No, they can't when you can get tens of TFLOPS per GPU for <$1000 that comes with a solid software ecosystem for all the major AI frameworks out of the box. That's the power of the gaming/industrial complex: NVDA can fund the development of AI chips and software from the nearly bottomless pockets of millions of gamers and cryptocurrency miners. ASIC startups figuratively have one bullet in their gun and they have to hit the bullseye to live.
Now when a Tesla GPU costs $9-10K instead of <$1000, that's a somewhat different story, but even then, NVDA tapes out a brand new architecture almost annually. Good luck keeping up with that pace ASIC startups. And that's exactly what happened to Nervana. Their ASIC was better than a Pascal GP100, but it's clobbered by Volta V100. So at best you get a 6-12 month lead on them.
In contrast, if you can right-size the transistor count for expensive and very specific workloads across 100K+ machines like companies with big datacenters and big data sets can do, I see an opportunity for all of the BigCos to build custom ASICs for their siloed data sets and graphs. That's what GOOG is doing and it seems to be working for them so far. FB is now hiring to do the same thing I suspect.
Yes but over the lifetime of a GPU, you'll spend more on power draw than the physical hardware. That's where the savings come from, or at least that's what I've been told.
A V100 costs ~300/yr in electricity. If you are buying at the scale of 100k units, but can price per operation, even by just 10% (for example, by dropping features you don't care about), that's a million dollars of electricity over the lifetime of your hardware.
Now when a Tesla GPU costs $9-10K instead of <$1000, that's a somewhat different story, but even then, NVDA tapes out a brand new architecture almost annually. Good luck keeping up with that pace ASIC startups. And that's exactly what happened to Nervana. Their ASIC was better than a Pascal GP100, but it's clobbered by Volta V100. So at best you get a 6-12 month lead on them.
In contrast, if you can right-size the transistor count for expensive and very specific workloads across 100K+ machines like companies with big datacenters and big data sets can do, I see an opportunity for all of the BigCos to build custom ASICs for their siloed data sets and graphs. That's what GOOG is doing and it seems to be working for them so far. FB is now hiring to do the same thing I suspect.