Top compute talent moves
Reports say Meta has hired three senior infrastructure leaders from OpenAI as firms reorganise compute and talent strategies, and OpenAI is reportedly revising its Stargate compute plans. The shifts underline intense competition for people who build and run large AI systems. (outlookbusiness.com)
Meta has recruited three senior OpenAI infrastructure leaders, deepening its push to control the data centers and computing power behind its artificial intelligence systems. (bloomberg.com) Bloomberg reported on April 10 that Peter Hoeschele, Shamez Hemani and Anuj Saharan are joining Meta after helping lead OpenAI’s Stargate buildout. Meta declined to comment, and Bloomberg said OpenAI had earlier thanked the three for their work and said it remained focused on hiring for infrastructure. (bloomberg.com) These are not model researchers moving between labs. They are executives tied to the work of finding power, financing, sites and hardware for giant artificial intelligence clusters, the warehouse-scale computer systems used to train and run tools like ChatGPT. (ft.com) Meta is spending at a scale that makes those hires easier to understand. The company told investors it expects 2026 capital spending of $115 billion to $135 billion, up from $72.22 billion in 2025, with the increase tied to Meta Superintelligence Labs and core infrastructure. (datacenterdynamics.com) OpenAI’s side of the story has shifted in recent weeks. Bloomberg reported on March 6 that Oracle and OpenAI dropped plans to expand a flagship Stargate site in Abilene, Texas, after talks dragged over financing and OpenAI’s changing needs. (bloomberg.com) OpenAI also paused a Stargate project in the United Kingdom. Bloomberg reported on April 9 that the company cited high energy costs and said it was reining in spending plans, while the Financial Times said regulatory uncertainty was also a factor. (bloomberg.com, ft.com) That is a sharp contrast with how Stargate was introduced. SoftBank and OpenAI said on January 22, 2025 that the venture aimed to invest $500 billion in United States artificial intelligence infrastructure over four years, with $100 billion to be deployed immediately. (group.softbank) OpenAI later said it had lined up more capacity, announcing in September 2025 that five additional United States sites put it on a path toward the full $500 billion, 10-gigawatt commitment by the end of 2025. The latest pauses do not erase that plan, but they show how quickly site-by-site decisions can change when power prices, financing and customer demand move. (openai.com) Meta has been building the organization that would use this capacity. CNBC reported on April 8 that Meta Superintelligence Labs, overseen by Alexandr Wang after Meta’s $14.3 billion Scale AI deal, released a new model called Muse Spark as Zuckerberg tries to catch up with OpenAI and Google. (cnbc.com) The immediate contest is no longer only about who has the best model. It is also about who can hire the people who know how to secure land, chips, electricity and contracts fast enough to keep those models running. (bloomberg.com, ft.com)
