Top OpenAI Infra Staff Departures
Reports say three senior infrastructure leaders tied to OpenAI’s 'Stargate' project have left for Meta, with Meta also planning a large infrastructure spend in 2026. Coverage links these moves to shifting talent flows between leading model builders. (newsbytesapp.com) (outlookbusiness.com)
Three senior OpenAI infrastructure leaders tied to Stargate have left for Meta, according to reports published April 11, adding a new twist to the fight over who builds the computing backbone for advanced artificial intelligence. (outlookbusiness.com) (newsbytesapp.com) The reported departures are Peter Hoeschele, Shamez Hemani, and Anuj Saharan. Outlook Business said Hoeschele helped launch Stargate, while NewsBytes said all three moved to Meta Platforms as Mark Zuckerberg expands the company’s infrastructure push. (outlookbusiness.com) (newsbytesapp.com) Stargate is OpenAI’s umbrella project for giant data-center buildouts, the warehouse-scale facilities packed with chips that train and run artificial intelligence models. OpenAI said in January 2025 that SoftBank, OpenAI, Oracle, and MGX were the initial equity funders, with SoftBank handling financing and OpenAI handling operations. (openai.com) OpenAI then widened the buildout in 2025. In July 2025, it said Oracle and OpenAI would develop 4.5 gigawatts of additional United States data-center capacity, and in October 2025 it said five new United States Stargate sites would bring planned capacity to nearly 7 gigawatts and more than $400 billion in investment over three years. (openai.com 1) (openai.com 2) The latest reports say some of that plan is being reworked. Outlook Business reported that OpenAI had scaled back parts of Stargate, including a United Kingdom pause and changes around the Abilene, Texas, site, though OpenAI’s own recent posts still describe Abilene as operating and additional projects as underway. (outlookbusiness.com) (openai.com 1) (openai.com 2) Meta is spending at a scale that helps explain the hiring. CNBC reported on February 6 that Meta told investors its 2026 capital spending could reach $135 billion, and CNBC reported on March 26 that Meta raised its El Paso, Texas, data-center investment to $10 billion, with 1 gigawatt of capacity planned by 2028. (cnbc.com 1) (cnbc.com 2) That spending is aimed at the same bottleneck every model builder faces: enough power, land, chips, and networking to keep training systems getting larger. Meta said in March that its data centers are a core part of its artificial intelligence and national-competitiveness push in the United States. (about.fb.com) The talent fight has not been limited to infrastructure jobs. Outlook Business previously cited Wall Street Journal reporting that Meta had also hired three OpenAI researchers from the Zurich office for its superintelligence unit, showing that the competition now reaches both the people designing models and the people securing the machines underneath them. (outlookbusiness.com) OpenAI is still publicly presenting Stargate as an active long-term platform, including a May 2025 announcement for Stargate United Arab Emirates with a 1-gigawatt cluster in Abu Dhabi and 200 megawatts expected to go live in 2026. That leaves the clearest immediate fact as a personnel shift: Meta is hiring deeply into the teams that know how to turn artificial intelligence ambition into power, chips, and concrete. (openai.com) (outlookbusiness.com)