Mercor’s APEX-Agents benchmark finds top AI models score under 25% accuracy on realistic consulting, legal, and finance tasks.
The post New study shows AI isn’t ready for office work appeared first on Digital Trends.
This week on the GeekWire Podcast: Newly unsealed court documents reveal the behind-the-scenes history of Microsoft and OpenAI, including a surprise: Amazon Web Services was OpenAI’s original partner. We tell the story behind the story, explaining how it all came to light.
Plus, Microsoft CEO Satya Nadella debuts a new AI catchphrase at Davos, startup CEO Dave Clark stirs controversy with his “wildly productive weekend,” Elon Musk talks aliens, and the latest on Seattle-area physical AI startups, including Overland AI and AIM Intelligent Machines.
Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.
With GeekWire co-founders John Cook and Todd Bishop; edited by Curt Milton.
Bicycles for the mind. … Information at your fingertips. … Managers of infinite minds?
Microsoft CEO Satya Nadella riffed on some famous lines from tech leaders past this week in an appearance at the World Economic Forum in Davos, Switzerland, and offered up his own trippy candidate to join the canon of computing metaphors.
Nadella traced the lineage in a conversation with former UK Prime Minister Rishi Sunak.
And now? “All of us are going to be managers of infinite minds,” Nadella said. “And so if we have that as the theory, then the question is, what can we do with it?”
He was referring to AI agents — the autonomous software that can take on tasks, work through problems, and keep going while you sleep. Microsoft and others have been talking for the better part of a year now about people starting to oversee large fleets of them.
Nadella said it’s already reshaping how teams are structured. At Microsoft-owned LinkedIn, the company has merged design, program management, product management, and front-end engineering into a single new role: full-stack builders. Overall, he called it the biggest structural change to software teams he’s seen in a career that started at Microsoft in the 1990s.
“The jobs of the future are here,” Nadella said, putting his own spin on a famous line often attributed to sci-fi writer William Gibson. “They’re just not evenly distributed.”
Nadella’s comments came during a live stream for LinkedIn Premium members, hosted from Davos by LinkedIn VP and Editor in Chief Daniel Roth, after Sunak mentioned his two teenage daughters, and the world they’ll enter. Young people may not manage lots of people at age 20 or 21, he said, “but they will be managing a team of agents.”
Sunak was referencing an essay by Goldman Sachs CIO Marco Argenti in Time.
The agentic shift, Argenti wrote, requires “moving from being a sole performer to an orchestra conductor” — your team now includes AI agents that “must be guided and supervised with the same approach you would apply to a new, junior colleague.”
Nadella agreed, saying “we do need a new theory of the mind” to navigate what’s coming, before he offered up his new metaphor about managing infinite minds.
In other remarks at Davos, Nadella made headlines with his warning that AI’s massive energy demands risk eroding its “social permission” unless it delivers tangible benefits in health, education, and productivity. Energy costs, he added, will decide the AI race’s winners, with GDP growth tied to cheap power for processing AI tokens.
Whether “infinite minds” catches on like “bicycles” and “fingertips” remains to be seen. But it’s definitely more psychedelic. And if this shift is stranger than what came before, maybe we do need a mind-expanding metaphor to make sense of it all.
The launch of the AI lab that would redefine Microsoft caught the tech giant by surprise.
“Did we get called to participate?” Satya Nadella wrote to his team on Dec. 12, 2015, hours after OpenAI announced its founding. “AWS seems to have sneaked in there.”
Nadella had been Microsoft CEO for less than two years. Azure, the company’s cloud platform, was five years old and chasing Amazon Web Services for market share. And now AWS had been listed as a donor in the “Introducing OpenAI” post. Microsoft wasn’t in the mix.
In the internal message, which hasn’t been previously reported, Nadella wondered how the new AI nonprofit could remain truly “open” if it was tied only to Amazon’s cloud.
Within months, Microsoft was courting OpenAI. Within four years, it would invest $1 billion, adding more than $12 billion in subsequent rounds. Within a decade, the relationship would culminate in a $250 billion spending commitment for Microsoft’s cloud and a 27% equity stake in one of the most valuable startups in history.
New court filings offer an inside look at one of the most consequential relationships in tech. Previously undisclosed emails, messages, slide decks, reports, and deposition transcripts reveal how Microsoft pursued, rebuffed and backed OpenAI at various moments over the past decade, ultimately shaping the course of the lab that launched the generative AI era.
More broadly, they show how Nadella and Microsoft’s senior leadership team rally in a crisis, maneuver against rivals such as Google and Amazon, and talk about deals in private.
For this story, GeekWire dug through more than 200 documents, many of them made public Friday in Elon Musk’s ongoing suit accusing OpenAI and its CEO Sam Altman of abandoning the nonprofit mission. Microsoft is also a defendant. Musk, who was an OpenAI co-founder, is seeking up to $134 billion in damages. A jury trial is scheduled for this spring.
OpenAI has disputed Musk’s account of the company’s origins. In a blog post last week, the company said Musk agreed in 2017 that a for-profit structure was necessary, and that negotiations ended only when OpenAI refused to give him full control.
The recently disclosed records show that Microsoft’s own leadership anticipated the possibility of such a dispute. In March 2018, after learning of OpenAI’s plans to launch a commercial arm, Microsoft CTO Kevin Scott sent Nadella and others an email offering his thoughts.
“I wonder if the big OpenAI donors are aware of these plans?” Scott wrote. “Ideologically, I can’t imagine that they funded an open effort to concentrate ML [machine learning] talent so that they could then go build a closed, for profit thing on its back.”
The latest round of documents, filed as exhibits in Musk’s lawsuit, represents a partial record selected to support his claims in the case. Microsoft declined to comment.
Microsoft’s relationship with OpenAI has been one of its key strategic advantages in the cloud. But the behind-the-scenes emails make it clear that Amazon was actually there first.
According to an internal Microsoft slide deck from August 2016, included in recent filings, OpenAI was running its research on AWS as part of a deal that gave it $50 million in computing for $10 million in committed funds. The contract was up for renewal in September 2016.
Microsoft wanted in. Nadella reached out to Altman, looking for a way to work together.
In late August, the filings show, Altman emailed Musk about a new deal with Microsoft: “I have negotiated a $50 million compute donation from them over the next 3 years!” he wrote. “Do you have any reason not to like them, or care about us switching over from Amazon?”
Musk, co-chair of OpenAI at the time, gave his blessing to the Microsoft deal in his unique way, starting with a swipe at Amazon founder Jeff Bezos: “I think Jeff is a bit of a tool and Satya is not, so I slightly prefer Microsoft, but I hate their marketing dept,” Musk wrote.
He asked Altman what happened to Amazon.
Altman responded, “Amazon started really dicking us around on the T+C [terms and conditions], especially on marketing commits. … And their offering wasn’t that good technically anyway.”
Microsoft and OpenAI announced their partnership in November 2016 with a blog post highlighting their plans to “democratize artificial intelligence,” and noting that OpenAI would use Azure as its primary cloud platform going forward.
Internally, Microsoft saw multiple benefits. The August 2016 slide deck, titled “OpenAI on Azure Big Compute,” described it as a prime opportunity to flip a high-profile customer to Azure.
The presentation also emphasized bigger goals: “thought leadership” in AI, a “halo effect” for Azure’s GPU launch, and the chance to recruit a “net-new audience” of developers and startups. It noted that OpenAI was a nonprofit “unconstrained by a need to generate financial return” — an organization whose research could burnish Microsoft’s reputation in AI.
But as the ambition grew, so did the bill.
In June 2017, Musk spoke with Nadella directly to pitch a major expansion. OpenAI wanted to train AI systems to beat the best human players at competitive esports, Valve’s Dota 2. The computing requirements were massive: 10,000 servers equipped with the latest Nvidia GPUs.
“This would obviously be a major opportunity for Microsoft to promote Azure relative to other cloud systems,” Musk wrote in an email to OpenAI colleagues after the call.
Nadella said he’d talk about it internally with his Microsoft cloud team, according to the email. “Sounds like there is a good chance they will do it,” Musk wrote.
Two months later, Altman followed up with a formal pitch. “I think it will be the most impressive thing yet in the history of AI,” he wrote to Nadella that August.
Microsoft’s cloud executives ran the numbers and balked. In an August 2017 email thread, Microsoft executive Jason Zander told Nadella the deal would cost so much it “frankly makes it a non-starter.” The numbers are redacted from the public version of the email.
“I do believe the pop from someone like Sam and Elon will help build momentum for Azure,” Zander wrote. “The scale is also a good forcing function for the fleet and we can drive scale into the supply chain. But I won’t take a complete bath to do it.”
Ultimately, Microsoft passed. OpenAI contracted with Google for the Dota 2 project instead.
Microsoft’s broader relationship with OpenAI was starting to fray, as well. By January 2018, according to internal emails, Microsoft executive Brett Tanzer had told Altman that he was having a hard time finding internal sponsors at Microsoft for an expanded OpenAI deal.
Altman started shopping for alternatives. Around that time, Tanzer noted in an email to Nadella and other senior executives that OpenAI’s people “have been up in the area recently across the lake” — a reference to Amazon’s Seattle headquarters.
The internal debate at Microsoft was blunt.
Scott wrote that OpenAI was treating Microsoft “like a bucket of undifferentiated GPUs, which isn’t interesting for us at all.” Harry Shum, who led Microsoft’s AI research, said he’d visited OpenAI a year earlier and “was not able to see any immediate breakthrough in AGI.”
Eric Horvitz, Microsoft’s chief scientist, chimed in to say he had tried a different approach. After a Skype call with OpenAI co-founder Greg Brockman, he pitched the idea of a collaboration focused on “extending human intellect with AI — versus beating humans.”
The conversation was friendly, Horvitz wrote, but he didn’t sense much interest. He suspected OpenAI’s Dota work was “motivated by a need to show how AI can crush humans, as part of Elon Musk’s interest in demonstrating why we should all be concerned about the power of AI.”
Scott summed up the risk of walking away: OpenAI might “storm off to Amazon in a huff and shit-talk us and Azure on the way out.”
“They are building credibility in the AI community very fast,” the Microsoft CTO and Silicon Valley veteran wrote. “All things equal, I’d love to have them be a Microsoft and Azure net promoter. Not sure that alone is worth what they’re asking.”
But by the following year, Microsoft had found a reason to double down.
In 2019, OpenAI restructured. The nonprofit would remain, but a new “capped profit” entity would sit beneath it — a hybrid that could raise capital from investors while limiting their returns.
Microsoft agreed to invest $1 billion, with an option for a second billion, in exchange for exclusive cloud computing rights and a commercial license to OpenAI’s technology.
The companies announced the deal in July 2019 with a joint press release. “The creation of AGI will be the most important technological development in human history, with the potential to shape the trajectory of humanity,” Altman said. Nadella echoed that sentiment, emphasizing the companies’ ambition to “democratize AI” while keeping safety at the center.
So what changed for Microsoft between 2018 and 2019?
In a June 2019 email to Nadella and Bill Gates, previously disclosed in the Google antitrust case, Scott cited the search giant’s AI progress as one reason for Microsoft to invest in OpenAI. He “got very, very worried,” he explained, when he “dug in to try to understand where all of the capability gaps were between Google and us for model training.”
Nadella forwarded Scott’s email to Amy Hood, Microsoft’s CFO. “Very good email that explains why I want us to do this,” Nadella wrote, referring to the larger OpenAI investment, “and also why we will then ensure our infra folks execute.”
Gates wasn’t so sure. According to Nadella’s deposition testimony, the Microsoft co-founder was clear in “wanting us to just do our own” — arguing that the company should focus on building AI capabilities in-house rather than placing such a large bet on OpenAI.
Nadella explained that the decision to invest was eventually driven by him and Scott, who concluded that OpenAI’s specific research direction into transformers and large language models (the GPT class) was more promising than other approaches at the time.
Hood, meanwhile, offered some blunt commentary on OpenAI’s cap on profits — the centerpiece of its new structure, meant to limit investor returns and preserve the nonprofit’s mission. The caps were so high, she wrote, that they were almost meaningless.
“Given the cap is actually larger than 90% of public companies, I am not sure it is terribly constraining nor terribly altruistic but that is Sam’s call on his cap,” Hood wrote in a July 14, 2019, email to Nadella, Scott, and other executives.
If OpenAI succeeded, she noted, the real money for Microsoft would come from Azure revenue — far exceeding any capped return on the investment itself.
But the deal gave Microsoft more than cloud revenue.
According to an internal OpenAI memo dated June 2019, Microsoft’s investment came with approval rights over “Major Decisions” — including changes to the company’s structure, distributions to partners, and any merger or dissolution.
Microsoft’s $1 billion made it the dominant investor. Under the partnership agreement, major decisions required approval from a majority of limited partners based on how much they had contributed. At 85% of the total, Microsoft had an effective veto, a position of power that would give the company a pivotal role in defining the future of the company.
In September 2020, Musk responded to reports that Microsoft had exclusively licensed OpenAI’s GPT-3. “This does seem like the opposite of open,” he tweeted. “OpenAI is essentially captured by Microsoft.”
Nadella seemed to take the criticism seriously.
In an October 2020 meeting, according to internal notes cited in a recent court order, Microsoft executives discussed the perception that the company was “effectively owning” OpenAI, with Nadella saying they needed to give thought to Musk’s perspective.
In February 2021, as Microsoft and OpenAI negotiated a new investment, Altman emailed Microsoft’s team: “We want to do everything we can to make you all commercially successful and are happy to move significantly from the term sheet.”
His preference, Altman told the Microsoft execs, was “to make you all a bunch of money as quickly as we can and for you to be enthusiastic about making this additional investment soon.”
They closed the deal in March 2021, for up to $2 billion. This was not disclosed publicly until January 2023, when Microsoft revealed it as part of a larger investment announcement.
By 2022, the pressure to commercialize was explicit.
According to a transcript of her deposition, Mira Murati, then OpenAI’s vice president of applied AI and partnerships, had written in contemporaneous notes that the most-cited goal inside the company that year was a $100 million revenue target. Altman had told employees that Nadella and Scott said this needed to be hit to justify the next investment, as much as $10 billion.
Murati testified that Altman told her “it was important to achieve this goal to receive Microsoft’s continued investments.” OpenAI responded by expanding its go-to-market team and building out its enterprise business.
Then everything changed.
On Nov. 30, 2022, OpenAI announced ChatGPT. The chatbot became the fastest-growing consumer application in history, reaching 100 million users within two months. It was the moment that turned OpenAI from an AI research lab into a household name.
Microsoft’s bet was suddenly looking very different.
OpenAI’s board learned about the launch on Twitter. According to deposition testimony, board members Helen Toner and Tasha McCauley received no advance notice and discovered ChatGPT by seeing screenshots on social media.
McCauley described the fact that a “major release” could happen without the board knowing as “extremely concerning.” Toner testified that she wasn’t surprised — she was “used to the board not being very informed” — but believed it demonstrated that the company’s processes for decisions with “material impact on the mission were inadequate.”
Altman, according to one filing, characterized the release as a “research preview” using existing technology. He said the board “had been talking for months” about building a chat product, but acknowledged that he probably did not send the board an email about the specific release.
As its biggest investor, Microsoft pushed OpenAI to monetize the product’s success.
In mid-January 2023, Nadella texted Altman asking when they planned to activate a paid subscription.
Altman said they were “hoping to be ready by end of jan, but we can be flexible beyond that. the only real reason for rushing it is we are just so out of capacity and delivering a bad user experience.”
He asked Nadella for his input: “any preference on when we do it?”
“Overall getting this in place sooner is best,” the Microsoft CEO responded, in part.
Two weeks later, Nadella checked in again: “Btw …how many subs have you guys added to chatGPT?”
Altman’s answer revealed what they were dealing with. OpenAI had 6 million daily active users — their capacity limit — and had turned away 50 million people who tried to sign up. “Had to delay charging due to legal issues,” he wrote, “but it should go out this coming week.”
ChatGPT Plus launched on Feb. 1, 2023, at $20 a month.
A week earlier, Microsoft made its landmark $10 billion investment in OpenAI. The companies had begun negotiating the previous summer, when OpenAI was still building ChatGPT. The product’s viral success validated Microsoft’s bet and foreshadowed a new era of demand for its cloud platform.
Ten months later, it nearly collapsed.
On Friday afternoon, Nov. 17, 2023, OpenAI’s nonprofit board fired Altman as CEO, issuing a terse statement that he had not been “consistently candid in his communications with the board.” Greg Brockman, the company’s president and cofounder, was removed from the board the same day. He quit hours later.
Microsoft, OpenAI’s largest investor, was not consulted. Murati, then OpenAI’s chief technology officer and the board’s choice for interim CEO, called Nadella and Kevin Scott to warn them just 10 to 15 minutes before Altman himself was told.
“Mira sounded like she had been run over by a truck as she tells me,” Scott wrote in an email to colleagues that weekend.
The board — Ilya Sutskever, Tasha McCauley, Helen Toner, and Adam D’Angelo — had informed Murati the night before. They had given her less than 24 hours to prepare.
At noon Pacific time, the board delivered the news to Altman. The blog post went live immediately. An all-hands meeting followed at 2 p.m. By Friday night, Brockman had resigned. So had Jakub Pachocki, OpenAI’s head of research, along with a handful of other researchers.
A “whole horde” of employees, Scott wrote, had reached out to Altman and Brockman “expressing loyalty to them, and saying they will resign.”
Microsoft didn’t have a seat on the board. But text messages between Nadella and Altman, revealed in the latest filings, show just how influential it was in the ultimate outcome.
At 7:42 a.m. Pacific on Saturday, Nov. 18, Nadella texted Altman asking if he was free to talk. Altman replied that he was on a board call.
“Good,” Nadella wrote. “Call when done. I have one idea.”
That evening, at 8:25 p.m., Nadella followed up with a detailed message from Brad Smith, Microsoft’s president and top lawyer. In a matter of hours, the trillion-dollar corporation had turned on a dime, establishing a new subsidiary from scratch — legal work done, papers ready to file as soon as the Washington Secretary of State opened Monday morning.
They called it Microsoft RAI Inc., using the acronym for Responsible Artificial Intelligence.
“We can then capitalize the subsidiary and take all the other steps needed to operationalize this and support Sam in whatever way is needed,” Smith wrote. Microsoft was “ready to go if that’s the direction we need to head.”
Altman’s reply: “kk.”
The company calculated the cost of absorbing the OpenAI team at roughly $25 billion, Nadella later confirmed in a deposition — enough to match the compensation and unvested equity of employees who had been promised stakes in a company that now seemed on the verge of collapse.
By Sunday, Emmett Shear, the Twitch co-founder, had replaced Murati as interim CEO. That night, when the board still hadn’t reinstated Altman, Nadella announced publicly that Microsoft was prepared to hire the OpenAI CEO and key members of his team.
“In a world of bad choices,” Nadella said in his deposition, the move “was definitely not my preferred thing.” But it was preferable to the alternative, he added. “The worst outcome would have been all these people leave and they go to our competition.”
On Tuesday, Nov. 21, the outcome was still uncertain. Altman messaged Nadella and Scott that morning, “can we talk soon? have a positive update, ish.” Later, he said the situation looked “reasonably positive” for a five-member board. Shear was talking to the remaining directors.
Nadella asked about the composition, according to the newly public transcript of the message thread, which redacts the names of people who ultimately weren’t chosen.
“Is this Larry Summers and [redacted] and you three? Is that still the plan?”
Summers was confirmed, Altman replied. The other slots were “still up in air.”
Altman asked, “would [redacted] be ok with you?”
“No,” Nadella wrote.
Scott was more emphatic, giving one unnamed person a “strong no,” and following up for emphasis: “Strong strong no.”
The vetting continued, as Nadella and Scott offered suggestions, all of them redacted in the public version of the thread.
Nadella added Smith to the thread. One candidate, the Microsoft president wrote, was “Solid, thoughtful, calm.” Another was “Incredibly smart, firm, practical, while also a good listener.”
At one point, Scott floated a joke: “I can quit for six months and do it.” He added a grinning emoji and commented, “Ready to be downvoted by Satya on this one, and not really serious.”
Nadella gave that a thumbs down.
The back-and-forth reflected a delicate position. Microsoft had no board seat at OpenAI. Nadella had said publicly that the company didn’t want one. But the texts showed something closer to a shadow veto — a real-time screening of the people who would oversee the nonprofit’s mission.
By evening, a framework emerged. Altman proposed Bret Taylor, Larry Summers, and Adam D’Angelo as the board, with himself restored as CEO. Taylor would handle the investigation into his firing.
Smith raised a concern. “Your future would be decided by Larry [Summers],” he wrote. “He’s smart but so mercurial.” He called it “too risky.” (Summers resigned from the OpenAI board in November 2025, following revelations about his correspondence with Jeffrey Epstein.)
Altman wrote, “id accept it given my conversations with him and where we are right now.” He added, “it’s bullshit but i want to save this … can you guys live with it?”
Nadella asked for Summers’ cell number.
At 2:38 p.m., Altman texted the group: “thank you guys for the partnership and trust. excited to get this all sorted to a long-term configuration you can really depend on.”
Nadella loved the message.
Two minutes later, Smith replied: “Thank you! A tough several days. Let’s build on this and regain momentum.”
Altman loved that one.
Nadella had the last word: “Really looking forward to getting back to building….”
Later that night, OpenAI announced Altman’s return with the newly constituted board.
“We are encouraged by the changes to the OpenAI board,” Nadella posted on X. “We believe this is a first essential step on a path to more stable, well-informed, and effective governance.”
The crisis was resolved, but the underlying tensions remained.
On December 27, 2024, OpenAI announced it would unwind its capped-profit structure. Internally, this initiative was called “Project Watershed,” the documents reveal.
The mechanics played out through 2025. On September 11, Microsoft and OpenAI executed a memorandum of understanding with a 45-day timeline to finalize terms.
Microsoft’s role was straightforward but powerful. Its approval rights over “Major Decisions” including changes to OpenAI’s structure. Asked in a deposition whether those rights covered a recapitalization of OpenAI’s for‑profit entity into a public benefit corporation, Microsoft corporate development executive Michael Wetter testified that they did.
The company had no board seat. “Zero voting rights,” Wetter testified. “We have no role, to be super clear.” But under the 2019 agreement, the conversion couldn’t happen without them.
The timing mattered. A SoftBank-led financing — internally called Project Sakura — was contingent on the recapitalization closing by year-end. Without the conversion, the funding could not proceed. Without Microsoft’s approval, the conversion could not proceed.
Valuation became a key focus of negotiations. Morgan Stanley, working for Microsoft, estimated OpenAI’s value at $122 billion to $177 billion, according to court filings. Goldman Sachs, advising OpenAI, put it at $353 billion. The MOU set Microsoft’s stake at 32.5 percent. By the time the deal closed after the SoftBank round, dilution brought it to 27 percent.
OpenAI’s implied valuation was $500 billion — a record at the time (until it was surpassed in December by Musk’s SpaceX). As Altman put it in his deposition, “That was the willing buyer-willing seller market price, so I won’t argue with it.”
For Microsoft, it was a give-and-take deal: the tech giant lost its right of first refusal on new cloud workloads, even as OpenAI committed to the $250 billion in future Azure purchases.
At the same time, the agreement defused the clause that had loomed over the partnership: under prior terms, a declaration of artificial general intelligence by OpenAI’s board would have cut Microsoft off from future models. Now any such declaration needs to be made by an independent panel, and Microsoft’s IP rights run through 2032 regardless.
The transaction closed on Oct. 28, 2025. The nonprofit remained (renamed the OpenAI Foundation) but as a minority shareholder in the company it had once controlled.
Six days later, OpenAI signed a seven-year, $38 billion infrastructure deal with Amazon Web Services. The company that had “sneaked in there” at the founding, as Nadella put it in 2015, was back — this time as a major cloud provider for Microsoft’s flagship AI partner.
In a post this weekend, OpenAI CFO Sarah Friar made the shift explicit: “Three years ago, we relied on a single compute provider,” she wrote. “Today, we are working with providers across a diversified ecosystem. That shift gives us resilience and, critically, compute certainty.”
Revenue is up from $2 billion in 2023 to more than $20 billion in 2025. OpenAI is no longer a research lab dependent on Microsoft’s cloud. It’s a platform company with leverage.
In December 2015, Nadella had to ask whether Microsoft had been called to participate in the OpenAI launch. A decade later, nothing could happen without the Redmond tech giant.
But OpenAI will no longer be theirs alone.
The Microsoft Library in Redmond has long been a quiet anachronism in the middle of the high-tech campus, a place where authors gave talks and employees checked out old-fashioned paper books, including titles recommended by CEO Satya Nadella and other execs.
That chapter of the company’s history is now closing.
The Verge broke the news Thursday that Microsoft’s traditional library is going away as part of what Microsoft described internally as a shift to a “modern, AI-powered learning experience.”
Responding to an inquiry from GeekWire, the company confirmed that its libraries in Redmond, Hyderabad, Beijing, and Dublin closed as of this week and “are being repurposed into collaborative spaces for group learning and experimentation,” where employees can explore emerging technologies.
“We’re evolving Microsoft Library locations and services to better support how employees learn, stay current, and build new skills,” a Microsoft spokesperson said via email. The changes are already underway and will roll out fully in the coming weeks, according to the company.
In an internal FAQ obtained by The Verge, Microsoft described the new approach as a “Skilling Hub” and acknowledged that the decision “affects a space many people valued.”
The shift also includes cuts to employee subscriptions for newspapers and industry reports. Publications affected include The Information and Strategic News Service, which had provided global reports to Microsoft employees for more than two decades.
Microsoft said it continues to offer access to more than 20 digital resources and subscriptions, “prioritizing those most valuable to employees.”
Strategic News Service didn’t mince words about Microsoft’s AI-focused rationale.
“Technology’s future is shaped by flows of power, money, innovation, and people — none of which are predictable based on LLMs’ probabilistic regurgitation of old information,” Berit Anderson, the company’s chief operating officer, told The Verge.
The library has moved around over the decades, from the original Building 4 to Building 92 most recently. The news of the closure drew a nostalgic response on X from Steven Sinofsky, the former Windows president, who called the library “a crown jewel of the early days.”
“They bought every PC book and two copies of every software,” Sinofsky wrote. “If you found one you needed that they didn’t have, they acquired it.”
[Editor’s Note: Agents of Transformation is an independent GeekWire series and March 24, 2026 event, underwritten by Accenture, exploring the people, companies, and ideas behind AI agents.]
It was “like bringing a Porsche into a world of Model Ts.”
That’s what Microsoft said in its 1990 annual report about the shift from MS-DOS to Windows. But the bigger breakthrough for the company wasn’t the graphical interface. It was Windows’ 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 Microsoft’s specs, and your software would become a first-class citizen on the computers that were arriving “on every desk and in every home,” as the company’s original mission statement put it.
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 what’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.
“We 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,” Microsoft CEO Satya Nadella wrote in a blog post this week looking ahead to 2026. “This is the engineering sophistication we must continue to build to get value out of AI in the real world.”
Elements of this are already emerging elsewhere.
But Microsoft’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.
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 parallels to the past only go so far. Traditional PC applications ran in their own windows, worked with their own files, and didn’t touch the rest of the system for the most part.
“Agents are going to need to be able to scratchpad their work,” 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 user’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 agent’s instructions — 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 “understand the security implications of enabling an agent on your computer” before turning them on.
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 isn’t the center of the computing world anymore. Smartphones, browsers, and cloud platforms have fragmented the landscape in ways that didn’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 — an attempt to make Windows the home for agents that serve individual users on their own machines. That’s a harder sell when the PC is competing with phones, browsers, and cloud apps for people’s attention. Microsoft can build the platform, but it can’t guarantee that developers will show up the way they did 35 years ago.
And unlike in the 1990s, Microsoft can’t count on users to embrace what it’s building. There’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.
In October, for example, Microsoft announced new features including “Hey Copilot” voice activation, a redesigned taskbar with Copilot built in, and the expansion of “Copilot Actions” agentic capabilities beyond the browser to the PC itself.
“They’re thinking about revenue first and foremost,” 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 Microsoft’s most recent fiscal year, Windows and Devices generated $17.3 billion in revenue — essentially flat for the past three years.
That’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, Microsoft’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 company’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 doesn’t matter as much as keeping it out of the junkyard.
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, it’s not without precedent.
As we sat down to reflect on the past year, we rewound all the way back to January — 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.
“Now what’s happening is intelligence is becoming free,” he said, “and that’s even more profound than computing becoming free.”
As we looked through GeekWire’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.
“We look forward to joining Matt on his private island next year.” — Kiana Ehsani, CEO of Vercept, after her co-founder Matt Deitke left to join Meta for a reported hundreds of millions of dollars.
Proud Seattleite and grammarian Ken Jennings on Jeopardy!, correcting a contestant: “Sorry, Dan, we are sticklers in Seattle. It’s Pike Place — no s.”
Ambika Singh, CEO and founder of Armoire, accepting the Workplace of the Year award at the GeekWire Awards: “It 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.”
Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.
Audio editing by Curt Milton
Microsoft is pouring $17.5 billion into India — its largest investment in Asia — to boost the country’s AI infrastructure and diffusion, the company announced Tuesday.
The funding, planned over four years between 2026 and 2029, comes after an earlier $3 billion commitment announced earlier this year.
Microsoft’s aim is to help advance India’s cloud and AI infrastructure, skilling and ongoing operations. The tech giant said the partnership will help India make the leap from “digital public infrastructure to AI public infrastructure in the coming decade,” toward “a future that is more equitable and uniquely Indian in its scale and impact.”
Microsoft CEO Satya Nadella is in the country this month as part of a multi-city “India AI” tour. He met with Prime Minister Narendra Modi in New Delhi on Tuesday and will deliver a keynote address on Wednesday: “Leading in the New Age of AI.”
The investment will target three primary areas:
Microsoft also announced that 310 million informal workers in India will benefit from advanced AI capabilities being integrated into two key digital public platforms of the Ministry of Labour and Employment — e-Shram and the National Career Service.
Microsoft employs 22,000 people across Bengaluru, Hyderabad, Pune, Gurugram, Noida and other cities, representing numerous company business lines.
Elsewhere on Tuesday, Microsoft President Brad Smith announced new commitments to Canada, adding $5.4 billion over the next two years to its continued investment in building out digital and AI infrastructure in the country.
Smith highlighted many of the same goals the company outlined for India, including boosting skills training and digital sovereignty in Canada.
December 7 carries historical weight well beyond the tech world, but for those who covered Microsoft in the ’90s, the date has another resonance. Thirty years ago today, Bill Gates gathered more than 200 journalists and analysts at Seattle Center to declare that the company was going “all-in” on the internet.
As managing editor for Microsoft Magazine at the time, I was there, and I remember it well. Three decades later, I can’t help but see the parallels to Microsoft’s current AI push.
The moves that Microsoft kicked off that day to build internet connectivity into all its products would reverberate throughout the next decade, helping to lay the foundation for the dot-com boom years and arguably the eventual rise of cloud computing.
The release of Internet Explorer 2.0 as a free, bundled browser, the internet-enablement of Microsoft Office, the complete revamping of the still-new MSN online service, Microsoft’s licensing of Java from Sun Microsystems and a focus on how the internet might be used commercially were all pieces of the Microsoft plan unveiled that day.
“The internet is the primary driver of all new work we are doing throughout the product line,” Bill Gates told the assembled technology press in 1995. “We are hard core about the internet.”
Substitute the word “AI” for “internet” and you have a statement that current Microsoft CEO Satya Nadella could have made at any moment in the last couple of years.
“Fifty years after our founding, Microsoft is once again at the heart of a generational moment in technology as we find ourselves in the midst of the AI platform shift,” Nadella wrote in his 2025 annual letter to shareholders. “More than any transformation before it, this generation of AI is radically changing every layer of the tech stack, and we are changing with it.”
Whether you are using the Microsoft Azure cloud platform; running a Windows 11 PC, tablet, or laptop; spending time on LinkedIn; or using Microsoft 365, you will find AI baked in.
Comparing then and now, there are insights in both the similarities and the differences, and lessons from Microsoft’s mid-’90s missteps and successes that are still relevant today.
The challenge of navigating the shift to a new generation of technology in a large, fast-moving company is the biggest similarity between now and 30 years ago.
Microsoft was a lot smaller in 1995, but it was still the dominant force in the software industry of its day. When the company launched Windows 95 in August of 1995, it came with the first versions of both Internet Explorer and MSN. Within four months, it had to ship new, better versions of those products alongside a whole lot of other changes.
The push for speedy change grew out of something the company had been telling its senior leaders for several months prior to the launch of Windows 95: It had to move fast and do more if it was going to catch up in a race that it couldn’t afford to lose.
Gates’ famous “internet tidal wave” memo from May 26, 1995 (which later became an antitrust exhibit) spelled out both the threat and opportunity — calling the internet “the most important single development to come along since the IBM PC was introduced in 1981.”
Later in the memo, Gates acknowledged a significant problem: Microsoft would have to explain why publishers and internet users should use MSN instead of just setting up their own website — and he admitted that the company didn’t have a great answer.
Fast forward to March 2023, a few months after Microsoft partner OpenAI launched ChatGPT, when Satya Nadella made the scale of the AI era clear in a speech on the future of work.
“Today is the start of the next step in this journey, with powerful foundation models and capable copilots accessible via the most universal interface: natural language,” Nadella said. “This will radically transform how computers help us think, plan, and act.”
Of course, Microsoft CEOs have learned a lot over the last 30 years, including the importance of not pointing out the company’s shortcomings in memos that could end up being seen by the rest of the world. Nadella offered nothing like Gates’ MSN admission. But his comments about the size of the AI challenge and opportunity were a direct parallel to the urgency that Gates expressed about the internet 30 years ago.
In the world of PC operating systems and software, Microsoft in the 1990s was king — with few competitors that came even close to the kind of market share it enjoyed. It was arguably late in making a bet-the-company pivot to the internet, but doing so from a very strong position.
Thirty years later, amid the rise of artificial intelligence, Amazon, Google, Nvidia, OpenAI, and Anthropic are part of a more complex network of competitors and partners.
Back in 1995, the big competition was perceived as coming from Netscape and other fast-moving internet startups — and Microsoft was the behemoth battling the insurgents.
The New York Times’ headline about the 1995 event summed up the framing: “Microsoft Seeks Internet Market; Netscape Slides.” As The Seattle Times put it, “Microsoft plays hardball — Game plan for the Internet: Crush the competition.” Many others echoed the theme.
I saw that competitive dynamic first-hand at the press event, when by a stroke of luck I ended up sitting beside Bill Gates at lunch. I recall him being a little annoyed by questions about the Java licensing deal with Sun and the broader press interest in the Netscape/Microsoft narrative. He wanted to focus on the broader impact of the day’s announcements.
He stressed, for example, that the licensing by Microsoft of Sun’s Java programming language for use with Microsoft’s Internet Explorer browser was not really a big deal.
“Java you can recreate trivially,” Gates told me, brushing off the licensing deal as a routine business decision, not much different than many others Microsoft made over the years.
The scale is also drastically different. For example, my January 1996 cover story for Microsoft Magazine quoted Gates explaining how the “150 million users of Windows” would benefit from the internet integration it was undertaking across 20 new products and technologies.
In today’s terms, those numbers look tiny. In a blog post earlier this year, Microsoft executive vice president Yusuf Mehdi said Windows now powers more than 1.4 billion monthly active devices. That doesn’t include Microsoft’s massive cloud computing business, Microsoft 365, LinkedIn, Xbox, and its already-significant AI-attributable revenue from Copilot.
The investment gap is more dramatic, even adjusted for inflation. Microsoft poured more than $88 billion into capital expenditures last fiscal year, much of it on AI infrastructure. In 1995, the company’s $220 million deal with NBC to launch MSNBC sounded like a lot of money.
That MSNBC deal, however, highlights another important contrast between the present and the past. In 1995, no one really knew where the internet (and the web) was going to go. Fortunes were made and lost trying to predict which business models would work online.
Tim Bajarin, CEO of the consultancy Creative Strategies and a longtime industry analyst, says Microsoft is better positioned now than it was in 1995. The difference: we already have the underlying architecture for useful AI applications. That wasn’t true with the internet back then.
“We didn’t see the value proposition until we saw the role of applications built on a web-based architecture,” Bajarin said. “That is what is significantly different.”
Microsoft’s AI push, Bajarin said, will succeed only if it delivers genuine value — implementations that solve real problems and show clear return on investment.
Recent headlines suggest not everyone is convinced. ‘No one asked for this’: Microsoft’s Copilot AI push sparks social media backlash, declared Germany’s PC-WELT magazine. It’s the same question Gates couldn’t answer about MSN in 1995: Why should anyone use this?
Perhaps the biggest lesson on the competition front is that there is no guarantee of longevity or relevance in tech. Only one of the competitors listed in the December 1995 New York Times story is still around – IBM – and it is a vastly different company than it was then.
There is one more lesson, about the cost of success. Microsoft’s aggressive internet push worked — but it also triggered a Department of Justice investigation that lasted from 1998 to 2001. Competing hard is essential. Competing too hard has consequences.
But that’s a story for another decade.
Microsoft’s annual shareholder meeting Friday played out as if on a split screen: executives describing a future where AI cures diseases and secures networks, and shareholder proposals warning of algorithmic bias, political censorship, and complicity in geopolitical conflict.
One shareholder, William Flaig, founder and CEO of Ridgeline Research, quoted two authorities on the topic — George Orwell’s 1984 and Microsoft’s Copilot AI chatbot — in requesting a report on the risks of AI censorship of religious and political speech.
Flaig invoked Orwell’s dystopian vision of surveillance and thought control, citing the Ministry of Truth that “rewrites history and floods society with propaganda.” He then turned to Copilot, which responded to his query about an AI-driven future by noting that “the risk lies not in AI itself, but in how it’s deployed.”
In a Q&A session during the virtual meeting, Microsoft CEO Satya Nadella said the company is “putting the person and the human at the center” of its AI development, with technology that users “can delegate to, they can steer, they can control.”
Nadella said Microsoft has moved beyond abstract principles to “everyday engineering practice,” with safeguards for fairness, transparency, security, and privacy.
Brad Smith, Microsoft’s vice chair and president, said broader societal decisions, like what age kids should use AI in schools, won’t be made by tech companies. He cited ongoing debates about smartphones in schools nearly 20 years after the iPhone.
“I think quite rightly, people have learned from that experience,” Smith said, drawing a parallel to the rise of AI. “Let’s have these conversations now.”
Microsoft’s board recommended that shareholders vote against all six outside proposals, which covered issues including AI censorship, data privacy, human rights, and climate. Final vote tallies have yet to be released as of publication time, but Microsoft said shareholders turned down all six, based on early voting.
While the shareholder proposals focused on AI risks, much of the executive commentary focused on the long-term business opportunity.
Nadella described building a “planet-scale cloud and AI factory” and said Microsoft is taking a “full stack approach,” from infrastructure to AI agents to applications, to capitalize on what he called “a generational moment in technology.”
Microsoft CFO Amy Hood highlighted record results for fiscal year 2025 — more than $281 billion in revenue and $128 billion in operating income — and pointed to roughly $400 billion in committed contracts as validation of the company’s AI investments.
Hood also addressed pre-submitted shareholder questions about the company’s AI spending, pushing back on concerns about a potential bubble.
“This is demand-driven spending,” she said, noting that margins are stronger at this stage of the AI transition than at a comparable point in Microsoft’s cloud buildout. “Every time we think we’re getting close to meeting demand, demand increases again.”
Login