Š” ๊ฒƒ์ด ๊ฐ€์žฅ ํƒ€๋‹นํ•˜๋‹คโ€๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค. ์ž์ธ์€ โ€œH100๊ณผ H200 GPU์— ๋Œ€ํ•œ ์ˆ˜์š”๊ฐ€ ๊ณต๊ธ‰์„ ํฌ๊ฒŒ ์›ƒ๋Œ๋ฉด์„œ, AWS๊ฐ€ ๋ณด์žฅํ˜• GPU ๋ฌผ๋Ÿ‰์— ์‚ฌ์‹ค์ƒ ํ”„๋ฆฌ๋ฏธ์—„์„ ์ ์šฉํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์ „์ฒด ์ธํ”„๋ผ ์šฉ๋Ÿ‰์ด ์•„๋‹ˆ๋ผ, ๊ธด๊ธ‰ํ•˜๊ฒŒ GPU ์šฉ๋Ÿ‰์„ ํ•„์š”๋กœ ํ•˜๋Š” ๊ณ ๊ฐ์„ ๋Œ€์ƒ์œผ๋กœ ๋” ๋†’์€ ์ธํ”„๋ผ ๋ฐ ์ž๋ณธ ๋น„์šฉ์„ ํšŒ์ˆ˜ํ•˜๋ ค๋Š” ์ „๋žตโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ณด์žฅํ˜• GPU ์˜ˆ์•ฝ ์„œ๋น„์Šค๊ฐ€ ์ƒˆ๋กœ์šด ๊ฒฝ์Ÿ ์‹œ์žฅ์œผ๋กœ ๋ถ€์ƒ

GPU ํด๋Ÿฌ์Šคํ„ฐ์— ๋Œ€ํ•œ ์ ‘๊ทผ์ด ๋ณด์žฅ๋˜๋ฉด ๊ธฐ์—…์€ AI ์ธํ”„๋ผ ๊ณ„ํš ๊ณผ์ •์—์„œ ๋ถˆํ™•์‹ค์„ฑ์„ ์ค„์ด๊ณ , ํ–ฅํ›„ ๊ณต๊ธ‰ ๋ณ€๋™์„ฑ์— ๋Œ€๋น„ํ•œ ์•ˆ์ •์„ฑ์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋‹ค. ์—”๋น„๋””์•„ H100๊ณผ H200 ๋“ฑ ๊ณ ์‚ฌ์–‘ GPU์— ๋Œ€ํ•œ ์ˆ˜์š”๊ฐ€ ๊ธ‰์ฆํ•˜๋ฉด์„œ ๊ณต๊ธ‰ ๋ถ€์กฑ์ด ์‹ฌํ™”๋˜์ž, ์ฃผ์š” ํด๋ผ์šฐ๋“œ ์—…์ฒด๋Š” ๊ธฐ์—…์˜ GPU ์‚ฌ์šฉ์„ ๋ณด์žฅํ•˜๋Š” ์˜ˆ์•ฝ ์„œ๋น„์Šค๋ฅผ ์ ์ฐจ ํ™•๋Œ€ํ•˜๊ณ  ์žˆ๋‹ค.

AWS ์™ธ์—๋„ ๊ตฌ๊ธ€๊ณผ ๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ(MS)๊ฐ€ ์œ ์‚ฌํ•œ ์„œ๋น„์Šค๋ฅผ ์ œ๊ณตํ•˜๊ณ  ์žˆ๋‹ค. ๋‹ค๋งŒ ์ด๋“ค ์„œ๋น„์Šค๋Š” ๋ณด๋‹ค ์ „ํ†ต์ ์ธ ์˜ˆ์•ฝ ๋ชจ๋ธ๊ณผ ์Šค์ผ€์ค„๋ง ๋ฐฉ์‹์œผ๋กœ ๊ตฌ์„ฑ๋ผ ์žˆ๋‹ค๋Š” ์ ์—์„œ AWS์˜ ์ ‘๊ทผ ๋ฐฉ์‹๊ณผ๋Š” ์ฐจ์ด๋ฅผ ๋ณด์ธ๋‹ค.

์˜ˆ๋ฅผ ๋“ค์–ด ๊ตฌ๊ธ€ํด๋ผ์šฐ๋“œ๋Š” ๊ธฐ์—…์ด ์บ˜๋ฆฐ๋” ๊ธฐ๋ฐ˜ ๋„๊ตฌ๋ฅผ ํ™œ์šฉํ•ด GPU ์šฉ๋Ÿ‰์„ ๊ณ ์ •๋œ ๋ธ”๋ก ๋‹จ์œ„๋กœ ์‚ฌ์ „์— ์˜ˆ์•ฝํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•˜๋Š” ์Šค์ผ€์ค„๋ง ๊ธฐ๋Šฅ์„ ๋„์ž…ํ–ˆ๋‹ค. ๊ทธ๋ ˆ์ดํ•˜์šด๋“œ ๋ฆฌ์„œ์น˜์˜ CEO ๊ฒธ ์ˆ˜์„ ์• ๋„๋ฆฌ์ŠคํŠธ์ธ ์‚ฐ์นซ ๋น„๋ฅด ๊ณ ๊ธฐ์•„๋Š” โ€œํ˜•์‹๋งŒ ๋ณด๋ฉด AWS์˜ ์šฉ๋Ÿ‰ ๋ธ”๋ก๊ณผ ์ƒ๋‹นํžˆ ์œ ์‚ฌํ•ด ๋ณด์ธ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์ ‘๊ทผ ๋ฐฉ์‹์€ ๋‹ค๋ฅด๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ณ ๊ธฐ์•„๋Š” โ€œ๊ตฌ๊ธ€์€ ์ด๋ฅผ ํ”„๋ฆฌ๋ฏธ์—„ ์ƒํ’ˆ์ด ์•„๋‹ˆ๋ผ, ์ „์ฒด ์ž์› ์Šค์ผ€์ค„๋Ÿฌ์˜ ์ผ๋ถ€๋กœ ๋‹ค๋ฃจ๊ณ  ์žˆ๋‹ค. ๋ณด์žฅ์€ ์œ ์ง€๋˜์ง€๋งŒ, ๊ฐ€๊ฒฉ ๊ตฌ์กฐ๊ฐ€ ์ง€๋‚˜์น˜๊ฒŒ ์„ธ๋ถ„ํ™”๋˜๊ฑฐ๋‚˜ ์ˆ˜์‹œ๋กœ ๋ณ€ํ•œ๋‹ค๋Š” ๋А๋‚Œ์€ ์ ๋‹คโ€๋ผ๊ณ  ํ‰๊ฐ€ํ–ˆ๋‹ค. ์ด์–ด โ€œ๊ฐ€๊ฒฉ ๊ฒฝ์Ÿ๋ณด๋‹ค๋Š” ์Šค์ผ€์ค„๋ง์„ ํ†ตํ•ด ์ฐจ๋ณ„ํ™”๋ฅผ ์‹œ๋„ํ•˜๋Š” ๊ฒƒ์ฒ˜๋Ÿผ ๋ณด์ด๋ฉฐ, ์ผ๋ถ€ ์›Œํฌ๋กœ๋“œ๋ฅผ GPU ๋Œ€์‹  TPU๋กœ ์ „ํ™˜ํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ์ ๋„ ์‹œ์Šคํ…œ ์ „๋ฐ˜์˜ ์œ ์—ฐ์„ฑ์„ ๋†’์—ฌ์ค€๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋ฐ˜๋ฉด MS์˜ ์• ์ €๋Š” ๋ฆฌ์ „ ๋‹จ์œ„์˜ ์šฉ๋Ÿ‰ ์˜ˆ์•ฝ์— ๋” ๋ฌด๊ฒŒ๋ฅผ ๋‘๊ณ  ์žˆ๋‹ค. ๊ณ ๊ธฐ์•„๋Š” โ€œํŠน์ • ๋ฆฌ์ „๊ณผ ๊ฐ€์šฉ ์˜์—ญ์—์„œ ํŠน์ • ๊ฐ€์ƒ๋จธ์‹ (VM) ์œ ํ˜•์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•˜์ง€๋งŒ, ์žฅ๊ธฐ ๊ณ„ํš๊ณผ ๋Œ€๊ทœ๋ชจ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ์•ฝ์ •์„ ์„ ํ˜ธํ•˜๋Š” ๊ตฌ์กฐโ€๋ผ๋ฉฐ โ€œ๋ณด์žฅ๋œ ์šฉ๋Ÿ‰์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋Š” ๋Œ€์‹ , ์‹ค์ œ ์‚ฌ์šฉ ์—ฌ๋ถ€์™€ ๊ด€๊ณ„์—†์ด ํ•ด๋‹น ์ž์›์„ ์œ ์ง€ํ•˜๋Š” ๋น„์šฉ์„ ๋ถ€๋‹ดํ•ด์•ผ ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด์–ด โ€œ์ด๋Š” ์š”๊ธˆ ์ž์ฒด๋ณด๋‹ค ์žฅ๊ธฐ๊ฐ„ ์•ฝ์ •๊ณผ ์ž์› ์„ ์ ์— ํ”„๋ฆฌ๋ฏธ์—„์„ ๋ถ™์ด๋Š” ๋˜ ๋‹ค๋ฅธ ์ข…๋ฅ˜โ€๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

๊ฐ€๊ฒฉ ์ธ์ƒ์—๋„ ๋‹จ๊ธฐ ์˜ํ–ฅ์€ ์ œํ•œ์ 

์ „๋ฌธ๊ฐ€๋“ค์€ ์ด๋Ÿฌํ•œ ๋ณด์žฅํ˜• GPU ์˜ˆ์•ฝ ์„œ๋น„์Šค๊ฐ€ ์ผ๋ฐ˜ ์ปดํ“จํŒ… ์ž์›๊ณผ ๋น„๊ตํ•˜๋ฉด ์ „์ฒด ํด๋ผ์šฐ๋“œ ์ง€์ถœ์—์„œ ์ฐจ์ง€ํ•˜๋Š” ๋น„์ค‘์€ ํฌ์ง€ ์•Š๋‹ค๊ณ  ๋ณด๊ณ  ์žˆ๋‹ค. ๋‹ค๋งŒ ๊ธฐ์—…์˜ ์ „๋žต์  AI ํˆฌ์ž ๊ด€์ ์—์„œ๋Š” ์ƒ๋Œ€์ ์œผ๋กœ ๋น„์ค‘์ด ๋†’์•„ ์˜๋ฏธ๊ฐ€ ํฌ๋‹ค๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

AWS๊ฐ€ ๊ฐ€๊ฒฉ ์ธ์ƒ์„ ๊ฐ€์žฅ ๋จผ์ € ๊ณต์‹ํ™”ํ–ˆ์ง€๋งŒ, ๊ณ ๊ธฐ์•„๋Š” ๋‹ค๋ฅธ ํด๋ผ์šฐ๋“œ ์—…์ฒด๋„ ํ‘œํ˜„ ๋ฐฉ์‹๋งŒ ๋‹ค๋ฅผ ๋ฟ ๋™์ผํ•œ ์ „๋žต์„ ๋”ฐ๋ฅด๊ณ  ์žˆ๋‹ค๊ณ  ํ‰๊ฐ€ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ๊ฒ‰์œผ๋กœ ๋“œ๋Ÿฌ๋‚˜๋Š” ์„ค๋ช…์€ ๋‹ค๋ฅผ ์ˆ˜ ์žˆ์ง€๋งŒ, ์‹ค์ œ๋กœ๋Š” ๊ฐ™์€ ๋ฐฉํ–ฅ์œผ๋กœ ์›€์ง์ด๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์ž์ธ์€ ๋Œ€๋ถ€๋ถ„์˜ ๊ธฐ์—…์—์„œ ์ด๋ฒˆ ๊ฐ€๊ฒฉ ์ธ์ƒ์ด ์ฆ‰๊ฐ์ ์ธ ์›Œํฌ๋กœ๋“œ ์ด์ „์œผ๋กœ ์ด์–ด์งˆ ๊ฐ€๋Šฅ์„ฑ์€ ๋‚ฎ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. EC2 ์šฉ๋Ÿ‰ ๋ธ”๋ก ์„œ๋น„์Šค๊ฐ€ ์ „์ฒด GPU ์ง€์ถœ์—์„œ ์ฐจ์ง€ํ•˜๋Š” ๋น„์ค‘์ด ํฌ์ง€ ์•Š๊ณ , ๋ฐ์ดํ„ฐ ์ข…์†์„ฑ, ์ด๋ฏธ ๊นŠ๊ฒŒ ์ž๋ฆฌ ์žก์€ ML์˜ต์Šค(MLOps) ํ™˜๊ฒฝ, ๊ทœ์ œ ๋ฐ ์ปดํ”Œ๋ผ์ด์–ธ์Šค ์š”๊ฑด, ๊ทธ๋ฆฌ๊ณ  ์ถ•์ ๋œ ์กฐ์ง ์—ญ๋Ÿ‰์ด ์—ฌ์ „ํžˆ ์›Œํฌ๋กœ๋“œ๋ฅผ AWS์— ๋ฌถ์–ด๋‘๊ณ  ์žˆ๋‹ค๋Š” ์ด์œ ์—์„œ๋‹ค. ์„ฑ์ˆ™๋„๊ฐ€ ๋†’์•„์ง„ ML ํ•™์Šต ํ™˜๊ฒฝ์„ AWS ์™ธ๋ถ€๋กœ ์ด์ „ํ•˜๋Š” ์ž‘์—…์ด ๋ณต์žกํ•˜๊ณ  ์‹œ๊ฐ„์ด ๋งŽ์ด ๋“œ๋Š” ๋งŒํผ, ์ด๋ฒˆ ๊ฐ€๊ฒฉ ์ธ์ƒ์˜ ์˜ํ–ฅ์€ ๊ธฐ์กด ์‹œ์Šคํ…œ๋ณด๋‹ค๋Š” ์‹ ๊ทœ AI ์›Œํฌ๋กœ๋“œ์—์„œ ๋” ํฌ๊ฒŒ ๋‚˜ํƒ€๋‚  ์ „๋ง์ด๋‹ค.
dl-ciokorea@foundryco.com

How Microsoft is betting on AI agents in Windows, dusting off a winning playbook from the past

The cover of Microsoftโ€™s 1990 annual report, showing Microsoft Word for Windows 3.0, reflected the companyโ€™s confidence as Windows was emerging as a true platform.

[Editorโ€™s Note:ย Agents of Transformationย is an independent GeekWire series and March 24, 2026 event, underwritten by Accenture, exploring the people, companies, and ideas behind AI agents.]

It was โ€œlike bringing a Porsche into a world of Model Ts.โ€ย 

Thatโ€™s what Microsoft said in its 1990 annual report about the shift from MS-DOS to Windows. But the bigger breakthrough for the company wasnโ€™t the graphical interface. It was Windowsโ€™ ability to serve as a platform for applications made by others.

Windows 3.0, released that year, made third-party software easier to find and launch, and offered developers a clear bargain: build to Microsoftโ€™s specs, and your software would become a first-class citizen on the computers that were arriving โ€œon every desk and in every home,โ€ as the companyโ€™s original mission statement put it.ย 

Thirty-five years later, AI feels less like a car and more like a rocket ship. But Microsoft is hoping that Windows can once again serve as the platform where it all takes off.

A new framework called Agent Launchers, introduced earlier this month as a preview in the latest Windows Insider build, lets developers register agents directly with the operating system. They can describe an agent through whatโ€™s known as a manifest, which then lets the agent show up in the Windows taskbar, inside Microsoft Copilot, and across other apps.

The long-term promise for Windows users is autonomous assistants that operate on their behalf, directly on their machines. Beyond routine tasks like assembling a PDF or organizing files, agents could monitor email and calendars to resolve scheduling conflicts, or scan documents across multiple apps to pull together a briefing for an upcoming meeting.

Achieving that level of autonomy requires more than just a clever interface. It will take deep, persistent memory that operates more like the human brain.

Microsoft CEO Satya Nadella this week framed AI agents as a new layer of computing infrastructure that requires greater engineering sophistication. Windows is one of the places where Microsoft is attempting to implement that vision. (GeekWire File Photo / Kevin Lisota)

โ€œWe are now entering a phase where we build rich scaffolds that orchestrate multiple models and agents; account for memory and entitlements; enable rich and safe tools use,โ€ Microsoft CEO Satya Nadella wrote in a blog post this week looking ahead to 2026. โ€œThis is the engineering sophistication we must continue to build to get value out of AI in the real world.โ€

Elements of this are already emerging elsewhere.

  • Googleโ€™s Gemini and Anthropicโ€™s Claude offer desktop-style agents through browsers and native apps, with extensions that can read pages, fill forms, and take limited actions on a userโ€™s behalf.
  • Amazon is developing โ€œfrontier agentsโ€ aimed at automating business processes in the cloud.ย 
  • Startups like Seattle-based Vercept are building standalone agentic apps that coordinate work across tools.ย 

But Microsoftโ€™s Windows team is betting that agents tightly linked to the operating system will win out over ones that merely run on top of it, just as a new class of Windows apps replaced a patchwork of DOS programs in the early days of the graphical operating system.ย 

Microsoft 365 Copilot is using the Agent Launchers framework for first-party agents like Analyst, which helps users dig into data, and Researcher, which builds detailed reports. Software developers will be able to register their own agents when an app is installed, or on the fly based on things like whether a user is signed in or paying for a subscription.

The risks posed by PC agents

The parallels to the past only go so far. Traditional PC applications ran in their own windows, worked with their own files, and didnโ€™t touch the rest of the system for the most part.

โ€œAgents are going to need to be able to scratchpad their work,โ€ Microsoft CTO Kevin Scott said recently on the South Park Commons Minus 1 podcast, explaining that agents will need to retain a history of user interactions and tap into the necessary context to solve problems.

Agents are meant to maintain this context across apps, ask follow-up questions, and take actions on a userโ€™s behalf. That requires a different level of trust than Windows has ever had to manage, which is already raising difficult questions for the company.

Microsoft acknowledges that agents introduce unique security risks. In a support document, the company warned that malicious content embedded in files or interface elements could override an agentโ€™s instructions โ€” potentially leading to stolen data or malware installation.

To address this, Microsoft says it has built a security framework that runs agents in their own contained workspace, with a dedicated user account that has limited access to user folders. The idea is to create a boundary between the agent and what the rest of the system can access.

The agentic features are off by default, and Microsoft is advising users to โ€œunderstand the security implications of enabling an agent on your computerโ€ before turning them on.

A different competitive landscape

Even if Microsoft executes perfectly, the landscape is different now. In the early 1990s, Windows became dominant because developers flocked to the platform, which attracted more users, which attracted more developers. It was a virtuous cycle, and Microsoft was at the heart of it.

But Windows isnโ€™t the center of the computing world anymore. Smartphones, browsers, and cloud platforms have fragmented the landscape in ways that didnโ€™t exist back then. Microsoft missed the mobile era almost entirely, and the PC is now one screen among many.

In the enterprise, Microsoft has better footing. Azure, Microsoft 365 Copilot, and a growing ecosystem of business-focused agents give the company a strong position, competing against Google, Amazon, OpenAI and others for cloud-based AI agents and services.

Agent Launchers is a different bet โ€” an attempt to make Windows the home for agents that serve individual users on their own machines. Thatโ€™s a harder sell when the PC is competing with phones, browsers, and cloud apps for peopleโ€™s attention. Microsoft can build the platform, but it canโ€™t guarantee that developers will show up the way they did 35 years ago.

And unlike in the 1990s, Microsoft canโ€™t count on users to embrace what itโ€™s building. Thereโ€™s a growing sentiment that these AI capabilities are being pushed into Windows not because users want them, but because Microsoft needs to justify its massive AI investments.ย 

In October, for example, Microsoft announced new features including โ€œHey Copilotโ€ voice activation, a redesigned taskbar with Copilot built in, and the expansion of โ€œCopilot Actionsโ€ agentic capabilities beyond the browser to the PC itself.ย 

โ€œTheyโ€™re thinking about revenue first and foremost,โ€ longtime tech journalist and Microsoft observer Ed Bott said on the GeekWire Podcast at the time. The more users rely on these AI features, he explained, the easier it becomes for the company to upsell them on premium services.

There is a business reality driving all of this. In Microsoftโ€™s most recent fiscal year, Windows and Devices generated $17.3 billion in revenue โ€” essentially flat for the past three years.ย 

Thatโ€™s less than Gaming ($23.5 billion) and LinkedIn ($17.8 billion), and a fraction of the $98 billion in revenue from Azure and cloud services or the nearly $88 billion from Microsoft 365 commercial.

By comparison, in fiscal 1995, five years after the launch of Windows 3.0, Microsoftโ€™s platforms group (which included MS-DOS and Windows) represented about 40% of its total revenue of $5.9 billion. Windows was the growth engine for the company.

Windows is unlikely to play that kind of outsized role again. But AI integration is the companyโ€™s best bet to return the OS to growth. Whether that ultimately looks like a restored Porsche or a rocket ship on the launchpad probably doesnโ€™t matter as much as keeping it out of the junkyard.

The stories that defined 2025: AI dreams, brutal realities, and Seattle tech at a turning point

An illustration by ChatGPT based on its interpretation of our year-end GeekWire Podcast discussion.

The past year may go down as one of the most consequential in technology history, in both the Seattle tech community and the world. But in some ways, itโ€™s not without precedent.

As we sat down to reflect on the past year, we rewound all the way back to January โ€” when, as part of a larger discussion with Bill Gates, we asked the Microsoft co-founder to compare the early days of the PC with these early years of AI.

Gates reflected on the PC era as a moment of computing becoming free, effectively.

โ€œNow whatโ€™s happening is intelligence is becoming free,โ€ he said, โ€œand thatโ€™s even more profound than computing becoming free.โ€

As we looked through GeekWireโ€™s top stories of the year, almost every one felt like a subplot to that larger narrative. On this special year-end episode of the GeekWire Podcast, we reviewed the articles that resonated most with readers, and compared notes to make sense of it all.

Listen below, and continue reading for episode notes and links.

Enigma of success: โ€˜Brutal realityโ€™ of tech cycles

  • Best of times, worst of times: Massive AI infrastructure spending alongside widespread layoffs.
  • Satya Nadella on the Stargate announcement: โ€œIโ€™m good for my $80 billion.โ€œ
  • The unexpected way AI is affecting jobs โ€” not by replacing workers directly, but by pressuring companies to cut costs as they pour money into infrastructure.
  • MIT study: 95% of projects using generative AI have failed or produced no return.
  • Worker stress: Mandates to use AI, but no playbook on how.
  • One tech veteranโ€™s take: โ€œThe enigma of success is a polite way of describing the brutal reality of tech cycles. โ€ฆ The challenge, and opportunity for leadership, is whether the bets actually compound into something durable, or just become another slide deck for next yearโ€™s reorg.โ€
  • Bill Radke on KUOW: โ€œThe tech industry had quite a year. Amazon ordered their workers back to the office. You must come back to the office. Are you here? Good. Youโ€™re laid off. Not all of you. Just the humans.โ€œ

A pivotal year for Amazon

  • Andy Jassyโ€™s explanation: Not financially driven, not even really AI driven โ€” itโ€™s culture.
  • After rapid growth, Amazon trying to get back to operating like โ€œthe worldโ€™s largest startup.โ€
  • The new motto seems to be: Get small and nimble, faster.
  • Can Amazon find that next pillar of business, as Jeff Bezos used to say?

Coding is dead, computer science is not

Seattleโ€™s future as a tech hub

Sense of place: More important for some, less for others

  • Amazon brings employees back five days a week; Microsoft announces three days starting in 2026.
  • Rebooting Redmond: The conclusion of our Microsoft 50th anniversary series explored the new campus and what it signals.
  • Yet many startups are more distributed and diffuse than ever โ€” sometimes itโ€™s hard to even pin down where their headquarters are.
  • Statsig, entirely in-office in Bellevue, acquired by OpenAI for $1.1 billion.
  • The perennial question: Why donโ€™t more of these companies become Seattleโ€™s next tech giant?

M&A and IPOs: Base hits, not home runs

  • Didnโ€™t see as much deal activity as some predicted for 2025.
  • GeekWire deals list reflects smaller acquisitions, not blockbusters.
  • One tech IPO from Washington state: Kestra Medical Technologies, $202 million in March.
  • Complex alchemy of interest rates, regulation, and market conditions.

AI becomes real

  • Brad Smith at Microsoftโ€™s annual meeting: Asked Copilotโ€™s researcher agent to produce a report on an issue from seven or eight years ago. Fifteen minutes later: 25-page report with 100 citations.
  • Whatโ€™s happening now: the shift from individual productivity to team productivity, from people using AI to organizations figuring it out.
  • As companies implement AI agents, we move from desktop/individual applications to true enterprise services, playing to Seattleโ€™s strengths.

Quote of the Year

โ€œWe look forward to joining Matt on his private island next year.โ€ โ€” Kiana Ehsani, CEO of Vercept, after her co-founder Matt Deitke left to join Meta for a reported hundreds of millions of dollars.

Stickler of the Year

Proud Seattleite and grammarian Ken Jennings on Jeopardy!, correcting a contestant: โ€œSorry, Dan, we are sticklers in Seattle. Itโ€™s Pike Place โ€” no s.โ€

Feel-Good Moment of the Year

Ambika Singh, CEO and founder of Armoire, accepting the Workplace of the Year award at the GeekWire Awards: โ€œIt is not a surprise to any of you that we are losing community outside of these walls in this country. But here, it feels alive and well.โ€

Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.

Audio editing by Curt Milton

โŒ