Three senior infrastructure heavyweights who helped shape OpenAI's Stargate data center plans have walked out the door and are set to land at Meta Platforms, turning up the heat in the Bay Area's AI arms race. The trio - Peter Hoeschele, Shamez Hemani and Anuj Saharan - are expected to help build a new Meta compute organization focused on scaling data centers and tightening supplier partnerships. Their exits underline how the real fight in AI is increasingly about hardware and talent, not just clever model code.
The Information first reported the executives' departures and outlined their new roles at Meta. Coverage republished by Business Standard adds that Meta declined to comment and that the OpenAI employees did not respond to requests for comment.
Meta's model push and big spending
Meta has started rolling out a new model while signaling record infrastructure spending as it races to support larger AI workloads. Meta's newsroom introduced Muse Spark on W,ednesday positioning it as part of a broader push under Meta Superintelligence Labs. The company has told investors it expects 2026 capital expenditures in the range of $115 billion to $135 billion to fund those labs and related data center projects, a jaw-dropping budget even by Big Tech standards. In its latest earnings release, Meta frames this spending as central to both its product roadmap and its underlying infrastructure plans for the year.
OpenAI pauses some builds and reshapes teams
OpenAI, meanwhile, is tapping the brakes on parts of its own Stargate rollout and reshuffling how it manages big compute projects. Bloomberg reported that OpenAI is "pausing its Stargate artificial intelligence infrastructure project in the UK," citing rising energy costs and regulatory uncertainty as the key reasons. Industry reporting also indicates that OpenAI is leaning more heavily on cloud partnerships and has recently tapped Sachin Katti to lead its industrial compute efforts, according to Data Center Dynamics.
Why the shuffle matters
Control of physical compute - the sites, suppliers and power deals that actually keep these models running - has become one of the core battlegrounds of the AI era. Leaders who know how to execute at a massive scale help shorten the path from research prototype to production system. Reporting indicates the new Meta compute unit will sit alongside Alexandr Wang's Superintelligence Labs and report to leaders responsible for both technical design and strategic supplier relationships. As The Information notes, this round of hiring is about locking in operational know-how as much as it is about adding marquee research talent.
For Bay Area engineers and vendors, the shakeup likely translates into hotter demand for specialized build-and-deploy skills and more pressure on local permitting, power and real estate markets as companies scramble to secure capacity. OpenAI and several outlets have reported that the company thanked the departing staff and emphasized it remains focused on its broader infrastructure goals while shifting some plans toward cloud partnerships and more disciplined spending, per coverage in Business Standard and others.
