Meta, the social media behemoth, has expanded its AI infrastructure deal with CoreWeave by adding a new $21 billion cloud computing agreement, strengthening a partnership that now totals around $35 billion. This builds on a previous $14.2 billion deal signed in September 2025. The latest contract runs through December 2032 and gives the Mark Zuckerberg-led firm access to large-scale GPU-based cloud systems needed for AI model training and real-time deployment. The move comes as Meta increases spending to support its growing AI products and services across its apps.
Under the latest deal, CoreWeave will provide Meta with large-scale GPU cloud infrastructure designed specifically for AI workloads. These systems are optimized for both training and inference tasks. Training involves building and refining large AI models using huge datasets, while inference refers to running those models in real time for billions of user interactions every day. As Meta scales its AI products globally, inference has become increasingly resource-intensive, requiring low-latency, high-throughput computing environments that can handle constant demand across multiple applications simultaneously.
It is important to note that CoreWeave’s infrastructure is built around clusters of high-end GPUs, primarily supplied by Nvidia, which dominate the global AI chip market. These chips are essential for modern machine learning workloads, especially large language models that require parallel processing at extreme scale. The partnership is expected to include access to next-generation GPU architectures, like advanced Nvidia systems designed to significantly improve performance and energy efficiency for AI computation.
Meanwhile, despite the aggressive internal expansion, Meta continues to rely on external providers like CoreWeave to bridge immediate capacity gaps. Even the firm recently revealed that it could spend up to $135 billion on AI infrastructure alone in 2026.
At the same time, CoreWeave, which made its public market debut last year, has rapidly emerged as a key player in the AI infrastructure space. The company is often described as a ‘neocloud’ provider because its operations are optimized for AI workloads rather than traditional enterprise cloud services. By 2024, CoreWeave operated around 32 data centers. The firm competes with giants like Amazon Web Services (AWS) and Google Cloud. In 2025, the company gained further attention after securing several major partnerships, including a notable $11.9 billion deal with OpenAI.