CoreWeave has signed a $14 billion agreement with Meta to supply computing power, the latest in a string of multi-billion-dollar infrastructure deals underscoring how artificial intelligence spending is shifting from software to the hardware backbone that supports it.
The contract, announced Tuesday, commits Meta to pay about $14.2 billion through December 14, 2031, with an option to extend into 2032 for additional capacity, CoreWeave said in a filing. Shares of the cloud computing company jumped 15% following the disclosure, pushing its valuation—which stood at $60 billion as of the last close—further into the ranks of high-growth AI service providers.
Under the agreement, Meta will secure access to Nvidia’s latest GB300 systems housed in CoreWeave’s data centers, according to CEO Michael Intrator in a Bloomberg News interview.
Register for Tekedia Mini-MBA edition 18 (Sep 15 – Dec 6, 2025): registration continues.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab: From Technical Design to Deployment.
For Meta, the deal represents more than just raw computing power. It ties directly into its consumer-facing ambitions, including its recently launched Ray-Ban smart glasses, while complementing its vast spending on U.S. data centers and top AI engineering hires, where salaries now rival those of professional athletes.
For CoreWeave, Meta adds another pillar customer beyond Microsoft, its largest client. Just last week, the server provider inked its third multi-billion-dollar expansion deal with OpenAI, strengthening its position as the preferred partner for companies racing to scale AI.
Some analysts believe that the wave of long-term contracts marks a new phase in the AI boom. “Nvidia’s chips are in CoreWeave’s data centers, which will then be used by companies like Meta. These kinds of deals spark bubble concerns because of how insular the industry appears, and the massive dollar amounts involved,” said Emarketer’s Jacob Bourne.
He cautioned that valuations have been inflated by “circular” financing, where AI firms both consume and invest in each other’s services.
At the same time, the expansion of AI demand beyond the so-called Magnificent Seven—Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta, and Tesla—suggests the risk of a sudden bubble burst may be tempered. Bourne noted that startups, cloud providers, and enterprises outside Big Tech are beginning to sign similar agreements, ensuring broader market participation.
The Meta-CoreWeave deal is also emblematic of a global trade divide in AI infrastructure. In the U.S., private-sector partnerships dominate, with hyperscalers like Meta, Microsoft, and OpenAI locking in long-term cloud deals to secure scarce Nvidia chips. In contrast, China has leaned heavily on state-backed investment through firms such as Huawei, Baidu, and Alibaba, which are scaling their own data centers to meet domestic AI demand while racing to reduce reliance on U.S.-controlled chipmakers.
This shift—away from flashy AI applications toward the invisible but capital-intensive infrastructure—has reshaped investment priorities. CoreWeave’s 8x manufacturing capacity increase in the past 18 months, with plans to expand another 4x in the next six to eight months, illustrates just how rapidly the market is evolving. After Taiwan Semiconductor Manufacturing (TSMC) produces its chip wafers, CoreWeave packages them in the U.S., an added signal of the strategic value of onshore production.
Meta’s deal, alongside CoreWeave’s tie-ups with Microsoft and OpenAI, confirms that AI’s next frontier lies in securing compute power before shortages bite harder. It’s a defensive move to maintain leadership in the global AI race for U.S. firms, underlining how access to chips and infrastructure has become the new battleground in trade and technology competition.
