3 ms·

googles multiple businesses and gemini isn't the largest one. anthropic is the anchor external customer of tpu's and nvidia is worth more than all of google. I

googles multiple businesses and gemini isn't the largest one. anthropic is the anchor external customer of tpu's and nvidia is worth more than all of google. If tpu's actually breakout as a viable alternative over the next few years for multiple clients the business could easily be worth as much as search, maybe more.

billisonline 6mo ago
> If tpu's actually breakout as a viable alternative over the next few years Why haven't they broken out yet, I wonder, if they're more efficient for inference and LLM costs are now weighted towards inference over training?
- chris_st 6mo ago
  Possibly because they just haven't been able to manufacture enough of them yet to be a viable business to others? They're fighting everyone else for foundry space and time.
- zaphar 6mo ago
  You essentially have to run in google to use them and that probably limits their ability to breakout. Anthropic might be doing this deal as a way to shore up their supply chain and cost of both inference and training by leveraging Google's hardware and chip manufacturing expertise.
  - ai-x 6mo ago
    Several customers like Citadel, run TPUs in their own datacenters (closer to Exchanges)
  - lanthissa 6mo ago
    every tpu thats been made is in use and sold at a high margin, demand is not the issue.
- lanthissa 6mo ago
  there are literally not enough tpu's on earth for them to break out, every tpu thats been made is in use, the spike in demand is recent and google has heavy competition for foundry space.
- AnggaSP 6mo ago
  TPUs are not that portable and easy for both inferencing and training. It has since improved a lot with their effort on the torch backend (XLA/TorchTPU) and JAX though. But as far as i know it currently supports just that + tensorflow (which nobody uses it anymore, least here). And last we tried, so much of our kernels needs rework that it’s not worth the effort. This may change since ironwood but we haven’t tried that generation.
nikcub 6mo ago
Google cloud also need to be able to offer Anthropic models on Vertex otherwise they just won't be competitive. Microsoft is in the same boat with Azure.
- shimman 6mo ago
  Google Cloud also needs to show constant quarterly growth so what better way than simply buying it and fudging the numbers?

billisonline 6mo ago

> If tpu's actually breakout as a viable alternative over the next few years Why haven't they broken out yet, I wonder, if they're more efficient for inference and LLM costs are now weighted towards inference over training?

chris_st 6mo ago
Possibly because they just haven't been able to manufacture enough of them yet to be a viable business to others? They're fighting everyone else for foundry space and time.
zaphar 6mo ago
You essentially have to run in google to use them and that probably limits their ability to breakout. Anthropic might be doing this deal as a way to shore up their supply chain and cost of both inference and training by leveraging Google's hardware and chip manufacturing expertise.
- ai-x 6mo ago
  Several customers like Citadel, run TPUs in their own datacenters (closer to Exchanges)
- lanthissa 6mo ago
  every tpu thats been made is in use and sold at a high margin, demand is not the issue.
lanthissa 6mo ago
there are literally not enough tpu's on earth for them to break out, every tpu thats been made is in use, the spike in demand is recent and google has heavy competition for foundry space.
AnggaSP 6mo ago
TPUs are not that portable and easy for both inferencing and training. It has since improved a lot with their effort on the torch backend (XLA/TorchTPU) and JAX though. But as far as i know it currently supports just that + tensorflow (which nobody uses it anymore, least here). And last we tried, so much of our kernels needs rework that it’s not worth the effort. This may change since ironwood but we haven’t tried that generation.

nikcub 6mo ago

Google cloud also need to be able to offer Anthropic models on Vertex otherwise they just won't be competitive. Microsoft is in the same boat with Azure.

shimman 6mo ago
Google Cloud also needs to show constant quarterly growth so what better way than simply buying it and fudging the numbers?