Three senior executives who built the foundation of OpenAI's ambitious Stargate data center initiative are leaving the company to join Meta, according to a Bloomberg report published Friday. The departures mark the latest sign of turbulence in OpenAI's infrastructure strategy as it prepares for a potential IPO.
Who Is Leaving
Peter Hoeschele, who spearheaded early planning and operations for the Stargate project, has already exited OpenAI. Shamez Hemani, who worked on computing strategy, partnerships, and business development, and Anuj Saharan, who handled supply chain logistics and compute strategy, are also departing.
An OpenAI spokesperson said in a statement: "We're grateful for the contributions Peter, Shamez, and Anuj have made to OpenAI and wish them the very best in what comes next."
A Strategic Pivot Away From Owned Infrastructure
The exits come as OpenAI has fundamentally shifted its infrastructure approach. The original Stargate vision called for a $100-billion-plus custom supercomputer with millions of GPUs. Instead, OpenAI is now partnering with Microsoft Azure and Oracle to rent compute capacity rather than building proprietary facilities.
The UK arm of the Stargate project, which involved a partnership with NVIDIA and Nscale, has been paused due to escalating energy costs and regulatory hurdles. Active data center projects continue in the United States, Norway, and the UAE, but the broader ambition has been scaled back.
OpenAI has brought in Sachin Katti to lead the compute division, signaling a new chapter for the company's infrastructure strategy.
Meta's Aggressive Infrastructure Play
The timing of the hires is notable. Meta has been aggressively expanding its AI infrastructure, recently signing a $21 billion expanded agreement with CoreWeave for AI cloud capacity through 2032. That deal brought Meta's total CoreWeave commitments to approximately $35 billion. Meta's AI-related capital expenditures for 2026 are projected between $115 billion and $135 billion — up roughly 60 to 87 percent from last year's approximately $72 billion.
Recruiting the architects of a rival's data center strategy gives Meta valuable institutional knowledge as it races to build out computing resources for its own AI ambitions, including the recently launched Muse Spark model.
What It Means
The departures strip Stargate of its foundational team and raise questions about continuity for OpenAI's remaining infrastructure projects. For Meta, the acquisitions strengthen an already aggressive push to dominate AI compute. The broader industry takeaway: as the economics of building custom AI data centers prove punishing, even the most well-funded labs are reconsidering whether owning the hardware is worth the cost.
