Microsoft has updated a page on its Windows Health Dashboard to say that if you're on Windows 10 version 20H2, it's soon going to automatically move you to version 21H2. The news comes as no surprise, since this is typical behavior when a version of the OS is nearing the end of support.

Specifically, Windows 10 version 20H2 - only including Home and Pro editions - is no longer supported after May 10, 2022. That gives Microsoft four months to move everyone to something newer. Remember, Windows 10 feature updates don't work like they did before. The only time you're forced to install a feature update is when your current version is nearing the end of support.

The company said the following:

Considering this servicing timeline, we started a rollout for machine learning (ML) training targeting devices on Windows 10, version 20H2 that are approaching end of servicing to update automatically to Windows 10, version 21H2.

It's worth noting that this is a harmless update. That's a bold claim given the vast array of Windows 10 machines on the planet, but there's a reason that we can say that. Every feature update that was offered after Windows 10 version 2004 is the exact same bits.

Starting with Windows 10 version 20H2, Microsoft started taking an approach where feature updates were offered with enablement packages. The new features, if there are any, are delivered in a monthly cumulative update, but they're hidden. The enablement package just lights them up and increases the build number by one. That's why Windows 10 version 2004, 20H2, 21H1, and 21H2 were all getting the exact same cumulative updates, unless Microsoft arbitrarily ended support for one of them like it has for version 2004.

And of course, version 20H2 will come to the end of support in May for Home and Pro SKUs. Since this is an 'H2' update, Enterprise and Education SKUs have another year of support if they really want to stick to 20H2.

For now, Microsoft is describing the automatic rollout of Windows 10 version 21H2 as being in the first phase of machine learning training, meaning that it's basically going to automatically upgrade some machines, see how it goes, and continue the process based on that. As mentioned above, there shouldn't be any issues blocking this from rapidly becoming a widespread rollout.