RE: LeoThread 2025-08-17 12:27

People are still talking about training compute for the AI world. That is a drop in the bucket compared to the inference compute that will be required. Jensen Huang, CEO of NVIDIA, mentioned this earlier this year. He said that inference compute will need to scale a billion-fold over the next decade.

With embedded AI, a company needs to understand not only robotics (or whatever the embodied agent is) but manufacturing and acquire the inference.

There are few companies that can pull all this expertise together.