Abhinav pivoted from a career in banking to pursue his first love in writing. Even while working full-time, he continued contributing as an editor-at-large, a role he has held for more than 7 years. A lifelong tech enthusiast who has built three gaming and productivity powerhouse PCs since 2018, his passion for technology keeps him closely following the semiconductor industry, from NVIDIA and AMD to ARM. His MSc dissertation explored how artificial intelligence will reshape the future of work, reflecting his curiosity about the wider social impact of emerging technologies.
Windows 11 version 26H2 is here, with some big and many small changes to Microsoft's desktop client operating systems. If all you've read about it comes from Microsoft's release notes, chances are, you probably shrugged and went on with your day. To be fair, that's a valid response to most of the things you'd see on the list. After all, not many would rush to hit the update button for a movable taskbar, voice typing to rename a file, or new File Explorer quirks.
However, there's one thing that's generating some interest in the Windows subreddits and user forums. Since installing 26H2, people are watching their PCs idle with a couple of gigabytes less in RAM use, which has sparked conversations around perhaps a lighter, more optimized OS. There does, however, appear to be a catch, and it's what most headlines have skipped over.
Windows 11 26H2 is reducing RAM usage across many PCs
The newly updated OS is finally judicious with system memory
Perhaps one of the most controversial effects of 26H2 that isn't in the Microsoft documentation became a subject of discourse a few days after the initial rollout. A Reddit user with 48GB of RAM reported that their on-boot memory usage fell from 6.9GB to roughly 4.8GB after the update, and the comments underneath filled up with corroborating numbers from other people installing the update on their own machines. Most other users reported a drop anywhere between 1GB and 3GB.
Windows Latest also ran their own test on a 16GB laptop and reported similar findings, stating that the laptop typically crossed 10GB of idle usage before the update, whereas the usage fell down to 7GB after the fact.
Regarding what's doing the work here, the answer doesn't seem to be all that clear at the moment, especially because Microsoft hasn't claimed any of these optimizations. However, the company has spent most of 2026 talking about reducing the operating system's memory footprint.
Pavan Davuluri specifically acknowledged as much on an episode of Microsoft's Inside Windows podcast back in May that his team has "clear efforts around driving better utilization of memory", along with work on compressing memory and making the broader platform more efficient. This change, in effect, does seem to be in alignment with Microsoft's priority of reducing memory usage in the wake of the memory crisis.
The 8GB machines are still waiting their turn
Every reported drop has come from a 16GB machine or bigger
Among the talks surrounding memory optimization on the latest build, a subset of users have found themselves reporting no gains since installing 26H2 on their machines, and it's the set that would find themselves most heavily impacted by the change. Almost all reported drops in memory utilization have come from systems running 16GB of RAM or more, with those using 32-48GB of RAM reporting the biggest wins.
Those with 8GB machines, in the meanwhile, are either staying put or getting lost in the noise. Amongst several Reddit threads and user forums, that seems to be the case based on user reporting. So far, all major publications reporting on the subject have noticed a similar pattern amongst those running 8GB of RAM.
Although the reports are largely based on first-person accounts from other users, the move does seem to reflect Microsoft's objective of optimizing memory for systems with "8GB and above", which Davuluri named as one of four Windows priorities through the end of 2026 in a July post. If indeed, the memory optimizations under the hood of 26H2 are real, there is a chance that they are skipping over the 8GB tier that Microsoft committed to lifting, leaving those users right where they were before 26H2 came into play.
I tested 26H2 on my own laptop from 2022
Is it really that judicious with memory?
Before installing 26H2, I benched my Lenovo Legion 5 Pro on 25H2 to get a baseline reading following a clean restart and a five-minute wait for the system to settle. To measure idle memory usage, I launched the Windows Task Manager and recorded the memory usage in the Memory tab. My idle usage was at 8.5GB out of the available 31.9GB, with 23.3GB available and 13.5GB in the standby cache. The machine had my usual startup apps loaded, without any other memory-heavy workloads running in the foreground.
Upon installing 26H2, I was able to confirm what many others on similar systems had previously reported with the new build. Idle usage had dropped to 6.7GB, with 25.1GB available and 11.7GB cached. That's the exact 1.8GB drop in the in-use memory on the same machine with no other variables changed.
Interestingly enough, I noticed that the cached memory also fell by exactly as much as in-use memory did, which seems to strongly suggest that Windows is holding less in standby rather than reshuffling the memory. When it comes to systems with an already generous memory capacity, all signs, so far, seem to point at the fact that Windows 11 26H2 really is that judicious with memory.
A gift for the abundant, but nothing for the rest?
I can't help but notice the impossible irony to all of this. Microsoft itself has spent the greater part of 2026 publicly acknowledging the memory crisis, launched 8GB Surface laptops, and yet, the 26H2 update, which happens to be the one that shows real promise in memory optimization, completely misses that one tier that would experience the relief the most. Maybe there are further optimizations on the horizon for those users, but for now, it feels like a slap in the face of the users looking for some relief in the midst of the DRAM crisis.





