Three OpenAI leaders from Stargate project moved to Meta during AI race

Three top OpenAI managers who previously worked on the Stargate project—a program to build data centers for artificial intelligence costing hundreds of billions of dollars—have moved to work at Meta Platforms (Meta is recognized as an extremist organization and its activities are banned in Russia). The transition coincided with a sharp increase in Meta's spending on AI infrastructure and a revision of some Stargate projects within OpenAI itself, reports 3DNews.

What Is the Stargate Project

Stargate is a large-scale infrastructure program by OpenAI to build data centers specifically designed for training and operating advanced AI models, with a stated budget in the hundreds of billions of dollars. It is precisely such projects—and not the models themselves—that today determine which company is capable of training next-generation models faster than competitors: access to computational resources has become as strategic a resource as access to data once was. Managers responsible for implementing such a program from within OpenAI possess rare expertise—from negotiations with energy and equipment suppliers to managing multibillion-dollar capital expenditures at specific construction sites.

The coincidence of several significant personnel departures from one program is rarely coincidental for a company of OpenAI's scale: such waves typically result from either dissatisfaction with the pace or conditions of implementing specific projects, or deliberate poaching by competitors ready to offer more attractive terms precisely to specialists who have already proven their effectiveness on comparably complex tasks.

In recent years, the AI industry is often described as a time of unprecedented talent competition: major technology companies offer specialists with experience in cutting-edge AI projects compensation packages that previously were the norm only for top executives of public companies. The transition of specifically infrastructure managers—not just model researchers—shows that this competition has also spread to those responsible for the physical side of the AI industry: power supply, construction and operation of data centers.

• Who—three OpenAI managers who participated in the Stargate project.
• New employer—Meta Platforms.
• Stargate project—a program to build AI data centers costing hundreds of billions of dollars.
• Coinciding events—Meta's increased spending on AI infrastructure and revision of some Stargate projects within OpenAI.
• Source—3DNews.

What Lies Behind the Reshuffles

The transition of specifically infrastructure managers, rather than model researchers, is telling: it indicates that competition between the largest AI companies is increasingly shifting to the plane of physical infrastructure—construction and operation of data centers, rather than just algorithm development. The coincidence of the transition with the revision of some Stargate projects within OpenAI can be interpreted two ways: either the company itself adjusted priorities for certain facilities, or the outflow of precisely these specialists became one of the reasons why some plans had to be revised.

Public details about which specific positions the specialists transferred to and which specific Stargate projects were revised are not disclosed in the article, however the very fact of the simultaneous coincidence of these two events—the outflow of talent and revision of infrastructure plans—already becomes a subject of discussion in the industry, where each such movement of top managers is traditionally viewed as an indicator of the company's internal priorities.

What This Means for the AI Infrastructure Race

Massive investments in own data centers have become for the largest AI companies a way to insure themselves against the shortage of computational resources, which in recent years has constrained the pace of training new models for all market players. Poaching experienced managers of such programs from a direct competitor allows the new company to reduce the time it takes to build its own expertise from scratch—instead of learning from its own mistakes, it receives a ready team that has already gone through this path within one of the most ambitious infrastructure projects in the industry. For OpenAI, in turn, the loss of precisely these personnel at a time of revising Stargate plans may mean additional difficulties in implementing the remainder of the program.

For the labor market in AI infrastructure, such transitions also have a demonstration effect: they signal to other specialists in the field that experience participating in megaprojects like Stargate converts to offers from direct competitors, and thus increase the already high cost of such expertise on the labor market in general.