Even as a mere AI consumer I'm seeing the compute constraints begin to bite as I scale up my usage - and I don't really use AI much for my day-to-day work! That's coding to a large extent but if that's what I'm experiencing then I presume (and hear) that those using these tools professionally are already hitting token/compute ceilings.
We're also starting to see the large providers like OpenAI and Anthropic choking on the demand. Presumably at some point the GPU chokepoint will begin to bite and limit what frontier models they can deliver.
Even as a mere AI consumer I'm seeing the compute constraints begin to bite as I scale up my usage - and I don't really use AI much for my day-to-day work! That's coding to a large extent but if that's what I'm experiencing then I presume (and hear) that those using these tools professionally are already hitting token/compute ceilings.
We're also starting to see the large providers like OpenAI and Anthropic choking on the demand. Presumably at some point the GPU chokepoint will begin to bite and limit what frontier models they can deliver.