For years, Meta has been one of the biggest buyers of AI compute on the planet. Now it is turning itself into a seller.

On July 1, 2026, Bloomberg reported that Meta is building a new cloud computing business to sell excess AI capacity to enterprise customers, putting it in direct competition with Amazon Web Services, Microsoft Azure, and Google Cloud. Meta’s stock jumped roughly 9 percent on the news, closing at approximately $612 per share on three times its average daily trading volume.

The move formalises what had been telegraphed internally. Earlier this year, Mark Zuckerberg established a new top-level initiative called Meta Compute, a programme to coordinate the company’s infrastructure ambitions at a scale it describes as “tens of gigawatts this decade, and hundreds of gigawatts or more over time.” The initiative is co-led by Santosh Janardhan, Meta’s head of global infrastructure, and Daniel Gross, a prominent AI operator and investor.

What Meta Is Planning to Offer

According to reports, Meta Compute will operate on two distinct layers.

The first is raw compute access: enterprises would be able to rent GPU clusters optimised for AI training and inference workloads, similar to what AWS, CoreWeave, and Together AI offer today. Meta has accumulated more than 600,000 H100-equivalent GPUs across its data centres, more than any single non-cloud enterprise customer in the world.

The second layer is hosted model access: companies would be able to run Meta’s Llama family of large language models through Meta-managed infrastructure, without needing to manage the hardware or model deployment themselves. This mirrors how Amazon Bedrock and Google Vertex AI work for third-party models today.

Pricing has not been confirmed, but multiple analyst estimates suggest Meta could afford to undercut existing cloud providers by 20 to 30 percent, given that it has already paid for the infrastructure to power its own products at scale and can amortise the cost over external revenue.

Why Now

The timing is not accidental. AI infrastructure has become a genuine constraint for enterprises trying to scale beyond pilots. Hyperscaler GPU capacity remains tight, wait times for reserved instances stretch months in some regions, and costs for frontier model inference have climbed steadily as models grow more capable.

Meta is sitting on a potential solution to all three problems. It has the compute. It has open-source models that many enterprises are already testing. And it has a financial incentive to offset the enormous capital expenditure it is making, and the company has committed more than $60 billion to AI infrastructure in 2026 alone.

The market recognised this immediately. A 9 percent stock jump on a single day is a meaningful signal that investors see the cloud business as a structural upgrade to Meta’s revenue model, not just a side project.

What This Means for Business

If Meta enters the enterprise cloud market in earnest, the most immediate effect for business leaders is leverage.

More competition among cloud providers means more room to negotiate pricing on GPU capacity and AI inference costs. Companies currently locked into a single hyperscaler relationship will have a credible alternative to bring to the table. The history of cloud pricing suggests that when a fourth serious competitor enters, rate cards across the whole market tend to move.

For businesses specifically interested in open-source AI, Meta’s offer could be particularly attractive. Running Llama models on Meta’s own infrastructure would mean dealing with a single vendor for the model and the compute, potentially simplifying licensing questions and reducing the operational overhead of managing fine-tuned models across third-party clouds.

There are legitimate reasons to wait and see. The service has not launched publicly. Meta’s enterprise sales and support capabilities are unproven compared with AWS or Azure. And any business adopting a new infrastructure provider needs to weigh switching costs, compliance obligations, and service reliability against the pricing upside.

But the direction of travel is clear: the AI compute market is getting more competitive, and that is generally good news for any business investing in AI. The cost of running AI agents, fine-tuned models, and inference workloads has been one of the most stubborn barriers to ROI for enterprise deployments. If a genuine fourth competitor holds pricing down, more of those deployments become financially viable.

Watch for a formal Meta Compute enterprise launch announcement in the coming weeks, likely tied to a partner programme or developer preview event.