Meta launches Meta Compute cloud to lease excess AI GPU capacity to clients

Meta Platforms is preparing to launch a dedicated cloud computing business, tentatively named Meta Compute, according to a Bloomberg report dated July 1, 2026. The unit would allow external clients to rent Meta’s excess AI computing capacity—including powerful GPUs and specialized hardware used for training large language models. This represents a major pivot for CEO Mark Zuckerberg, who had earlier insisted that all of Meta’s massive AI infrastructure would be used internally for its own products like LLaMA models, recommendation engines, and the metaverse. The shift suggests that Meta’s investment in AI data centers has outpaced its own needs, creating a surplus that can now be monetized.

If realized, Meta Compute would enter a fiercely competitive cloud market dominated by Amazon Web Services, Microsoft Azure, and Google Cloud. However, Meta has a unique advantage: its infrastructure is purpose-built for large-scale AI workloads, including the kind of high-bandwidth, low-latency compute required for training frontier models. The news triggered optimism among investors, with Meta’s stock rising on the prospect of a new high-margin revenue stream. While details on pricing, availability, and target customers remain sparse, the move signals that AI capacity leasing is becoming a viable business beyond the major cloud incumbents, and could reshape the economics of AI model development.