Meta and CoreWeave: A $21 Billion Partnership to Dominate AI

Meta Invests Heavily in CoreWeave's AI Infrastructure

Meta has made a significant strategic decision by committing to invest a total of $35 billion in CoreWeave's AI infrastructure. This move highlights the growing importance of computing power in the field of artificial intelligence, surpassing even the models themselves in terms of priority. Major tech companies are seeking to secure access to these computing resources in the long term, and Meta is no exception.

The company recently announced a new allocation of $21 billion to CoreWeave, aimed at ensuring dedicated cloud capabilities for AI until December 2032. This agreement, unveiled on April 9, brings Meta's total commitment to CoreWeave to $35 billion. In light of increasing demand and tensions over GPUs, this decision underscores that infrastructure is now at the heart of global competition in AI.

A Hybrid Strategy for Meta

Meta is adopting a hybrid approach to avoid bottlenecks in its operations. The company combines building its own data centers with outsourcing some of its critical needs. This portfolio strategy allows Meta to balance control, flexibility, and speed of execution.

Internally, Meta is investing in a data center project in Texas, estimated at $10 billion. For this year, it plans to spend between $115 and $135 billion on investments. However, these infrastructures take time to become operational. In the meantime, CoreWeave plays a crucial role as an operational relay.

Rather than waiting for its internal capabilities to be ready, Meta is immediately renting available GPU resources. CoreWeave, specialized in accelerated cloud infrastructure, can deploy clusters much faster than a hyperscaler under construction.

Mike Intrator, CEO of CoreWeave, emphasized that the operational risk is too high to rely on a single approach. "Meta will continue to build, but also work with us," he stated. This redundancy has become a form of assurance in the face of growing demand.

A key aspect of this agreement concerns inference, often less publicized than model training. Inference is the use of models in production, meaning the moment they generate real-time responses for users.

Exclusive Access to New NVIDIA Chips

One of the strategic elements of the partnership between Meta and CoreWeave is the early access to the NVIDIA Vera Rubin platform. These next-generation chips succeed Blackwell and provide an immediate competitive advantage to those who deploy them first.

In the current market, every performance gain can translate into significant savings or a better user experience. This early access allows Meta to optimize its models before these technologies become widely available.

This type of partnership also shows that companies are looking to secure critical hardware capabilities over several years. Instead of simply purchasing cloud services, tech giants are seeking to lock in access to computing resources well ahead of their actual needs.

According to several analyses, including those reported by CNBC and The Street, this strategy aims to avoid a GPU shortage in the face of exponential demand. Computing is becoming a rare commodity, and thus an important differentiation lever.

CoreWeave has positioned itself precisely in this niche, providing specialized infrastructure for players like OpenAI, Google, or Microsoft. This specialization allows it to quickly secure significant long-term contracts.

Transformation of CoreWeave's Business Model

With this agreement with Meta, CoreWeave is diversifying its revenue sources. Historically, the company was heavily dependent on Microsoft, which accounted for 62% of its revenue in 2024. With Meta's arrival, no single client is expected to exceed 35% of revenue going forward. This more balanced distribution reduces business risk and enhances CoreWeave's credibility with the markets.

On the day of the announcement, CoreWeave's stock rose by about 5%, while Meta's gained nearly 3%. Investors see this agreement as a positive signal, both in terms of financial visibility and strategic positioning.

Listed on the Nasdaq in March 2025, CoreWeave now benefits from a long-term revenue projection. With contracts secured until 2032, this covers several technology cycles, a rare advantage in such a volatile sector.

To support this expansion, CoreWeave plans to raise several billion dollars through bonds. The company announced a $3 billion issuance of convertible bonds maturing in 2032, as well as a separate issuance of $1.25 billion in unsecured bonds maturing in 2031.

These funds will be used to support growth, but also to refinance existing debt. This massive reliance on external financing underscores the capital intensity of the sector. With $35 billion committed by Meta, CoreWeave is becoming a central player in global AI infrastructure.

Moreover, this agreement shows that, from a sector valuation perspective, it is now about having the capacity to secure long-term contracts with the largest tech players. In the AI economy, value no longer solely resides in algorithms, but in the ability to run them at scale, reliably and continuously. Meta seems to have grasped this challenge well by launching this $21 billion bet with CoreWeave.