Meta Platforms and Anthropic, the artificial intelligence company behind Claude, are in preliminary negotiations over a substantial compute infrastructure deal that could be worth as much as $10 billion over a two-year period, according to sources close to the discussions. The potential agreement represents a significant strategic move for Meta as it pivots toward generating revenue streams from its substantial computing infrastructure, an area that has been a heavy investment focus for the technology giant. News of the talks emerged on Friday, though the companies have not publicly confirmed the negotiations, with Meta declining immediate comment and Anthropic also remaining silent on the matter.
The proposed arrangement would see Anthropic make monthly payments to Meta throughout the two-year contract, though neither party has confirmed exact financial terms and both retain the ability to exit the agreement early should circumstances change. The source familiar with the negotiations indicated that discussions remain in their nascent stages, and there remains considerable uncertainty about whether a final deal will materialise. Meta's consideration of such a proposal marks a departure from its traditional business model, as the company has historically relied primarily on advertising revenue rather than infrastructure leasing arrangements.
This potential arrangement comes as Meta faces a critical decision about its strategic direction in the artificial intelligence economy. During the company's shareholder meeting in May, Chief Executive Officer Mark Zuckerberg explicitly stated that expanding into cloud computing services was "definitely on the table," revealing that technology firms approach Meta regularly seeking access to either its AI models or surplus computing capacity. The statement signalled management's serious intent to monetise the company's substantial infrastructure investments, which have grown significantly as Meta competes in the generative AI space. More recently, Bloomberg News reported that Meta is actively building out a cloud business unit specifically designed to sell excess computing power and provide hosting services for developers building AI applications.
The competitive landscape for AI infrastructure services remains intensely crowded and rapidly evolving. Meta's potential move into compute leasing would pit it against emerging specialised providers such as CoreWeave and Nebius, which have built business models around providing on-demand computing power to technology firms. By leveraging its existing infrastructure and operational expertise, Meta could theoretically offer competitive advantages in terms of scale and reliability, though the company would need to establish sales and support operations for an entirely new business line. The infrastructure leasing market has become increasingly attractive to technology companies seeking alternative revenue sources as traditional business models face saturation and regulatory headwinds.
Anthropicâs interest in securing additional computing capacity reflects the intense computational demands of developing and deploying advanced large language models at scale. As one of the fastest-growing artificial intelligence companies, Anthropic requires enormous processing power to train its Claude models and support the growing user base leveraging the platform. The startup reportedly initiated these discussions with Meta in June, indicating its proactive approach to securing the infrastructure necessary to support its expansion plans. Anthropicâs strategy mirrors similar arrangements being pursued by other major players in the AI space seeking to secure reliable, long-term access to computing resources.
Anthropicâs approach parallels a strategy recently adopted by Elon Muskâs SpaceX, which in May secured an agreement to access the complete computing capacity of its Colossus 1 data center located in Memphis, Tennessee. That arrangement demonstrated the potential value of long-term infrastructure partnerships in supporting ambitious AI initiatives. SpaceXâs willingness to make such a commitment underscores the critical importance of reliable, dedicated computing resources for organisations pursuing cutting-edge artificial intelligence development. The precedent established by SpaceX suggests that similar arrangements may become increasingly common as the demand for AI infrastructure outpaces available capacity.
Meta's consideration of this deal reflects broader industry recognition that artificial intelligence infrastructure represents a fundamental bottleneck limiting the pace of AI advancement. The extraordinary computational requirements of modern large language models have created intense competition for access to high-performance computing resources. Companies that can reliably provision such infrastructure possess significant leverage and opportunity to generate substantial revenue. For Meta, which has invested tens of billions of dollars in building data centre capacity, converting that infrastructure into a revenue-generating asset represents a logical extension of existing investments.
The potential complications surrounding such an arrangement stem from Meta's lack of established infrastructure for selling computing services to external parties. The company would need to establish pricing models, service level agreements, billing systems, and customer support operations to execute such a business successfully. These operational requirements, while manageable for an organisation of Meta's scale and sophistication, nonetheless represent a material undertaking requiring dedicated management attention and resource allocation. The source indicated that these factors have contributed to the complexity of ongoing negotiations.
For Malaysian and Southeast Asian technology markets, this development carries implications for the region's aspirations to establish itself as an AI hub. The concentration of compute infrastructure ownership and leasing among a handful of technology giants suggests that most organisations in the region will likely depend on leasing arrangements rather than building proprietary infrastructure. This dependency could shape the trajectory of AI adoption across the region, potentially limiting innovation to applications built atop existing platforms rather than fundamentally new approaches. Understanding these infrastructure dynamics becomes essential for technology policymakers and companies across Southeast Asia seeking to participate meaningfully in the AI economy.
Meta's stock pared earlier losses following the news, closing down more than 2 percent on Friday as part of a broader technology sector decline. Extended trading saw shares decline only marginally, suggesting investors viewed the potential deal as modestly positive for the company's long-term prospects. The muted market reaction likely reflects lingering uncertainty about whether negotiations will successfully conclude and what the actual financial impact might prove to be. Regardless of the immediate stock market response, the talks signal Meta's serious commitment to diversifying its revenue model beyond advertising and positioning itself as a significant player in the growing AI infrastructure market.
