$21 billion. That's how much Meta is committing to CoreWeave in one of the largest AI infrastructure contracts on record.
CoreWeave is a cloud provider built entirely around NVIDIA GPU clusters. Unlike Amazon Web Services or Google Cloud, which offer dozens of products, CoreWeave does exactly one thing: rent high-performance computing capacity for AI workloads. The company went public in early 2025 and has been signing long-term anchor contracts with major AI players ever since. A reported $11.9 billion deal with Microsoft was the prior headline. The Meta deal roughly doubles it.
What Meta Is Buying
Meta already owns more AI hardware than most governments. The company has spent tens of billions building its own data centers and acquiring NVIDIA H100 and H200 chips directly. CEO Mark Zuckerberg has committed publicly to spending $60-65 billion on AI infrastructure in 2025 alone. So why pay CoreWeave on top of that?
The answer is timing. Building a data center from scratch takes years. CoreWeave has capacity available now - or at least, committed capacity on a faster timeline than Meta can build internally. Locking in external compute is a hedge against delays and a way to handle demand spikes without leaving model training or product features waiting on construction crews.
The Open/Closed Contradiction
There's a tension worth naming here. Meta has spent the past two years positioning itself as the open-source alternative to OpenAI and Google - releasing Llama models, publishing research, framing closed AI as bad for the industry. That narrative holds up at the model layer. At the infrastructure layer, Meta is as dependent on proprietary hardware and cloud contracts as anyone else.
For CoreWeave, the deal is a business win that makes the company's post-IPO story substantially more credible. These long-term contracts are what justify the capital expenditure to keep expanding. Two anchor deals of this size - Microsoft and Meta - make CoreWeave a serious infrastructure player rather than a bet on continued GPU scarcity.