The companies that make RAM and flash memory chips are enjoying record profits because of the AI-induced memory crunch—and they’re also indicating that they don’t expect conditions to improve much if at all in 2026. And while RAM kits have been hit the fastest and hardest by shortages and price increases, we shouldn't expect SSD pricing to improve any time soon, either.
That's the message from Shunsuke Nakato (via PC Gamer), managing director of the memory division of Kioxia, the Japanese memory company that was spun off from Toshiba at the end of the 2010s. Nakato says that Kioxia’s manufacturing capacity is sold out through the rest of 2026, driving the market for both enterprise and consumer SSDs to a “high-end and expensive phase.”
“There is a sense of crisis that companies will be eliminated the moment they stop investing in AI, so they have no choice but to continue investing,” said Nakato, as reported by the Korean-language publication Digital Daily. Absent a big change in the demand for generative AI data centers, that cycle of investments will keep prices high for the foreseeable future.
Big Tech's AI-fueled memory shortage is set to be the PC industry's defining story for 2026 and beyond. Standalone, direct-to-consumer RAM kits were some of the first products to feel the bite, with prices spiking by 300 or 400 percent by the end of 2025; prices for SSDs had also increased noticeably, albeit more modestly.
The rest of 2026 is going to be all about where, how, and to what extent those price spikes flow downstream into computers, phones, and other components that use RAM and NAND chips—areas where the existing supply of products and longer-term supply contracts negotiated by big companies have helped keep prices from surging too noticeably so far.
This week, we're seeing signs that the RAM crunch is starting to affect the GPU market—Asus made some waves when it inadvertently announced that it was discontinuing its GeForce RTX 5070 Ti.
We’ve been waiting a long time to see how DirectStorage performs in the real world. Forspoken is the first game to support it and it was released this week after a multi-month delay. Now it’s in gamers’ hands and we finally have some numbers to pore over, thanks to some benchmarks a hardware testing site in Germany has posted. They’re not for loading times but for overall performance. As it turns out, offloading asset compression from the CPU to the GPU does impact gaming performance. Your mileage may vary, of course, but in the first tests, it’s up to an 11% penalty in frames per second.
The tests were performed by PC Games Hardware. To test DirectStorage 1.1, it set up a test bench with a Core i9-12900K and an RTX 4090. On the SSD side, they tested three models: SATA, and PCIe 3.0 and 4.0. Oddly, the testers didn’t say which models of SSDs they used for testing. Regardless, DirectStorage doesn’t work with SATA, so we’re able to glean the effects of the asset decompression happening on the GPU instead of the CPU. The tests were run in 4K and showed some clear results.
In an unexpected twist, the SATA SSD offered the highest fps, coming in at 83.2 on average. When switching to the fastest PCIe 4.0 SSD, the average frame rate was 11% slower at 74.4fps. The PCIe 3.0 drive was just as fast as PCIe 4.0, averaging a single fps more on average. Since they only tested at 4K, we don’t know if this situation is the same at lower resolutions. The good news for gamers is the 1% and 0.2% fps averages were essentially the same across all three drives. This would indicate that the player would not notice any performance spikes while playing.
Previously, it was reported that DirectStorage can lead to a huge increase in data transfer speeds. In that test, it was Intel’s GPU that was the fastest, beating out pricier GPUs from AMD and Nvidia. Clearly, more testing is needed across the GPU spectrum. We’d also be curious to see what a PCIe 5.0 SSD could do with Forspoken. Sadly, those drives are not quite ready yet. Also, keep in mind this is just one data point. Another YouTuber named Bang4BuckPC Gamer also has a SATA vs. PCIe 4.0 side-by-side, and in the majority of the scenes, the performance is the same. Sometimes, though, the NVME drive is noticeably faster than the SATA drive.
At this point, we need to see more SSDs and GPUs tested to see what the performance penalty is (if any). Though 11% is a higher number than expected, the game’s frame rate was still well above 60fps and it looks very smooth in the video. We also don’t think the RTX 4090 is the best GPU to test this on, as someone with that card never really has to worry about fps in any game, even at 4K. We’d be curious to see what the impact is on Windows 10 as well, as it has a watered-down version of DirectStorage.
As the famous poet Carl Johnson once said, “Ah sh*t, here we go again.” Apple’s newest MacBook Pros with M2 SoCs were just released and are now being subjected to teardowns. This has allowed us to peek under the hood at the SSD configurations. Sadly, it appears Apple is doubling down on hobbling the SSD speeds on the base models, 9to5mac reports. It did the exact same thing last year. Therefore, it’s not a huge surprise, but still a disappointment on a laptop with pricing that starts at $1,999.
What Apple has done this time is a replay of the M2 launch almost a year ago. When it announced the M2 MacBook Air and MacBook Pro it was discovered the base model used one NAND chip instead of the two in their M1-based predecessors. For example, on the M1 laptops, 256GB of storage was divided between two 128GB NAND modules. However, as time goes on, density goes up. So Apple switched to a single 256GB chip for the base models. The spot where the second chip used to be was empty and only filled if the customer upgraded its storage. Thanks to how SSDs rely on parallelism to boost performance, this meant the base model’s SSDs offered just half the speeds of machines with two NAND chips.
Benchmarks show slightly reduced write speeds but a pretty significant reduction in read speeds. (Image: 9to5mac)
Now we see it happening again on the new M2 machines as well. Despite the fact that the new M2 MacBook Pros offer 512GB of storage for the base models, Apple can now cover this amount with half the chips. Teardowns confirm the M1 MacBook Pro used four NAND modules and the M2 machines use just two. Benchmarks confirm a significant impact on performance, though 9to5mac’s speeds (above) are actually showing less of an impact than other reviewers are reporting.
For example, Max Tech compared the 16-inch M1 and M2 Pro laptops with 512GB of memory. The benchmarks show the M1 system is twice as fast in read speeds compared with the M2. In CrystalDiskMark, the M1 hit 6887MB/s, compared with 3462MB/s on the M2. For a 22GB file transfer test, the M2 system was 56 seconds slower than its M1 predecessor.
An Adobe Lightroom export of 499 photos with web browsing occurring in the background. (Credit: MaxTech)
In a multitasking and file export test in Adobe Lightroom, the M2 Pro showed noticeably slower performance than the M1 system as well. He exported 499 photos from Lightroom while clicking between 15 browser tabs. The M2 system took 70 seconds longer to export the photos. It also stuttered briefly during this process when switching tabs. The M1 never hitched at all in this multitasking comparison. Although, to be fair to Apple, the hitching was very brief. Also, in some CPU-based tasks, the improvement in CPU power of the M2 can make up for the SSD speeds, depending on the task.
As MaxTech states, when this happened with the M2 Air and 13-inch MacBook Pro, it was mostly excused as those were inexpensive machines. The M2 laptops cost over $2,000, though, with the 16-inch model starting at $2,500. How much this will impact people in the real world is hard to quantify. But we can sure see it in the benchmarks, both synthetic and real-world. At the very least, it would be handy if Apple informed people of the situation before they bought one.
Login
