OpenAI is building a custom AI chip codenamed Jalapeño, a direct move to reduce dependency on Nvidia and control its own compute stack. This mirrors what Google did with TPUs and what Amazon did with Trainium. The strategic logic is simple: inference at ChatGPT's scale means billions spent on third-party silicon, and owning that hardware layer changes the unit economics entirely.
Micron reported strong earnings with direct ties to Anthropic's infrastructure buildout, signaling that the memory bandwidth bottleneck is now a primary battleground in AI hardware. Meta's next flagship model is delayed, and the episode breaks down what that timeline slip actually indicates about training compute constraints. Google and A24 are collaborating on AI-assisted film production, and Anthropic's Claude is getting native Slack integration.
The Jalapeño chip discussion is the reason to watch this one in full. The hosts go beyond the announcement to map out what a vertically integrated OpenAI looks like across model training, inference, and data center strategy. If you care about where AI infrastructure power concentrates over the next five years, this episode is the relevant 22 minutes.
[WATCH ON YOUTUBE →]