Menlo Park: Meta Platforms Inc. pared sharp intraday losses on Friday after reports emerged that the company is in discussions with artificial intelligence startup Anthropic over a computing infrastructure lease agreement that could be worth up to $10 billion over two years. Although neither company has confirmed the negotiations, the reported deal has sparked investor interest as it could open a new revenue stream for Meta beyond its core digital advertising business.

The report comes at a time when technology companies are investing billions of dollars in artificial intelligence infrastructure to meet soaring demand for advanced computing power. If finalised, the agreement would represent one of the largest AI infrastructure leasing arrangements in the industry and further intensify competition in the rapidly growing market for high-performance computing.

Meta shares recover after report

Meta’s shares came under heavy pressure earlier in Friday’s trading session amid a broader sell-off in technology stocks.

The stock fell by as much as 6 per cent, touching around $626 during intraday trade before recovering following reports of the potential agreement with Anthropic. By late trading, Meta shares had climbed to $651.06, reducing losses to about 2 per cent compared to the previous close.

The market reaction reflected investor optimism that Meta may be able to generate meaningful revenue from its extensive AI infrastructure investments, helping offset concerns over the company’s significant capital expenditure.

Report outlines proposed agreement

According to a report by The New York Times, citing people familiar with the discussions, Anthropic proposed the arrangement to Meta in June.

Under the reported proposal, Anthropic—the developer of the Claude family of artificial intelligence models—would lease computing capacity from Meta and make monthly payments over a two-year period. Both companies would reportedly retain the option to terminate the agreement before its scheduled completion.

Neither Meta nor Anthropic has officially confirmed the discussions. Anthropic declined to comment on the report, while Meta did not immediately respond to media queries. Reuters also reported that it could not independently verify the negotiations.

AI infrastructure emerging as a business opportunity

If the agreement moves forward, it would signal Meta’s intention to monetise excess AI computing infrastructure in addition to supporting its own artificial intelligence projects.

The company has invested heavily in data centres, graphics processing units (GPUs) and other high-performance computing systems needed to train and deploy advanced AI models.

Until now, Meta has largely reserved its computing resources for internal development, including its Llama family of large language models and AI-powered services across Facebook, Instagram and WhatsApp.

Leasing spare computing capacity to third-party AI developers could create a new source of recurring revenue while improving returns on Meta’s infrastructure investments.

Zuckerberg has hinted at leasing capacity

Meta Chief Executive Officer Mark Zuckerberg previously acknowledged that outside companies have shown interest in purchasing access to the company’s computing infrastructure.

During the company’s May earnings call, Zuckerberg said businesses regularly approach Meta seeking AI computing capacity.

He noted that the company had prioritised its own AI requirements but indicated that leasing infrastructure could become a viable option if Meta ultimately builds more computing capacity than it requires.

His comments suggested that commercialising excess infrastructure had already become part of the company’s long-term strategic planning.

Massive AI spending under scrutiny

The reported agreement comes as investors closely monitor Meta’s substantial spending on artificial intelligence.

According to reports, Zuckerberg has indicated that the company expects to spend as much as $145 billion this year, with much of that investment directed towards AI infrastructure, including data centres, specialised processors and networking equipment.

While investors generally support Meta’s AI ambitions, the scale of its capital expenditure has raised questions about when those investments will begin generating meaningful financial returns.

A long-term computing lease with Anthropic could help demonstrate that Meta’s infrastructure investments are capable of supporting both internal AI development and commercial services for external customers.

Competition in AI infrastructure intensifies

A successful agreement would place Meta in more direct competition with specialist AI infrastructure providers such as CoreWeave and Nebius, which lease computing power to artificial intelligence companies.

Demand for high-performance computing has surged as AI developers race to build increasingly sophisticated language models and generative AI systems.

The reported Meta-Anthropic proposal is smaller than Anthropic’s previously reported $45 billion, three-year computing agreement with Elon Musk’s SpaceX, which also reportedly includes provisions allowing either party to terminate the arrangement before expiry.

As AI adoption accelerates globally, access to large-scale computing infrastructure has become one of the industry’s most valuable resources.

Although discussions between Meta and Anthropic remain unconfirmed, the reported negotiations highlight how major technology companies are exploring new business models to maximise returns from their rapidly expanding AI investments while meeting the growing demand for computing power.