The discussions mark a possible step toward a new business for Meta, which is building a cloud infrastructure unit to sell excess AI computing capacity from its massive data-center buildout [2]. The new line would put Meta in direct competition with established cloud providers such as Amazon Web Services, Microsoft Azure, and Google Cloud, officials said [1].
Meta has been rapidly expanding its data center network to support its own AI ambitions. A leaked internal memo disclosed plans to double the company’s overall computing capacity to 14 gigawatts by 2027, with 7 gigawatts to be deployed this year, according to reports [3][4]. The memo also detailed long-term supply contracts for memory, flash storage, and fiber-optic equipment amid a component shortage, the documents stated [4].
To fund this expansion, Meta borrowed $30 billion in debt, signaling that even cash-rich tech giants can no longer cover the astronomical costs of the AI race from revenue alone, according to an analysis [5]. The company has also signed multibillion-dollar deals with CoreWeave and Amazon for computing resources, and its own AI chips are slated to begin production in September [6][7][8]. Meta has said it may build more data centers than it needs based on the number of customers using its AI products, the report noted [1].
Meta CEO Mark Zuckerberg recently acknowledged that AI agent development over the past four months “hasn’t accelerated in the way we expected,” according to transcripts of his remarks [1]. The comment has raised questions about demand for Meta’s AI infrastructure, the report stated [1]. Selling excess computing power to companies such as Anthropic could provide a new revenue stream and potentially alleviate investor concerns about Meta’s multiyear data-center spending spree, analysts said [1].
At the same time, Meta has faced scrutiny over its data practices. The company launched Meta AI, a chatbot powered by Llama 4 that collects intimate details from conversations to monetize for targeted advertising, with critics warning that Meta AI turns private discussions into opportunities for product recommendations and ads [9]. Meta also cut about 8,000 jobs in May, part of a broader trend as AI adoption accelerates, according to reports [10][11]. The layoffs and the potential compute lease reflect a company under pressure to justify enormous capital outlays in a rapidly shifting market.
Meta is not the only company exploring the leasing of AI computing capacity. Elon Musk’s SpaceX, which acquired his AI startup xAI earlier this year, has been renting massive amounts of computing capacity from its Memphis data centers to Anthropic, according to people familiar with the matter [1]. That strategy could help xAI generate more than $50 billion in revenue by 2028 and $100 billion by 2030, company projections indicate [1]. The arrangement mirrors a broader trend where AI companies seek computing power from large infrastructure holders, industry observers said [1].
Other deals in the sector include Reflection AI signing a $1 billion compute deal with European AI infrastructure firm Nebius, and OpenAI in talks for a massive 10-gigawatt data center in Ohio with a potential buildout cost exceeding $500 billion [12][13]. Smaller and mid-sized companies often lack the financial resources to compete with giants like Meta and OpenAI, leaving them dependent on leasing capacity from larger players, one analyst noted [14].
Meta’s own AI models have not gained significant traction, according to recent industry reports. An assessment noted that “amid a fast-moving AI race… Meta’s models are nowhere to be found” [1]. The company has launched Muse Spark 1.1, an AI model for agentic coding, but acknowledged it is behind competitors like Anthropic and OpenAI, which have offered similar models for longer periods [15]. Meta is also developing in-house AI chips to lower GPU costs, but the project remains in early stages [6][16].
Anthropic, a direct competitor in AI development, would use Meta’s computing power while building its own systems. The deal, if completed, would mark an unusual arrangement between two competing firms, according to experts [1]. Anthropic CEO Dario Amodei has warned of the potential dangers of AI, cautioning that systems should not be dismissed as “just math” or “software,” and that the technology could lead to significant job displacement [17][18]. The evolving competitive dynamics highlight the immense capital and infrastructure demands shaping the AI industry.