Source item

Musk says they’re building their own chip for inference, reducing the need for $NVDA GPUs. It’ll be made in Taiwan and the U.S, maybe work with $INTC or they will build a "giantaic chip fab". My take: He’s talked about moving away from $NVDA before and it’s proven extremely hard to do. As for building a fab, I think they have better places to spend $20B.