This is another example of Jensen's pro-Neocloud strategy.

Enterprise intelligence largely runs through OpenAI, Anthropic and Google. They control placement and are moving into custom silicon: OpenAI has Jalapeño, Google has TPUs and Anthropic works with AWS on Trainium.

NVIDIA's counter is to commoditise models. Open weights, data and recipes let enterprises deploy their own intelligence, moving persistent inference onto neutral GPU infrastructure.

Bullish for neoclouds and$AGPU. If you missed the memory trade, dont miss this.nitter.net/NVIDIAAI/status/208716…