Meta CEO Mark Zuckerberg took to his social community Threads two weeks in the past to announce Meta Compute, a brand new “top-level initiative” led by the corporate’s most senior executives. In doing so, he reemphasized Meta’s dedication to being an AI infrastructure behemoth—and signaled that Meta has no intention of being an also-ran within the knowledge middle build-out race.
The brand new group is designed to safe the large quantities of computing energy—measured in gigawatts, every of which may energy a whole lot of 1000’s of properties—wanted for Meta’s drive to construct AI fashions that result in “superintelligence.”
“Meta is planning to construct tens of gigawatts this decade, and a whole lot of gigawatts or extra over time,” Zuckerberg wrote. “How we engineer, make investments, and accomplice to construct this infrastructure will turn into a strategic benefit.”
Beneath the brand new construction, Zuckerberg mentioned longtime Meta govt Santosh Janardhan will proceed to run the corporate’s technical structure, software program, customized chips, and the day-to-day constructing and operation of Meta’s huge knowledge middle community. In the meantime, Daniel Gross—one among Zuckerberg’s high-profile AI hires from final summer time, who was beforehand cofounder of Protected Superintelligence with former OpenAI chief scientist Ilya Sutskever—will lead a brand new group centered on the lengthy recreation: how a lot computing energy Meta will want years from now, the place it ought to be constructed, the right way to safe scarce chips and vitality, and the right way to mannequin the enterprise influence of these bets.
Zuckerberg additionally introduced a brand new Meta president and vice chair, Dina Powell McCormick, to work on creating partnerships with governments to finance and deploy knowledge facilities all over the world. She was beforehand deputy nationwide safety advisor for technique to President Trump.
A rising notion that Meta is taking part in catch-up
For some Meta-watchers, the Meta Compute announcement was puzzling. In spite of everything, Meta is already an AI infrastructure large. It has been a yr for the reason that firm broke floor on its Hyperion web site—a 4-million-square-foot knowledge middle campus in northeast Louisiana that Zuckerberg famously instructed President Trump was roughly corresponding to the scale of decrease Manhattan. Why, then, did Meta instantly must declare a brand new top-level group to do what it already seems to be doing at historic scale?
“This was sort of a head scratcher, I didn’t perceive it at first,” mentioned Patrick Moorhead, founder and chief analyst at Moor Insights and Technique. He advised the message about Meta Compute was actually meant for buyers and workers, signaling that Meta stays a critical contender in a membership led by Microsoft, Google, Oracle, OpenAI, and xAI. “It’s Meta saying, ‘Right here is our technique to roll this out,’” he mentioned.
Rick Pederson of Bow River Capital agreed that the announcement responds to a rising notion amongst market analysts that Meta is taking part in catch-up with Google and OpenAI within the AI arms race.
“It was a technique to focus on their focus, their intentionality surrounding the build-out of AI, computational capability and infrastructure,” he mentioned. “I’d think about that the opposite main hyperscalers have organizations one thing like this. However he took the chance to outline it.” Though the corporate spent over $70 billion on AI infrastructure final yr and plans to spend one other $600 billion within the subsequent two years, Google and OpenAI are spending at the least that quantity, he added. “So I feel this gave Zuckerberg an opportunity to speak about not simply the emphasis, but in addition how they’re going to do it.”
Doubling down on infrastructure as an funding portfolio
Some consultants mentioned they don’t seem to be in any respect shocked by the transfer to announce Meta Compute. “Meta is doubling down on infrastructure as an funding portfolio somewhat than merely a value middle,” mentioned Lane Dilg, former head of infrastructure coverage at OpenAI and founding father of boutique advisory agency Apeiro.
Because the AI growth accelerates, Meta not sees knowledge facilities, GPUs, energy contracts, and customized chips as mere plumbing for its merchandise. It treats them as strategic property—extra like a capital allocator than a tech firm. In impact, Dilg mentioned, Meta is positioning itself not simply in opposition to different hyperscalers, however in opposition to the world’s most refined funding platforms.
In that panorama, she defined, selecting Daniel Gross to co-lead the initiative is sensible. “Gross’s expertise constructing AI-native and agentic platforms issues, paired together with his compute experience,” she mentioned, pointing to the supercomputer he constructed with NFDG coinvestor Nat Friedman—now Meta’s head of merchandise. Their effort grew to become the Andromeda Cluster, a community of computing stockpiles totaling greater than 4,000 GPUs, which they made obtainable to their portfolio corporations at below-market charges.
Gross has dropped hints about his personal recruiting push for Meta Compute, saying in a latest submit on X that he’s hiring folks with backgrounds in “deep studying, provide chains, commodities, semiconductors, sovereigns, vitality, Excel, prediction markets, monitoring conditions, and so forth.” This means Meta is getting ready to hedge in opposition to volatility in energy and {hardware} prices—and to make long-term bets formed not simply by know-how, however by vitality markets, provide chains, and geopolitics.
Powell McCormick can be a key strategic rent for the Meta Compute effort, Umesh Padval, an skilled investor and board member, instructed Fortune. “Hyperscalers are actually centered on the right way to get energy and are investing in energy tasks financed by money and debt,” he mentioned. “Along with her banking and political background, she would work with the Meta knowledge middle group to finance and get approvals to allow quicker constructing of the compute capability.”
Critics say Meta’s capital-intensive mannequin would drag down returns
Not all are supportive of Meta’s transfer, nevertheless. In a social media submit shared the day Meta Compute was introduced, “Huge Brief” investor Michael Burry wrote: “Meta offers in, throwing away its one saving grace. Watch ROIC crash.” Burry’s warning displays a concern that Meta is abandoning its capacity to generate huge earnings with out sinking huge sums into bodily infrastructure. By embracing gigawatt-scale knowledge facilities, he argues, Meta is shifting towards a much more capital-intensive mannequin—one that might drag down returns and make the corporate look extra like a utility.
However since Meta has already poured tens of billions into AI knowledge facilities and signaled a whole lot of billions extra in long-term infrastructure commitments, it seems to be like that prepare left the station way back.
