Top OpenAI execs move to Meta

Meta just pulled in three senior OpenAI leaders who had been building the company’s giant data-center push, according to Bloomberg on April 11. They were part of the team behind Stargate, OpenAI’s umbrella for the computing campuses meant to feed its biggest artificial intelligence models. Those jobs are not about writing chatbot demos. They are about finding power, land, chips, cooling systems, financing, and construction crews for buildings that can draw electricity like small cities. Meta has been throwing money at exactly that problem. In its January 2026 annual results, the company said it expects 2026 capital spending of $115 billion to $135 billion, with the increase driven by Meta Superintelligence Labs and the core business. Mark Zuckerberg has also been promising campuses on a scale that used to sound absurd even in cloud computing. Bloomberg reported in July 2025 that Meta was building several gigawatt-size data centers, with the first Ohio project, called Prometheus, expected to come online in 2026. OpenAI’s side of the race has looked less smooth in the last few weeks. Bloomberg reported on April 9 that OpenAI paused its Stargate project in the United Kingdom, and Reuters said the pause was tied to high energy costs and an unfavorable regulatory environment. A month earlier, Bloomberg reported that Oracle and OpenAI had scrapped plans to expand a flagship Texas site in Abilene, and that the opening created room for Meta to consider leasing the expansion instead. Microsoft later took 900 megawatts at a separate project that had previously been earmarked for Oracle and OpenAI. That is why these hires look bigger than a normal talent raid. Meta is not just buying researchers who can improve a model in a lab; it is buying operators who know how to turn a chief executive’s promise of “hundreds of billions” into substations, permits, concrete, and live servers. The backdrop is that artificial intelligence has become a construction race as much as a software race. OpenAI said when Stargate was announced on January 21, 2025 that the venture intended to invest $500 billion over four years in United States infrastructure, starting with $100 billion immediately. Once projects reach that size, the scarce resource stops being only Nvidia chips. Bloomberg reported on April 1 that United States data-center expansion has been slowed by shortages of transformers, switchgear, and batteries, which means executives who can navigate supply chains and utility bottlenecks become as valuable as engineers who train models. So the story here is not simply that Meta hired away a few OpenAI names. It is that the fight over artificial intelligence is moving down the stack, from chatbots on screens to power contracts and steel in the ground, and Meta just grabbed people who know where those bottlenecks are.