Meta Invests $21B More in CoreWeave for AI Cloud

Meta has significantly expanded its partnership with CoreWeave, committing an additional $21 billion for dedicated AI cloud capacity. This new agreement, spanning from 2027 through 2032, elevates the total value of their infrastructure relationship to approximately $35 billion. The contract is specifically designed for AI inference workloads and will facilitate early deployments of Nvidia’s Vera Rubin platform across multiple sites. Concurrently, CoreWeave announced plans to raise $4.25 billion in new debt to fund its expansion. The market responded positively, with CoreWeave shares rising around 5% and Meta shares gaining roughly 3%.

The company’s journey to this point is a remarkable pivot. Founded in 2017 as a cryptocurrency mining operation called Atlantic Crypto, CoreWeave’s founders recognized that their accumulated GPU inventory was precisely what the emerging machine learning sector desperately needed. They rebranded in 2019, shifting focus to GPU cloud infrastructure. After going public in March 2025, CoreWeave secured a foundational $14.2 billion agreement with Meta later that year, establishing its credibility. This latest $21 billion expansion solidifies Meta as CoreWeave’s most significant commercial relationship, providing a revenue base that will sustain the company through the end of the decade.

This contract is strategically focused on inference, the process of running already-trained AI models, rather than the initial training phase. Meta’s open-weight Llama models mean the expensive training is largely complete; the ongoing challenge is serving these models to billions of users across its apps with sustained, low-latency compute. CoreWeave’s distributed infrastructure, including early access to Nvidia’s next-generation Vera Rubin hardware, provides this critical capacity. The deal supplements Meta’s own massive capital expenditure on AI infrastructure, which is guided to reach up to $135 billion in 2026. This follows a similar $27 billion deal with Nebius in March, illustrating Meta’s strategy of building a diversified multi-vendor infrastructure for flexibility and redundancy at hyperscale.

For CoreWeave, the expanded Meta relationship directly addresses a key post-IPO concern: excessive revenue concentration. Prior to this deal, Microsoft represented a dominant share of its revenue, a risk that made investors uneasy. CEO Michael Intrator stated that with the new Meta commitment, no single customer will account for more than 35% of total sales, materially improving the company’s risk profile. CoreWeave’s business remains deeply intertwined with Nvidia GPUs, a dependency extended into the next generation with the Vera Rubin deployments. The company has also broadened its base with a recent expansion of its agreement with OpenAI, further diluting its reliance on any one partner.

Funding this rapid growth requires substantial capital. Alongside the Meta announcement, CoreWeave detailed plans to raise $4.25 billion through a combination of convertible notes and high-yield senior unsecured notes, effectively junk-bond pricing for a portion of the debt. This brings CoreWeave’s total debt load to around $30 billion, triple its level from a year ago. The company justifies this structure by pointing to its more than $66 billion in contracted backlog, which it says provides the revenue visibility to service its obligations. This mirrors a broader trend in AI infrastructure financing, where the scale of investment now demands novel and large-scale debt instruments, as seen in other major projects like SoftBank’s funding for OpenAI. From a closet of Ethereum mining rigs, CoreWeave has become a fundamental pillar of global AI infrastructure, built one multi-billion dollar commitment at a time.