Meta to rent AI superclusters, competing with AWS and CoreWeave
Meta is planning to launch a cloud infrastructure business, branded as Meta Compute, to monetize its massive investment in AI data centers by selling excess compute capacity and model access. This move places Meta in direct competition with hyperscale cloud providers by marketing its AI-optimized hardware and proprietary models like Muse Spark.
Key Takeaways
- Meta Compute will offer a two-track service: raw compute capacity and a hosted-model API similar to AWS Bedrock.
- The initiative is led by infrastructure head Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and President Dina Powell McCormick.
- The 2026 capital expenditure guidance for Meta has ballooned to between $125 billion and $145 billion to support these facilities.
- Meta's Manhattan-sized data center in Ohio is scheduled to come online later this year to bolster available capacity.
Why It Matters
Meta is shifting from a pure AI consumer to a merchant of infrastructure, following a path recently cleared by SpaceX’s xAI. By offering raw 'bare-metal' compute and proprietary model access, Meta poses a direct threat to both hyperscale giants like AWS and specialized 'neoclouds' like CoreWeave. For the streaming and video industry, this increases the supply of localized high-end GPU clusters necessary for real-time generative video and multimodal reasoning. Watch for Meta to reclassify further capital assets as 'held-for-sale' as it formalizes third-party data center partnerships.
Additional Context
The pivot toward third-party leasing follows a period of massive consolidation and infrastructure scaling across the AI sector. Per Bloomberg (July 2026), Meta recently reclassified $2.04 billion in data center assets as 'held-for-sale' while seeking partners to co-develop facilities. This mirrors the aggressive commercialization seen at SpaceX, which transformed its Colossus supercluster in Memphis from a private training ground into a commercial hub. Per Forbes (June 2026), SpaceX secured over $80 billion in committed revenue through 2029 by leasing capacity to Anthropic, Google, and Reflection AI. Meta’s internal pressure to monetize compute is heightened by its unprecedented capital spending. Per CFO Susan Li in recent earnings (March 2026), Meta’s projected infrastructure spend for 2026 alone is between $115 billion and $135 billion, nearly double its 2025 outlays. This investment supports frontier projects like the 2,250-acre Hyperion campus in Louisiana, which is expected to deliver five gigawatts of power. CEO Mark Zuckerberg noted in May 2026 that external firms approach Meta 'almost every week' to lease spare capacity, suggesting immediate market demand for Meta's ultra-high-end GPU clusters. On the software side, the cloud business will leverage the Muse Spark model suite. Launched on April 8, 2026, by Meta Superintelligence Labs, Muse Spark is a natively multimodal reasoning model designed for high-speed output and complex tool-use. Per TechCrunch reporting, while Meta previously utilized third-party neoclouds like CoreWeave and Nebius for expansion—including a $27 billion deal with Nebius—the launch of Meta Compute signals a shift toward self-sufficiency and the eventual capture of developer workloads that once relied on those external partners.
Read full article at techcrunch.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source