Meta plans AI cloud business, challenging Amazon, Microsoft, and Google

Industry Background

Meta has invested tens of billions of dollars in AI infrastructure, building data centers and procuring GPUs to support its research goal of "superintelligence." However, such massive investments have raised investor concerns about the return cycle. Meanwhile, AWS, Azure, and Google Cloud have already generated billions of dollars in quarterly revenue by renting out computing power and AI model services. Meta's computing needs are still in a phase of rapid growth, but the company is also beginning to face the possibility of "excess computing capacity" — if the capacity under construction exceeds internal demand, idle computing resources will incur huge sunk costs.

According to sources familiar with the matter, Meta is planning an internal project called Meta Compute, responsible for building and operating AI infrastructure, and intends to sell computing power and AI model access to third-party customers. The plan is jointly led by Meta's Head of Infrastructure Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and company President Dina Powell McCormick.

Market Impact

Following the news, Meta's stock price rose as much as 9.3% intraday on July 1, 2026, to $615.55, marking its largest single-day gain since April. Investors view this as a positive signal that Meta is monetizing its massive AI investments. Conversely, cloud service provider CoreWeave's stock fell 14%, and Nebius Group dropped 17%, indicating that the market expects Meta's entry to intensify competition in the AI cloud market.

For enterprise customers, more cloud options could lead to lower prices and service innovation. However, Meta's cloud business is still in the planning stage, with specific pricing, features, and availability yet to be determined. In the short term, it will not pose a substantial impact on existing cloud giants.

Competitive Landscape

Meta's AI cloud business may adopt two models: one is a model hosting service similar to AWS Bedrock, providing developers with API access to Meta's own models (such as Muse Spark); the other is a "bare metal" computing power rental similar to CoreWeave, directly offering data center and chip computing capabilities.The current AI cloud market is dominated by AWS, Azure, and Google Cloud, which occupy the enterprise market with their mature sales, support, and compliance systems. "New cloud" vendors such as CoreWeave and Nebius focus on GPU computing power leasing, serving AI startups. In addition, SpaceX (xAI) has recently entered this field by leasing its large data center in Memphis to Anthropic, and expects to achieve over $50 billion in cloud revenue by 2028.

Meta's differentiating advantage lies in its vast social network user base and internal AI R&D capabilities, enabling it to offer services deeply integrated with its own models. However, the disadvantages are also clear: a lack of enterprise-level cloud service experience, customer support systems, and compliance certifications, which may take several years to build trust.