The podcast analyzes NVIDIA's strategy of investing in open-source AI models, questioning whether the stated $26 billion investment is genuine or a clever reframing of existing compute service agreements. The discussion explores the nuances of open-source versus proprietary models, referencing Intel's earlier Linux strategy and the recent departure of the Alibaba open-source AI team. It further considers the emerging role of model orchestration, where AI agents select the best models for specific tasks, and how companies like Perplexity are implementing this. The conversation touches on Meta's silicon program, questioning the feasibility of a six-month chip release cadence and the potential for smaller volume manufacturers like Intel to participate. Finally, the hosts highlight the importance of supply chain mastery in the current constrained environment, noting how capacity limitations can force innovation and new vendor relationships.
Sign in to continue reading, translating and more.
Continue
