Meta is not waiting for its own data centers to catch up with its AI ambitions. It is renting the capacity it needs right now.
CoreWeave and Meta announced an expanded agreement on April 9 in which Meta will pay approximately $21 billion for dedicated AI cloud capacity through December 2032, CoreWeave confirmed this in a press release. The new deal layers on top of a prior $14.2 billion arrangement between the two companies, bringing Meta’s total commitment to CoreWeave to $35 billion.
CoreWeave shares rose about 5% in early afternoon trading on April 9. Meta shares gained roughly 3%.
What the CoreWeave deal covers
The new capacity will be deployed across multiple locations and will include some of the initial deployments of the NVIDIA Vera Rubin platform. Vera Rubin is NVIDIA’s next-generation AI chip architecture following the Blackwell series. Getting early access to this hardware matters. Companies that deploy it first gain a performance advantage before it is widely available.
The agreement covers new capacity running through December 20, 2032, alongside the exercise of an existing option for additional capacity through April 10, 2032, as detailed in CoreWeave’s 8-K filing. The two companies’ Master Services Agreement dates back to December 2023. The new spending runs from 2027 to 2032, as Meta continues building its own infrastructure while leasing flexible capacity from CoreWeave, CNBC reported.
Why inference is driving the deal
A key focus of the agreement is inference, which is when trained AI models deliver real-time results to users rather than learning from data. This is different from training, where models are built. Inference is often where costs and delays are most visible when operating at scale.
More AI Stocks:
- Morgan Stanley sets jaw-dropping Micron price target after event
- Bank of America updates Palantir stock forecast after private meeting
- Morgan Stanley drops eye-popping Broadcom price target
Meta runs Llama-based models across Facebook, Instagram, WhatsApp, and Messenger, which together reach billions of users. Handling that workload demands fast, widely distributed infrastructure built specifically for AI. CoreWeave can deploy GPU clusters faster than Meta can construct new sites, which is a core reason for the relationship.
In March, Meta announced a $10 billion data center in Texas. This year, the company plans to spend between $115 billion and $135 billion on capital expenditures, nearly double what it spent in 2024, with most of that going to data centers. The CoreWeave deal fills the gap while those owned facilities are being built.
Why Meta keeps coming back
CoreWeave CEO Mike Intrator framed the relationship in direct terms.
“They’re going to continue to do it themselves, but they’re also going to continue to do it with us,” he said. “There’s just too much risk not to”.
Intrator said Meta’s return reflects how the company evaluates its options.
“They hired from across the space, people who have used infrastructure from all different folks, and they came back to us,” he added.
The CoreWeave CEO also addressed the deal’s broader significance. “This is another example that leading companies are choosing CoreWeave’s AI cloud to run their most demanding workloads,” Intrator said.
A Meta spokesperson described the arrangement as “part of our portfolio-based approach to infrastructure, as we invest in capacity for our AI ambitions,” according to CNBC.
Ratcliffe/Getty Images
What it means for CoreWeave
The deal significantly reshapes CoreWeave’s revenue concentration. Microsoft represented 62% of CoreWeave’s 2024 revenue. After this deal, no single customer will represent more than 35% of total sales, Intrator said.
CoreWeave also announced it intends to raise $3 billion through convertible senior notes due 2032, with an option for buyers to purchase an additional $450 million. A separate $1.25 billion issuance of senior unsecured notes due 2031 was also announced. Proceeds will be used for general corporate purposes and potentially to refinance existing debt.
Key terms of the Meta-CoreWeave deal:
- New agreement value: approximately $21 billion
- New capacity term: through December 20, 2032
- Prior arrangement: $14.2 billion, option term through April 10, 2032
- Total Meta commitment to CoreWeave: $35 billion
- Infrastructure: multiple locations featuring early NVIDIA Vera Rubin platform deployments
- Master Services Agreement originally signed: December 2023
The broader picture AI
The deal reflects an accelerating pattern across the AI industry. The largest technology companies are securing long-term computing capacity well in advance of when they need it. CoreWeave’s model, which specializes in GPU-accelerated cloud infrastructure for companies including Google, Microsoft, and OpenAI, has attracted a series of major long-term agreements in a short period.
For CoreWeave, which completed its Nasdaq listing in March 2025, the $35 billion in total Meta contracts alongside its existing customer base transforms the revenue picture considerably. The company is no longer primarily dependent on any single customer, and the 2032 term gives it visibility across multiple generations of AI hardware.
Related: Meta weighs drastic workforce decision after $135 billion guide