Meta could spin up its own cloud business to make use of the massive infrastructure investments it has made to train and run AI models, according to a report from Bloomberg. The social media giant has traditionally funded its risky hardware bets, like smart glasses and virtual reality, using its lucrative online advertising engine. Selling cloud infrastructure would place #Meta in direct competition with hyperscalers like Amazon, Google, and newly-minted cloud provider SpaceX.
The cloud business could offer multiple services, including selling access to AI models hosted on Meta’s infrastructure, or leasing the computing power of its high-performance data centers to other enterprises looking to train their own models. Offering a suite of services akin to AWS could help claw back some of Meta's heavy capital expenditures. As part of its aggressive AI roadmap, the company has committed to investing $600 billion in the US by 2028, alongside hiring top-tier talent for its AI superintelligence teams.
Meta Compute, an initiative focused on data centers and AI established in January, is currently spearheading the development of this new cloud division. Meta has yet to officially comment on these data center commercialization plans.
For now, Meta provides access to its AI models—such as the Muse Spark model referenced in reports—for free within its core apps like Facebook, WhatsApp, and Instagram. Paid tiers only unlock higher limits for image generation and reasoning. Meta's latest models are also expanding into wearables, starting with the recently launched Meta Glasses. Like Google, Meta is heavily focusing on AI agents capable of executing complex personal and professional workflows on behalf of users.
[AgentUpdate Depth Analysis] Meta’s transition into a cloud provider marks a pivotal shift from an infrastructure spender to an AI ecosystem enabler. By offering native cloud services, Meta can tightly couple its market-leading open-weights Llama models with optimized, first-party compute. This integration will significantly lower the hosting and inference costs for developers building advanced AI Agents. Furthermore, because AI Agents require low-latency loop executions and context retrieval, Meta’s massive global edge-compute network is uniquely positioned to power real-time agentic workflows across both mobile software and hardware like Ray-Ban Meta Glasses. This move challenges traditional cloud giants and establishes a highly competitive "Device-Model-Cloud" vertical integration for the future Agent economy.
