From 14.2 Billion to 21 Billion, Why Meta Continues to Bet on CoreWeave?
AI Podcast
CoreWeave and Meta have entered a $21 billion long-term AI cloud computing partnership, extending through December 2032. This agreement, Meta's largest with CoreWeave, signifies increased AI infrastructure investment and global demand for computing power. The collaboration includes the initial large-scale commercial deployment of Nvidia's Vera Rubin platform, optimizing Meta's AI inference tasks. This partnership significantly diversifies CoreWeave's customer base, reducing single-client revenue dependency.
TradingKey - On April 9, AI cloud computing provider CoreWeave ( CRWV) announced a partnership with Meta ( META) to enter into a long-term cooperation agreement worth $21 billion, providing the latter with dedicated AI cloud computing capacity through December 2032.
This represents the largest commercial partnership announced by CoreWeave since its Nasdaq listing last March. It not only further solidifies its market position as a core provider of AI infrastructure but also provides a strong endorsement for the large-scale commercial deployment of Nvidia's next-generation Vera Rubin platform.
Following the announcement, CoreWeave shares rose nearly 2.5% in pre-market trading, and Meta shares also climbed nearly 2%.
Analysts believe that this partnership upgrade marks a new high in Meta's investment in AI infrastructure, while also reflecting the continued strong global demand for AI computing power.
In fact, this is not the first collaboration between the two companies. As early as September 2025, Meta signed a $14.2 billion AI infrastructure partnership agreement with CoreWeave, effective through 2031. Combined with this new $21 billion agreement, the total scale of cooperation between the two parties has reached $35.2 billion, making Meta one of CoreWeave's most significant clients.
Under the terms of the agreement, CoreWeave will provide Meta with customized AI cloud computing capacity and adopt a global multi-data center distributed deployment architecture. Of particular note is that this partnership will include the initial deployment of Nvidia's next-generation Vera Rubin platform, marking the first large-scale commercial application of the platform.
CoreWeave stated that the computing resources provided will use a distributed multi-node deployment, aiming to comprehensively optimize the operational performance, system stability, and scalability of Meta's AI business to support its increasingly complex large-scale AI inference task requirements.
CoreWeave views this agreement as a "clear signal of the accelerating industry demand for high-performance AI infrastructure," believing that such infrastructure, capable of supporting large-scale complex AI workloads, is becoming the core competitive arena for technology companies.
Meta's Computing Power Deployment
As a provider specializing in GPU computing services, CoreWeave has deployed hundreds of thousands of Nvidia GPUs within its data centers. By supporting the operation of large AI models with high-performance computing power, it has become a critical infrastructure supplement for the rapid expansion of hyperscale cloud service providers.
Although tech giants such as Meta are constructing their own data centers, they still rely on CoreWeave's additional computing power to meet short-term explosive demand. Currently, global tech giants including Google, Microsoft, and OpenAI are all customers of CoreWeave.
In March of this year, Meta announced it would invest $10 billion to build a new data center in Texas, continuing to ramp up its own computing infrastructure. CoreWeave CEO Mike Intrator remarked that while tech giants have the capability to build their own computing power, they still choose to procure services from CoreWeave because it provides higher-quality computing solutions.
This expansion of cooperation is a microcosm of Meta's significant increase in AI infrastructure investment this year. According to Meta's latest financial disclosure, the company's capital expenditure for this year is expected to be between $115 billion and $135 billion, which is not only higher than previous Wall Street expectations but also approaches double its 2025 capital expenditure.
While Meta's core advertising business has benefited from AI development, it has long been in a catch-up position in the field of AI models. To change this, Meta invested $14.3 billion and spent nine months re-architecting its AI technology stack, and established a "Superintelligence Lab" focused on advanced AI model development.
On April 8 local time, Meta launched its first proprietary large model, Muse Spark. Following the announcement, the company's stock price surged nearly 9% at one point, adding over $100 billion to its market capitalization in a single day.
The Value of CoreWeave
As one of the earliest companies in the industry to deploy the NVIDIA GPU cloud rental model, CoreWeave’s success stems from its high sensitivity to the marketization of GPU resources. Through early, deep strategic ties with NVIDIA, the company not only secured valuable capital injections but also gained near-monopolistic priority within the computing power supply chain.
This "special treatment" ensures its ability to secure H100, H200, and the latest Blackwell architecture chips ahead of the market. In a context of extreme computing power scarcity, even giants like Microsoft ( MSFT) that build their own cloud infrastructure have been forced to seek CoreWeave’s support when facing business peaks.
Currently, the global AI industry is in a period of exponential expansion in demand for computing power. Traditional general-purpose data centers often struggle to handle massive model training and inference workloads. Although tech giants are aggressively investing in capacity expansion, the structural imbalance between supply and demand remains significant.
This environment has created a niche for vertical computing service providers such as Fluidstack, Nebius Group NV, Nscale, and CoreWeave. Compared to traditional cloud giants, these "new cloud" providers have seen their valuations skyrocket due to more flexible scheduling and specialized computing supply, becoming the sub-sector with the strongest growth momentum in the current AI value chain.
CoreWeave Achieves Breakthrough in Customer Diversification
Furthermore, this new agreement is of significant strategic importance for CoreWeave. As a computing power service provider that completed its IPO last year, CoreWeave was previously heavily dependent on a single customer, with Microsoft contributing 62% of its revenue in 2024, making its customer concentration risk a focal point of concern.
Meta stated that the partnership with CoreWeave is a key component of the company's "diversified infrastructure portfolio strategy," aimed at providing continuous computing power support for AI development.
CoreWeave CEO Mike Intrator revealed that following the new agreement with Meta, the proportion of the company's revenue from any single customer will drop below 35%.
Meanwhile, Intrator remains optimistic about the future of the partnership, believing that even as Meta continues to expand its own data centers, it will maintain a long-term collaboration with CoreWeave: "They will continue to build their own facilities, but they will also continue to work with us; they cannot rely on a single approach."
This content was translated using AI and reviewed for clarity. It is for informational purposes only.
Recommended Articles
