thats why high interactivity is so important.
and @tenstorrent makes it economically feasible even on very large models.
and all this agentic reality is also why we bet long on our own CPU efforts here.
Introducing Infinite Studio ♾.
Last week, @tenstorrent x @prodia announced the fastest Wan 2.2 video generation in the world.
We built a demo to show what that speed unlocks: directing an infinite movie in real time.
Demo 👇
Google has TPU
Amazon got inferentia and trainium
Microsoft wishes Maia worked
Same with Meta and MTIA
Tesla has AIHW. 🪦 Dojo
...
Every adequate company wants to be in charge of own destiny. You can’t do everything at once but given enough time..
~2012 everyone buys CPU from Intel. By now every hyperscaler has their custom arm based CPU that’s used a ton. See AWS reports on graviton usage.
Every car manufacturer wants to be like Tesla now.
Every lab wants to be a bit more like Google.
OpenAI works on a chip with Broadcom..
...
Google secured its own destiny with TPUs. And it will keep doing things to further increase adoption and increase pool of talent familiar with it. That’s going to both drive revenue and reinforce their own critical supply chain.
Can't say the same about Meta, OpenAI, or even Anthropic (if I am to consider it an independent lab)
...
I clearly see why NVidia is so dominant today. But it might not be like that forever. NVidia is a fine company, way deeper than many realize, but there are forces against it.
As time goes companies want to own more of their supply chain, especially the core technology.
Lots of challenges - hardware takes time, then remember about software, things get very expensive and require scarce engineering talent.
...
This is what we address at @tenstorrent .
Silicon you can actually own.