[/QUOTE]
Keep in mind that Quadro is a legacy brand and not really much of a thing now, the Tesla A2 is an AIB GPU/graphics card though that is destined for AI desktop/workstations and one of its main customers are
datacenters too. Jon Peddie wouldn't need to worry about including legacy Quadro but their methodology would need to exclude the A2 from those AIB sales whatever that methodology is.
This is a little more realistic to what I thought you were saying. I was under the impression that you said nvidia sell 30-40M alone:
That puts them closer to 24M for all their GPU sales. though I'm not sure where you're getting "gaming" from in that stat. It would be all AIB GPU regardless of their intended purpose, whether that's mining, Autodesk, editing, AI or whatever. The consumer would likely be for gaming but there would be businesses making massive AIB GPU purchases too without ever touching gaming. That's for both Geforce and cards like the Tesla A2.
Yes, it appears Quadro has gone the way of the Tesla range. Tensor appears to have been discontinued as well, for professional virtualization Nvidia now has the NVIDIA RTX Ada Generation. So GeForce for gaming GPU, RTX Ada for the Professional Virtualization segment, and Nvidia A and T series for datacenters. Jon Peddie tracks desktop AIB GPUs and utilize the PCI-E interface, so GeForce and RTX Ada would fall under this bracket, but A and T series GPUs would not, because they are specifically not for desktop, but more crucially none of them have actual video out and cannot be used as regular GPUs, they are for compute workloads only.
So for Nvidia for the last 4 quarters, they shipped 32.2 million GPUs (88% market share), with 29.5 million of those being GeForce, with RTX Ada making up the rest (actually GeForce would certainly be higher as RTX Ada cards cost significantly more, so that 9% revenue slice for Professional certainly be less cards, but lets go with the smaller number for simplicity sake). AMD shipped as well so the total desktop gaming (Radeon/GeForce) brand shipped for the 4 quarters was 33.5 million in total.
As for what these GPUs are used for, Mining on RTX 4000 is hardly worth it, AI needs bandwidth and VRAM (so really only the 4090 is good for that), Businesses overwhelming go for the professional range (RTX Ada), and again the Ada series is just better for modeling. I'm sure not every single GPU goes to gaming, but considering how these cards are actually marketed and that Nvidia has the different market segment offerings its pretty clear that gaming is probably by far the majority of use cases.
Looking over the past 10 years, so end 2014 to now, almost 500 million GeForce/Radeon (so not Quadro, Tesla, Tensor, Radeon Pro, etc) GPUs were shipped. Mining had two crazes that took millions away (but tons of those get resold as gaming GPUs, after each mining dip Nvidia and AMD both had major sales slumps from the second hand market), some users use these for non gaming reasons, and a ton have probably broken over the years. That being said, it is still a colossal number of gaming GPUs. Integrated GPUs are in the billions, but I think we can all agree that almost none of that is really used for gaming.