- Reorganization: OpenAI has split its Stargate computing initiative into three groups covering technical design, commercial partnerships, and facility management.
- Leadership: Former Intel AI chief Sachin Katti is among the new leaders appointed to oversee the reorganized infrastructure teams.
- Strategy Shift: OpenAI has abandoned plans to build its own data centers and now rents capacity from cloud providers including AWS and Google Cloud.
- Spending Target: The company is targeting roughly $600 billion in total compute spending through 2030, down from an earlier $1.4 trillion figure.
OpenAI has appointed new leaders to oversee its Stargate computing initiative, splitting operations into three groups after moving away from building its own data centers in favor of renting capacity from cloud providers.
According to The Information, citing two people familiar with the matter, the reorganization divides compute operations into three groups covering technical design, commercial partnerships with cloud providers and chip firms, and on-the-ground facility management. OpenAI’s infrastructure ambitions have shifted considerably since Stargate was announced in January 2025, and the company now relies on rented server capacity rather than owned facilities.
As a result, OpenAI can control design and commercial relationships without bearing the capital burden of owning physical facilities. For an AI company that has struggled to secure construction financing, this structure creates flexibility to work with multiple partners simultaneously.
How the Reorganization Works
In practice, OpenAI divided its computing effort into three groups: one handling technical design of data centers, another managing commercial partnerships with cloud providers and chip firms, and a third overseeing on-the-ground facility operations. Sachin Katti, who left Intel in November 2025 where he served as chief technology and AI officer, is among those appointed to lead the reorganized infrastructure teams.
Alongside the leadership changes, OpenAI walked away from an Oracle Stargate expansion in Abilene, Texas earlier this month after prolonged negotiations complicated by financing challenges and shifting demand forecasts, according to The Information. OpenAI now prefers newer Nvidia chips at fresh sites that can be built to current specifications.
Furthermore, walking away from Abilene while appointing new infrastructure leaders signals that OpenAI is resetting its compute strategy rather than patching an existing one. Katti’s experience leading Intel’s AI strategy positions him to evaluate chip-level decisions as OpenAI shifts toward next-generation hardware.
Prior Coverage and Context
WinBuzzer has tracked OpenAI’s infrastructure evolution extensively. Stargate was announced in January 2025 as a large-scale data center venture with SoftBank and Oracle. By mid-2025, however, the project stumbled, prompting OpenAI to quickly diversify its cloud partnerships and pursue urgent fundraising.
Consequently, Stargate’s trajectory from ambitious buildout to stalled joint venture to hybrid rental model illustrates the persistent tension between AI companies’ growing compute appetite and the capital intensity of building and operating physical infrastructure at scale. Each successive pivot has forced OpenAI to cede more control to outside partners.
From Building to Renting
Yet the depth of the stall was telling. More than a year after Stargate was announced, the Stargate data center project ran into serious trouble without hiring any staff or actively developing data centers, according to The Information. OpenAI tried to build its own facilities but could not secure financing, with lenders unwilling to back billion-dollar construction projects from a company still carrying heavy operating losses.
In response, OpenAI cut deals with AWS, Google Cloud, AMD, and Cerebras to cover its computing needs. CEO Sam Altman has framed the company’s partnership with Nvidia as central to its next phase, and OpenAI is now redirecting compute toward Nvidia’s Vera Rubin platform, targeting the first gigawatt of capacity in the second half of 2026.
Meanwhile, diversifying across multiple cloud providers and chip suppliers reduces OpenAI’s dependency on any single partner, but it also means the company now competes for AI customers against the same firms supplying its infrastructure. The reorganization is part of a broader hardware transformation that has seen OpenAI issue RFPs for custom hardware manufacturing.
According to investor materials, OpenAI is targeting roughly $600 billion in total compute spending through 2030, down from an earlier $1.4 trillion figure. Stargate’s total planned capacity still stands at nearly 7 gigawatts across all locations, representing over $400 billion in investment over three years.
Under a compromise with SoftBank, OpenAI will sign long-term leases and control facility design while SoftBank Energy develops and owns the projects. The arrangement lets OpenAI shape its infrastructure without financing construction itself.
Great Article – what’s not clear is implications to Oracle who is OpenAI’s key partner in the Stargate Joint Venture.