CoreWeave expands Meta deal

CoreWeave has added roughly $21 billion in backlog from Meta to lock in cloud capacity for AI, and is reportedly planning a $4.25 billion debt raise to support that growth. The move illustrates continued compute scarcity behavior among hyperscalers and the lengths buyers will go to secure long‑term capacity (thetechcapital.com, techstartups.com).

Meta just agreed to spend another $21 billion with CoreWeave for artificial intelligence cloud capacity through December 2032, taking the two companies’ total contracts to about $35.2 billion. CoreWeave said the expansion also pushed its overall revenue backlog to $66.8 billion. (coreweave.com, cnbc.com) CoreWeave is not a consumer app company. It is a landlord for artificial intelligence computing, renting out giant clusters of Nvidia graphics chips inside data centers to companies that need more computing power than they can build fast enough themselves. (cnbc.com, investors.coreweave.com) Meta is buying this capacity for inference workloads, which is the stage where a trained model answers questions, generates images, or serves recommendations for real users. Training builds the brain first, but inference is the part that has to run all day for millions or billions of requests. (investors.coreweave.com) That helps explain why this deal runs from 2027 to 2032 instead of being a short burst order. Meta is trying to lock in future access now, before the next wave of chips, power hookups, and data center space gets even harder to secure. (coreweave.com, thenextweb.com) The new contract includes some of the first deployments of Nvidia’s Vera Rubin platform across multiple locations. Nvidia describes Vera Rubin NVL72 as a rack-scale system that combines 72 Rubin graphics chips with 36 Vera central processors in one artificial intelligence supercomputer. (coreweave.com, nvidia.com) Meta can afford to sign early because it is already planning one of the biggest infrastructure spending years in corporate history. In its February 2026 results, Meta said 2026 capital spending would land between $115 billion and $135 billion, after spending $72.22 billion in 2025. (investor.atmeta.com) CoreWeave, meanwhile, has to build the capacity before it can rent it out. The company told investors in February that it expects $12 billion to $13 billion in 2026 revenue, but an even larger $30 billion to $35 billion in capital spending to stand up the hardware and facilities behind signed customer contracts. (cnbc.com, finance.yahoo.com) That is why the debt number matters. Reports on April 9 said CoreWeave was lining up a $4.25 billion debt raise after also disclosing a separate $3 billion convertible debt offering, showing how expensive it is to turn backlog into live computing capacity. (fool.com, cnbc.com) This is not CoreWeave’s first debt-heavy expansion tied to a customer promise. On March 31, Bloomberg reported that CoreWeave had raised $8.5 billion in a chip-backed loan secured by graphics processors and customer contracts, with Meta agreements backing at least part of that financing. (bloomberg.com) Put together, the picture is simple: Meta does not want to wait for its own construction timetable, and CoreWeave is willing to borrow heavily to become the overflow factory. When buyers sign seven-year checks for future chips, power, and buildings, it usually means the shortage is not in the model anymore. It is in the physical infrastructure underneath it. (coreweave.com, investor.atmeta.com, bloomberg.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.