Especially if you take the view that AI is the new compute. There so many chips that need to get upgraded Thats maybe one of the largest TAMs of all time that hasn't even been touched at this point.
Given how fucked it’s competition is and how the delta between CUDA/NCCL and everything else like rocm/zulda has only grown yeah, Nvidia will own the whole thing for minimum 10 years.
Everyone who tried to compete failed hard because no one has the money, and raw talent or ability to get that talent needed to beat Nvidia at the software game.
Can you provide an alternative prediction that you believe is more accurate?
> This game is far from over
Do you think a company like AMD or Intel is going to make massive gains in taking market share away from Nvidia? Or do you think it will be another company? Or something else entirely?
They always say "no, it'll be GOOGLE!" and try to bring up TPUsv6 (you still can't even play with ironwood/TPUv7 yet) as though they're competitive with blackwell let alone Reuben, as though anyone except google internal engineers get real value from them, and as though GCP doesn't have well over 1 million+ haswell/blackwell GPUs.
These same people unironically believe that Gemini was trained with zero GPUs for any part of the process (including all experiments), and it was all done on TPUs. They cite this as evidence that CUDA/Nvidia is dead.
It's stupid, it's wrong, but it's what they claim.
I'm going to be a lot wealthier than the Nvidia bears.
And if you don't believe this, just take one look at TPU pricing (remember, Trillium is TPUv6) and tell me this is competitive with Nvidia/Oracle/any neocloud with a straight face.
I'd argue the Nvidia moat is the most difficult moat to understand unless one is really in the weeds using GPUs. "all you need to do is design a similar chip" seems to fit in peoples heads nicely when they are even a little ways removed from the details.
People have been talking about competing and trying to compete with the nvidia GPU/Cuda stack for almost 20 years (since the start). There have been various efforts. They have all fallen flat. Nvidia isn't standing still, the target keeps moving forward.
To suggest Nvidia will have the game to themselves for another 10 years might turn out to be wrong, but it isn't naive. You are the naive one here.
There hasn't really been any significant money in competing with cuda however. Nvidia had a bit of a gaming premium over amd, and crypto really boosted all GPUs...but until about 3 years ago, there wasn't literally trillions of dollars on the line to replace cuda. There is now. Companies ARE replicating it. The Mi300 is very competitive on token throughput as far as I'm aware.
No one is sitting around. I'd argue if there was more wafer supply you'd see amd/others undercutting nvidia...but it's hard to when supply is incredibly constrained.
They are trying. They are not succeeding yet. Maybe in 10 years the gap will be closed. Maybe it will not. I'll guess the latter. Nvidia's situation has changed too - the R&D $$ they have to spend to defend are dramatically higher. Nothing stands still, it's harder to catch up than it seems.
