Meta is planning a cloud infrastructure business that will sell AI computing power and AI model access, according to a Bloomberg News report published Wednesday, citing people familiar with the matter. The plans are still in development and the strategy could change, the report noted, though the reaction in financial markets suggested investors saw it as a credible and significant shift in Meta’s business model.
The initiative, internally called Meta Compute, is reportedly being built under the leadership of Santosh Janardhan, Meta’s head of infrastructure, alongside Daniel Gross, a leader within Meta Superintelligence Labs. According to sources cited by Bloomberg, the company is weighing two distinct approaches. The first would involve offering developers access to AI models hosted on Meta’s existing infrastructure, including its Muse Spark model line, similar to the way Amazon Web Services operates its Bedrock service. Meta would run the data centres and chips powering the models while charging developers for access. The second approach under consideration would involve selling access to raw computing capacity directly, an offering more akin to so-called neocloud businesses such as CoreWeave.
Meta declined to comment when contacted, and Reuters said it could not independently verify the underlying details of the Bloomberg report.
Zuckerberg Had Already Signalled This Was Coming
The move was not entirely unforeseen. Meta CEO Mark Zuckerberg discussed the concept during the company’s earnings call in May, describing the demand he had been fielding from outside firms. “Almost every week, there are different companies that come to us from outside asking us to both stand up an API service, or asking if we have compute that they could buy from us at some premium to what we’ve bought it at,” he said at the time.
“We haven’t done that yet, because we think that we have a use for the compute, but obviously if we get to a point where we feel that we have overbuilt, then that is an option that we have, and that is partially what gives us confidence in investing in building this out,” Zuckerberg added, a comment that, in hindsight, appears to have previewed exactly the pivot now being reported. The financial scale backing that confidence is considerable. Meta has raised its 2026 capital expenditure guidance to between $125 billion and $145 billion to support data centre capacity, including a 1-gigawatt data centre under construction in the American Midwest and a 2,250-acre hyperscale campus in Louisiana named Hyperion.
The Market’s Verdict Was Immediate and Sharply Divided
Investor reaction to the report was swift. Meta shares rose more than 10 per cent, easing pressure on a stock that had underperformed the S&P 500 for the year, having declined nearly 15 per cent as of the day before the report. The rally reflected relief among investors who had grown uneasy about the scale of Meta’s AI spending without a clear path to generating returns on that infrastructure.
The reaction among neocloud providers was the inverse. CoreWeave and Nebius, two AI cloud providers that had each signed multi-billion-dollar contracts to supply Meta with compute capacity, fell 10.8 per cent and 12.4 per cent respectively. The stakes for these firms are considerable given the existing commercial relationship: CoreWeave holds a $21 billion supply contract with Meta, part of a total backlog that has grown to $99.4 billion, while Nebius has a deal with Meta worth up to $27 billion. Neither firm was competing with Meta when those contracts were signed. Both now potentially are.
Gil Luria, managing director at D.A. Davidson, characterised the competitive dynamic succinctly. “The impact of adding Meta’s capacity to the market is more likely to be on neoclouds than the big hyperscalers. Those companies like CoreWeave and Nebius rely on Meta for their growth and Meta may not need them anymore,” he said, drawing a comparison to SpaceX, which similarly moved from being purely a compute buyer to renting out capacity in its Memphis data centres to firms including Anthropic and Google.
A Broader Pattern of AI Giants Seeking Returns on Infrastructure
Meta’s reported pivot fits within a wider industry pattern of AI companies searching for ways to monetise infrastructure built at extraordinary scale and cost. If Meta proceeds, it would become a fourth major seller of large-scale AI compute and hosted models alongside AWS, Azure and Google Cloud, entering a market currently dominated by established hyperscalers with years of enterprise relationships and operational maturity.
The move also arrives alongside other signs that Meta is exploring new revenue models beyond its traditional advertising base. The company has separately been reported to be testing consumer price cuts on its AI offerings while simultaneously considering a $199.99 premium monthly tier for its Hatch AI agent, positioning it directly alongside Anthropic and OpenAI at the high end of the consumer AI market. Together, these moves suggest a company testing multiple monetisation paths simultaneously, even as it continues pouring unprecedented capital into the underlying infrastructure that makes all of them possible.
Whether Meta Compute ultimately launches, and at what scale, remains uncertain given that the plans are still in development. But the market’s reaction on Wednesday, a sharp rally for Meta paired with a sharp selloff for CoreWeave and Nebius, has already demonstrated how consequential the shift could be for the broader AI infrastructure ecosystem, where the line between customer and competitor is proving far less stable than it once appeared.
