Meta Platforms is exploring a significant infrastructure partnership with Anthropic, the artificial intelligence company behind Claude, with discussions centring on a potential lease agreement valued at up to $10 billion over a two-year period, according to sources with knowledge of the matter. The arrangement represents an early-stage exploration for both companies, though terms remain fluid and either party could potentially exit the agreement early depending on how negotiations progress.
The technology sector experienced a broad pullback on Friday, with Meta's shares declining more than 2 percent by market close, though the stock recovered slightly from deeper losses following disclosure of the discussions. Extended trading saw the shares maintain marginal losses as the broader market reassessed valuations across major technology holdings.
For Meta, such an arrangement would signal a strategic pivot toward generating revenue streams independent of its core advertising business, which has historically anchored the company's financial performance. By monetising its substantial computing infrastructure through cloud services, Meta could compete directly with emerging infrastructure providers such as CoreWeave and Nebius, which have gained traction as demand for AI computational resources accelerates globally. This diversification strategy reflects industry recognition that artificial intelligence deployment requires enormous processing capacity, creating new market opportunities for companies with substantial data centre infrastructure already in place.
Under the proposed structure, Anthropic would remit payments to Meta in monthly instalments throughout the two-year lease period, though the payment schedule and other commercial terms remain subject to negotiation and potential modification. The arrangement includes flexibility provisions that would permit either organisation to terminate early, reflecting the preliminary nature of discussions and the absence of certainty regarding final terms.
Anthropica initiated the proposal in June, seeking to secure reliable computing resources to support its AI model development and deployment operations. Meta is currently evaluating the proposal, though the company lacks established infrastructure for selling computing power commercially, creating operational and organisational complexities that have slowed negotiations. The absence of existing business systems and sales structures for cloud services has introduced complications into discussions, as Meta would essentially need to build out new organisational capabilities to execute such an arrangement.
This potential deal mirrors a strategic approach recently undertaken by SpaceX, Elon Musk's aerospace company, which concluded an agreement with Anthropic in May granting the AI firm access to the complete computing capacity of Colossus 1, SpaceX's advanced data centre facility in Memphis, Tennessee. That partnership demonstrated emerging market appetite for alternative cloud infrastructure providers outside traditional players, signalling openness to novel commercial arrangements involving computing resources.
Meta's Chief Executive Officer Mark Zuckerberg indicated during the company's May shareholder meeting that entering the cloud computing sector remained under active consideration. Zuckerberg specifically noted that technology companies approach Meta with considerable frequency seeking either direct access to the company's artificial intelligence models or spare computing capacity. This consistent inbound interest from potential customers provided strategic justification for Meta to develop formal cloud service capabilities.
Earlier in July, Bloomberg News reported that Meta had commenced building dedicated cloud infrastructure operations designed to monetise excess computing power while simultaneously offering hosting services for AI models developed by third-party software developers. This disclosure suggested Meta's intention to construct comprehensive cloud offerings beyond simple compute leasing, positioning the company as a potential competitor to established cloud providers.
The timing of these discussions reflects broader industry dynamics surrounding artificial intelligence adoption. As organisations across sectors accelerate AI implementation, computational demands have exceeded existing infrastructure capacity, creating pricing pressure and supply constraints. Companies possessing substantial, underutilised computing resources recognise significant commercial opportunities in monetising excess capacity during this period of exceptional demand growth.
For Malaysian and Southeast Asian technology stakeholders, Meta's infrastructure ambitions carry important implications. Regional companies seeking to develop or deploy advanced AI applications may eventually access Meta's computing resources at scale, potentially reducing dependence on overseas cloud providers and lowering costs for artificial intelligence projects. However, the arrangement also raises questions about data residency, regulatory oversight, and the concentration of critical computing infrastructure among dominant technology firms.
The preliminary status of negotiations means uncertainty persists regarding whether discussions will culminate in a binding agreement. Should talks succeed, however, the arrangement would represent a meaningful expansion of Meta's business model beyond social networking and advertising, establishing the company as a significant infrastructure provider within the rapidly evolving artificial intelligence economy.
Both Meta and Anthropic declined to provide comment regarding the discussions. The companies' reticence reflects standard practice when negotiations remain in early phases, with public disclosure potentially complicating ongoing talks or creating shareholder expectations that may ultimately go unfulfilled.
