Ask HN: The State of the Chip and AI/ML Market. Hype vs. Reality
I don't have enough experience in both fields: chipmaking and AI/ML to really understand the landscape. But to me, it seems that the chip situation is two-fold: there is need for high performing chips for model creation. But once you need to actually do inference, there comes less need for high performance chips and more commodity hardware can suffice. The way I see the NVIDIA price jump (as a proxy for the AI/ML market), is there seems to be a focus on model creation as the value generator, and perhaps an investors' lack of understanding of how large the surface area is for inference/model execution.
Could members of the community chime in here with their own analysis and thoughts? It seems that this kind of post could (if gets enough discussion) be a good aggregate of sources and thoughts on the state and future state of chip+AI/ML. Thanks!