Meta is building a cloud business to sell AI computing power, competing with Amazon and Google

Meta Platforms is developing plans to create a cloud business that will allow it to sell third-party customers access to AI computing and its own models. Bloomberg Intelligence reported this on July 1, 2026; in terms of the scale of ambitions, the project puts Meta in direct competition with Amazon Web Services, Microsoft Azure, and Google Cloud — the three historical leaders of the global cloud market.

What exactly Meta is preparing

According to Bloomberg Intelligence, the company is working on cloud infrastructure open to external customers. The offering should cover two key elements: computing resources — GPU clusters for training and inference of neural networks — and direct access to Meta's AI models, including presumably the Llama family.

In essence, this is about Meta's transition from being a consumer technology giant to the B2B cloud infrastructure segment. Today, this segment is controlled by three players: AWS, Microsoft Azure, and Google Cloud together occupy more than two-thirds of the global market, which exceeds $270 billion per year. The company has not yet disclosed specific terms, launch timelines, or pricing models.

The key difference between Meta and other potential cloud B2B contenders is the open ecosystem around the Llama model family. If the company offers cloud access to these models, it will automatically gain a huge ready-made audience of tens of millions of developers already working with its stack — without needing to win trust from scratch.

Why Meta has the resources to compete with AWS and Google

Several years of large-scale investments have transformed Meta into a company with hyperscaler-level infrastructure. In early 2025, the company announced planned capital expenditures of $60–65 billion per year — one of the largest such targets among global technology companies. Meta's assets include: extensive GPU clusters based on NVIDIA H100/H200, proprietary MTIA chips, and a sprawling network of data centers in North America, Europe, and Asia.

In parallel with building hardware, Meta has created strong positions in open-source AI. The Llama series of models has accumulated hundreds of millions of downloads and formed an active community of developers, companies, and researchers. Most cloud providers offer third-party models through APIs — Meta will be able to offer its own, and ones already well known to the market. This substantially lowers the barrier to attracting the first corporate clients.

What will change in the cloud AI services market

The appearance of a fourth major player will create additional price pressure — primarily in the niches of AI inference and pre-trained models, where competition is already intensifying thanks to specialized providers: CoreWeave, Lambda Labs, Together.ai. The major players — AWS, Azure, and Google Cloud — so far confidently maintain their market share, however, the struggle for developer audience is increasingly unfolding around token cost and the breadth of model catalogs.

For corporate clients, Meta could potentially offer an unconventional combination: hyperscaler-level infrastructure, open models with a broad ecosystem, and possible integration with the platform's advertising and social data. For AWS, Azure, and Google Cloud, this is an unconventional competitor — a company that for many years was itself their major customer and knows intimately how their ecosystem works from the inside.

What this means

Meta is moving from consuming AI infrastructure to selling it. If the plans are realized, the cloud AI services market will gain a fourth major player with giant proprietary infrastructure, strong open models, and direct access to a multi-million developer audience. The next signal to watch for is an official product announcement with specific terms and pricing policy.

*Meta is recognized as an extremist organization and is banned in the Russian Federation.