Meta Expands AI Infrastructure Bet with $21 Billion CoreWeave Agreement

Meta is significantly scaling its AI compute capacity, announcing a new $21 billion spending commitment with cloud infrastructure provider CoreWeave. This new agreement complements a previous $14.2 billion deal, bringing the total commitment to over $35 billion as the social media giant accelerates its push into advanced artificial intelligence. The new contract will be deployed between 2027 and 2032, while the prior arrangement remains in effect through 2031. CoreWeave, which specializes in Nvidia-powered GPU clusters, provides the critical high-performance compute necessary for training and deploying large-scale models like Meta's recently unveiled Muse Spark. The deal highlights Meta's aggressive capital expenditure strategy. The company expects to spend between $115 billion and $135 billion on capex this year, nearly doubling its 2025 spending. While Meta is building its own data centers—including a $10 billion facility in Texas—it continues to rely on third-party hyperscalers to meet what it describes as insatiable demand. For CoreWeave, the partnership is a strategic victory in revenue diversification. Previously, Microsoft accounted for 62% of the company's 2024 revenue; following this deal, no single customer will represent more than 35% of total sales. The company, which went public last year, continues to leverage heavy debt to fund expansion, borrowing an additional $8.5 billion in March to support new contracts.