Just to be pedantic (hey, it's the Internet, what else I'm supposed to do?), the driver was, in a way, already open source and has always been - it interacts with the kernel code and thus must be GPL v2. But previously only like the "glue" that binds the driver module to the kernel was open source, the rest of it was a BLOB (think a firmware that's tens of megabytes in size). So if/when things went wrong with the driver, there was no way to debug it, much to the annoyance of Linus et al.
Nothing is open source on the Nvidia side. The userspace is proprietary and it communicates with a blob that runs alongside the kernel and interacts with it. The entire ecosystem is obscure.
GPLv2 allows this, so much so that one of the reasons for creating GPLv3 was to avoid this kind of thing and for this reason, Linus Torvalds disagreed with Richard Stallman and kept the kernel in GPLv2. Without this, I believe Linux would have less hardware support. It's not just Nvidia's proprietary blob that exists in the kernel, there are several others, such as Broadcom's wireless drivers.
What Nvidia did was migrate most of the blobs to the GPU firmware and created open source kernel modules, although they are not yet present directly in the kernel. Maybe they aren't, but it's easier to manage now. There will still be the proprietary userspace and there you will have access to Nvidia features such as DLSS, ray tracing, CUDA.
Red Hat is creating an open userspace in Mesa 3D called NVK. As far as I've seen, it has better performance than the proprietary one (in one game), but it still doesn't have the same features, as they will depend on their implementation in Vulkan.
www.phoronix.com
Here's an interesting video explaining the Linux graphics stack, if you're interested in finding out more.