Meta just announced it’s building a new business unit called Meta Compute, designed to sell its surplus GPU and AI processing power to outside clients. The move, reported on July 1, 2026, sent Meta’s stock up roughly 9% as investors priced in what amounts to a brand-new revenue line for the social media giant.

Here’s the thing: when a company spending between $115 billion and $135 billion on capital expenditures in a single year decides to start renting out its spare computing capacity, every player in the AI infrastructure game should be paying attention.

The chip offensive

On March 11, 2026, Meta unveiled four new generations of its MTIA (Meta Training and Inference Accelerator) chips in a single announcement.

The MTIA 300 is already in production, handling ranking and recommendations training workloads. Three more generations, the MTIA 400, 450, and 500, are slated for deployment by the end of 2027.

Meta had previously deployed hundreds of thousands of its earlier MTIA 100 and 200 chips across its data centers. Now the company is layering on generative AI inference capabilities designed to handle a much broader range of workloads.

Around April 2026, Meta also announced a partnership with Broadcom to co-develop custom AI silicon, pursuing modular chiplet architectures that allow faster iteration cycles. Production of the newer MTIA models is targeted to begin as early as September 2026.

Meta Compute and the cloud wars

By offering raw infrastructure and hosted AI models to external clients, Meta is stepping directly into territory dominated by AWS, Google Cloud, and Microsoft Azure.

What this means for decentralized AI and crypto compute networks

Projects like Render Network, Akash Network, and io.net have built their value propositions around the idea that distributed computing can offer cheaper, more accessible, and censorship-resistant alternatives to centralized cloud providers. Meta’s $115 billion to $135 billion in planned 2026 spending is roughly equivalent to the total value locked across all of DeFi at various points in the past year.

The real advantages for decentralized alternatives lie in permissionless access, censorship resistance, and the ability to serve users and workloads that centralized providers won’t touch. Privacy-sensitive workloads, geographically distributed inference for edge applications, and compute access for developers locked out of traditional cloud platforms all represent potential moats.