Meta Platforms and Anthropic are in preliminary negotiations over a computing power lease arrangement potentially valued at $10 billion across a two-year period, according to sources close to the discussions. The deal would see Anthropic, the artificial intelligence company behind the Claude chatbot, lease substantial computational resources from Meta, with payments structured as monthly instalments. The arrangement remains fluid, with both parties maintaining flexibility to terminate the agreement should circumstances warrant an early exit. Neither company has publicly confirmed the talks, though the disclosure comes amid broader industry momentum toward infrastructure-as-a-service models for AI developers.
The potential transaction represents a significant strategic pivot for Meta, traditionally dependent on advertising revenues. By monetizing its infrastructure capacity, the technology conglomerate aims to generate supplementary income streams while competing against emerging cloud infrastructure providers such as CoreWeave and Nebius. This approach leverages Meta's substantial investments in data centre expansion and GPU acquisition—expenditures driven by the company's ambitious artificial intelligence development roadmap. The diversification strategy addresses investor concerns about the company's advertising-dependent business model and demonstrates management's intent to capture value from its computing infrastructure beyond internal applications.
Anthropically, which has been preparing for a public listing, initiated the proposal in June according to the source. The timing reflects the company's urgent need for additional computational resources as demand for its Claude AI models accelerates. As a company focused exclusively on developing and deploying advanced language models, Anthropic faces significant infrastructure demands but lacks the vast data centre networks that technology incumbents possess. Leasing computing power from Meta offers a pragmatic solution to capacity constraints while allowing the startup to preserve capital for product development and research.
Meta Chief Executive Officer Mark Zuckerberg signalled openness to cloud computing ventures during the company's May shareholder gathering, indicating that potential partners approach Meta approximately weekly seeking access to its artificial intelligence capabilities or surplus computing infrastructure. This receptiveness reflects strategic recognition that Meta's massive infrastructure investments—directed primarily toward training proprietary AI systems—could generate additional revenue through third-party utilisation. The computing resources currently allocated to Meta's own AI research represent expensive underutilised capacity during periods of reduced internal demand.
The discussions between the two firms follow a comparable arrangement that Anthropic negotiated with SpaceX in May, which granted the AI company access to Colossus 1, the company's dedicated data centre facility in Memphis, Tennessee. That agreement demonstrates the viability of computing power leasing arrangements and provides a template for negotiations between infrastructure owners and AI companies. SpaceX's willingness to monetize its computing capacity through such arrangements signals industry-wide recognition that data centre resources represent valuable commercial assets independent of the companies' primary business operations.
Meta's exploration of cloud computing services aligns with recent reporting that the company has commenced building infrastructure specifically designed to serve external customers. According to earlier industry coverage, Meta intends to offer excess computing capacity and host artificial intelligence models for third-party developers. This infrastructure buildout suggests management commitment to establishing a sustained cloud services operation rather than pursuing merely opportunistic arrangements. Such a business unit would require dedicated teams, technical support infrastructure, and commercial capabilities distinct from Meta's traditional operations.
The complexity surrounding these discussions stems partly from Meta's historical inexperience in commercial infrastructure leasing. Unlike established cloud providers such as Amazon Web Services or Microsoft Azure, Meta lacks operational frameworks for managing customer relationships around computing capacity sales, establishing service level agreements, and handling technical support obligations. These organisational and procedural gaps necessitate internal development before Meta can confidently commit to substantial multi-year arrangements, explaining why negotiations remain preliminary and potentially subject to significant modifications.
Meta's stock price fluctuated marginally following disclosure of the negotiations, declining slightly more than two percent on Friday alongside broader technology sector weakness. Extended trading sessions saw minimal further movement, suggesting investor ambivalence regarding the potential transaction's significance. Market participants may view the arrangement as strategically sound but immaterial relative to Meta's overall financial performance, or alternatively may harbour concerns about management distraction from core business operations.
For Southeast Asian technology observers, this development carries implications extending beyond the immediate transaction. The emergence of computing power as a tradeable commodity between technology giants reflects the increasing capital intensity of artificial intelligence development. Regional technology companies and startups may face growing challenges accessing affordable computing resources as major players establish proprietary infrastructure networks. Alternatively, increased competition among infrastructure providers could eventually reduce costs for developing economies seeking to build AI capabilities. The regional dimension of this trend warrants monitoring as artificial intelligence development becomes increasingly central to competitive advantage across Asia's technology sector.
