Meta Commits Tens of Billions for Nvidia AI Chips and CPUs

Meta is dramatically expanding its reliance on Nvidia, agreeing to buy millions of the chipmaker’s newest processors—ranging from standalone Grace CPUs to next‑generation Blackwell GPUs and forthcoming Vera Rubin systems—for its U.S. data‑center buildout. Axios reports the multi‑year commitment, whose exact dollar value isn’t disclosed but is framed as tens of billions, makes Meta the first Big Tech firm to deploy Nvidia’s CPUs at scale for running (inference) AI workloads, not just training models. The deal locks in scarce next‑gen compute at a time when Nvidia’s Blackwell chips are back‑ordered and hyperscale cloud and social platforms are expected to spend about $650 billion on data centers this year, with Meta alone projected to pour roughly $135 billion into AI. Shares of Meta and Nvidia rose on the news, while rival AMD’s stock fell, underscoring investors’ view that Nvidia remains the backbone of large‑scale AI infrastructure despite efforts by Google, Amazon, Microsoft and Meta to develop in‑house chips. Analysts say the move highlights both how concentrated AI hardware spending has become and how dependent major U.S. platforms still are on Nvidia for the next phase of the AI race.