Meta Platforms has expanded its long-term AI infrastructure agreement with CoreWeave to about $21 billion, signaling how fast demand for AI computing is growing.
The updated deal runs through December 2032. It builds on Meta’s earlier commitment from September 2025, which was valued at roughly $14.2 billion. This larger contract shows that Meta wants more outside computing power as its AI products move into wider public use.
Meta Pushes Harder Into AI Inference
The new agreement focuses heavily on inference workloads. In simple terms, inference is the stage where trained AI models actually respond to users, create content, and support live tools.
That matters because AI demand no longer stops at training large models. Companies now need enough infrastructure to run those models every day for millions of people. Therefore, access to dependable cloud capacity has become a major competitive advantage.
Reports also suggest Meta will use CoreWeave infrastructure in several locations, including systems tied to Nvidia’s upcoming Vera Rubin platform. That points to strong confidence in future AI demand.
For CoreWeave, this is another major win. The company has focused on becoming a specialist AI cloud provider instead of competing as a broad cloud platform.
That strategy seems to be paying off. As demand for GPUs and AI-ready data centers keeps rising, CoreWeave has attracted large, high-profile customers.
Key takeaways from the deal include:
- A contract value of about $21 billion
- A timeline that extends through December 2032
- A stronger focus on AI inference services
- More evidence that large tech firms want outside compute partners
Furthermore, the Meta deal could help CoreWeave diversify its revenue base, which investors have watched closely.
Investor Optimism Meets Financial Pressure
Investors initially welcomed the news, with CoreWeave shares reportedly rising in premarket trading. Meta shares also moved higher.
However, the announcement also comes with fresh attention on CoreWeave’s financial position. The company continues to manage large debt and lease obligations tied to its infrastructure expansion. It is also seeking added financing flexibility as it scales.
The bigger story is clear. The AI race is now about more than building smarter models. It is also about securing enough computing power to keep those models running at industrial scale.
