Meta Platforms and artificial intelligence startup Anthropic are in advanced discussions over a computing infrastructure lease arrangement that could be valued at up to $10 billion across a two-year period, according to sources privy to the negotiations. Under the proposed terms, Anthropic—the creator of the Claude Code AI tool—would pay Meta through monthly instalments spread across the duration of the agreement, with provisions allowing both parties to exit the arrangement prematurely if circumstances change. While the negotiations remain fluid and the final structure has yet to be finalized, the deal would represent a significant shift in how technology giants are leveraging their vast data centre capabilities.
The disclosure emerged as Meta's shares initially pared back their losses during Friday's broader technology sector downturn, though the stock ultimately closed down more than two percent for the session. Extended trading saw marginal declines persist, suggesting the market remained cautious about the company's near-term trajectory despite the potential revenue-generation opportunity. The modest equity market reaction underscores investor uncertainty about whether such infrastructure licensing deals can meaningfully offset advertising revenue pressures that tech platforms currently face.
For Meta, this potential arrangement addresses a strategic imperative to diversify its revenue base beyond its historically dominant advertising business. The social media giant possesses substantial computing infrastructure built to support its AI research and content recommendation systems, creating an opportunity to monetize excess capacity and compete more directly with specialized cloud computing providers. Companies such as CoreWeave and Nebius have emerged in recent years specifically to service the exploding demand for high-performance computing resources driven by generative AI adoption. By leasing its own infrastructure, Meta could capture a portion of this lucrative market while keeping capital investments closer to full utilization.
Anthrophe initiated the proposal in June, recognizing the computational demands required to operate its Claude models at scale. As an AI startup preparing for an initial public offering, Anthropic faces mounting infrastructure costs as it scales operations and responds to customer demand. Securing committed computing capacity from an established technology giant like Meta offers both cost certainty and operational flexibility compared to spot-market purchases on traditional cloud platforms. The arrangement also strengthens Anthropic's financial profile ahead of its planned public market debut by securing reliable supply agreements that can be highlighted to potential investors.
However, the negotiations have encountered complications stemming from Meta's lack of prior experience operating a commercialized computing services business. Unlike AWS, Microsoft Azure, or Google Cloud—which have spent years refining service delivery, contractual frameworks, and customer support infrastructure—Meta faces organizational and operational hurdles in serving external clients at scale. The company has not previously marketed computing power as a standalone product, requiring new business processes, technical support mechanisms, and financial reporting structures specifically dedicated to this revenue stream.
Meta's senior leadership has signalled institutional support for such initiatives. During the company's shareholder meeting in May, Chief Executive Officer Mark Zuckerberg explicitly acknowledged that entering cloud computing services remained very much a possibility, noting that technology firms approach Meta almost weekly seeking access to either its proprietary AI models or uncommitted computing capacity. This sentiment reflects growing market awareness of Meta's infrastructural advantages and the untapped revenue potential they represent. Subsequently, Bloomberg News reported that Meta was actively constructing the technical and commercial infrastructure necessary to launch a formal cloud business targeting developers and enterprises requiring AI model hosting services.
The potential Meta-Anthropic arrangement appears to follow a strategic template recently established by SpaceX. In May, Anthropic secured access to the full computational capacity of SpaceX's Colossus 1 data centre facility located in Memphis, Tennessee. That agreement demonstrated how AI companies facing computational bottlenecks can establish direct arrangements with infrastructure operators, bypassing traditional cloud intermediaries. The SpaceX precedent likely influenced Anthropic's approach to Meta and provided a working model for how such arrangements could be structured and executed.
For the broader Asian technology ecosystem, including Malaysian tech enterprises and regional cloud computing providers, Meta's potential pivot toward infrastructure services carries strategic implications. The intensifying competition for computing resources and the entry of major platform companies into the infrastructure rental market could reshape pricing dynamics and service availability across the region. Malaysian startups and enterprises dependent on cloud computing may face either improved pricing competition or potential supply constraints as major technology corporations prioritize serving each other or large-scale AI applications over smaller regional customers.
The discussions remain preliminary and subject to substantial change, with sources emphasizing that no binding agreement has been reached and negotiations may ultimately prove unsuccessful. Both Meta and Anthropic declined immediate comment when contacted by Reuters, suggesting the organizations prefer to avoid public positioning while sensitive commercial discussions continue. The confidential nature of ongoing talks means the actual financial terms, service specifications, and contractual obligations could differ materially from currently reported figures as both parties work through technical, legal, and commercial complexities.
