The fourth compute leader in a year
OpenAI has named Uday Ruddarraju to lead its infrastructure and compute businesses, including the Stargate data center program, according to Bloomberg's reporting on the move this week. He replaces Sachin Katti, who joined OpenAI from Intel in November 2025 as head of AI and took over the Stargate effort in March of this year. Katti is expected to remain involved with the organization only in an advisory capacity going forward, a soft landing that suggests the exit was managed rather than contentious, though OpenAI has not explained the reasoning publicly.
Ruddarraju is not new to the organization he now runs. He joined OpenAI last July as head of compute and infrastructure and was named CTO of Compute that same month, giving him over a year of internal context before this latest promotion. Before OpenAI, he led infrastructure engineering at the SpaceX-adjacent business after an earlier stint at xAI, Elon Musk's AI company. The promotion consolidates authority in someone with hands-on data center build and operations experience rather than a chip or silicon design background, which is itself a signal about what OpenAI believes it is currently short on.
A pattern, not an incident
Katti's exit follows a string of data center leadership departures across OpenAI this year, including Peter Hoeschele, Keith Heyde, Shamez Hemani, and Chris Malone, the company's former head of data centers who left after roughly 17 months on the job earlier in the fall. Each departure has arrived alongside its own internal reorganization, and each reorganization has left the formal reporting lines for infrastructure slightly different than the arrangement that came before it, making the org chart a moving target for anyone trying to track accountability.
That cadence matters more than any single name on the list. A company that has changed the leader of its largest cost center four or more times within roughly a year is telling observers something important about how difficult that job has become, not simply that individual executives happened to move on for unrelated personal reasons. For outside parties watching closely, the frequency of the reshuffles is the more useful signal here, not the credentials or pedigree of whoever happens to be holding the title this particular quarter.
The spending number quietly got bigger
Alongside the leadership change, OpenAI's projected compute spending through 2030 has climbed to roughly 750 billion dollars, up meaningfully from an earlier estimate near 600 billion dollars reported earlier this year. That increase happened while the team responsible for actually delivering the underlying infrastructure was being rebuilt for at least the third time within the same twelve-month stretch. Spending commitments and delivery capability are clearly not moving in lockstep here, and the widening gap between the two is precisely where operational and financial risk tends to concentrate fastest.
For anyone negotiating GPU, power, or colocation capacity in the same markets OpenAI is actively buying into, this disconnect is directly relevant to pricing and availability forecasts. A buyer that keeps raising its own spending target while its delivery organization keeps turning over leadership is a buyer whose near-term demand is genuinely hard to forecast with any confidence, and that unpredictability makes capacity markets around it structurally harder to price for everyone else competing for the same scarce supply of power and chips.
Why Ruddarraju is stocking the team with xAI alumni
Ruddarraju has brought in several former xAI and SpaceX colleagues as part of the reshuffle, including Brent Mayo to head data center build and delivery, along with site leads, infrastructure staff, and project managers drawn from that same operational background. That is a deliberate and fairly specific bet on a team known for shipping data centers quickly under Musk's companies, even though the methods, culture, and risk tolerance at xAI differ substantially from OpenAI's existing engineering organization and its public commitments to partners.
The subtext here, confirmed directly by the reporting around this move, is that OpenAI's own in-house data center projects have been running behind schedule, pushing the company to lean more heavily on external cloud partners to cover near-term demand gaps. Importing a build team with a reputation for speed is one way to try closing that gap internally, but it also means OpenAI is now running two distinct infrastructure strategies in parallel at once: accelerating in-house construction while simultaneously increasing reliance on third-party leased capacity.
What this means for anyone capacity-planning near OpenAI
If your roadmap assumes stable GPU or power availability in markets where OpenAI operates as a major tenant or anchor customer, this reshuffle is worth tracking as a leading indicator rather than dismissing it as background industry noise. Leadership churn at this frequency inside a buyer this large tends to precede changes in sourcing strategy, construction timeline slippage, or sudden swings in how aggressively that buyer competes for scarce capacity in a given region once new leadership sets its own priorities.
The practical move for your own organization is to diversify supplier relationships rather than assume any single hyperscaler-adjacent demand signal will remain stable through your planning horizon. OpenAI's spending target keeps climbing and its delivery team keeps changing hands at a pace that would alarm most enterprise boards. Plan your own capacity commitments assuming both of those trends continue for the foreseeable future, and build real contractual flexibility into your cloud and colocation agreements rather than betting on a market that stays calm around the industry's largest and most volatile single buyer of compute.
