OpenAI Compute Leadership: What Ruddarraju’s Move Means
OpenAI has put compute and data centre operations under Uday Ruddarraju. See what the leadership change means for infrastructure, investment and AI growth.
OpenAI's Infrastructure Bet Gets a New Leader
OpenAI's AI ambitions depend on more than model research. They hinge on getting enormous amounts of compute, reliable power, data center capacity, financing, and the right partnerships lined up. OpenAI has now put those pieces under one executive: Uday Ruddarraju. He previously led infrastructure at xAI and worked on Grok 3's training run across 100,000 GPUs.
This isn't just a reshuffle on paper. Ruddarraju will take responsibility for compute and data center operations as OpenAI pushes ahead with projects that require major power and capital. Sachin Katti - who had been overseeing infrastructure and led work including the $30 billion Georgia site - is stepping out of the operating role but will stay on as an adviser. The teams that reported to Katti are expected to remain.
So what changed, what Ruddarraju inherits, and why this appointment matters to building and running AI infrastructure? The limits of what we can know are real too: a new leader can align teams, but they can't erase permitting, financing, construction, or power-grid risk.
What changed in OpenAI's infrastructure leadership
OpenAI has consolidated its compute and infrastructure responsibilities under Uday Ruddarraju, according to Bloomberg. Ruddarraju told the publication he expects a smooth handover. Sachin Katti, who oversaw infrastructure, is leaving the operating role and staying with the company as an adviser. Business leaders who reported to Katti are expected to keep their roles.
That distinction matters. This is a leadership transition, not a wholesale restructuring of teams or projects. Keeping familiar operators in place can help preserve institutional knowledge even as decision-making shifts to someone with a different background. Still, a new executive moving to the center means new judgments about capacity, site priorities, and infrastructure timelines.
Katti came from Intel, where he was chief technology and AI officer, and took over the Stargate effort in March, Data Center Dynamics reports. He also led the Georgia facility, described as the first data center OpenAI designed and led itself. He oversaw an eight-gigawatt project in Ohio involving SB Energy and Nvidia. The source reports that Nvidia guaranteed up to $105 billion in lease and power payments for the Ohio project, citing an Axios report based on a securities filing.
Katti also ran GPT-Infra from July. The initiative uses OpenAI models for infrastructure-related work, including semiconductor design. That effort is moving to Richard Ho, OpenAI's chip chief. Ho has said he doesn't fully trust the models' design changes, a reminder that when you use AI in technical workflows, you still need human review and clear boundaries.
A handover with some continuity
Ruddarraju praised Katti's work maintaining commercial partnerships and said he'd need to spend more time on the commercial side. That points to the core reality of the role: infrastructure leadership isn't just about picking hardware or adding data center capacity. It's about coordinating partners, power arrangements, construction schedules, and the commercial terms that make a project viable.
Keeping existing business leaders in place may help reduce the risk of losing context midstream. At the same time, consolidating responsibility under a single executive can make accountability clearer. When compute capacity and data center development are tightly linked, separate leadership can also create friction over priorities, schedules, and who owns resource decisions.
The trade-off is simple. One reporting line can speed up coordination, but it can also concentrate operational responsibility. Whether this change helps will come down to execution: can teams make decisions quickly, surface risks early, and keep long-running projects moving through external dependencies?
Why Ruddarraju's experience matters
Ruddarraju joined OpenAI in July 2025 and became chief technology officer for compute in July 2026. Before that, he ran infrastructure at xAI until July 2025. At xAI, he worked on the Grok 3 training run across 100,000 GPUs, the source says.
That background is relevant because large-scale AI training and data center operations are tightly connected problems. A model team may want more compute, but compute has to be procured, housed, powered, and made available on a usable timeline. If one part of that chain slips, it can ripple through the entire training and deployment schedule. For an AI company, compute is a constraint on what you can train, serve, and build next - not an abstract spreadsheet line.
Running infrastructure at xAI doesn't automatically make every OpenAI project easier. Different organizations choose different technologies, rely on different partners, and face different delivery constraints. Still, someone who's worked on a major GPU training operation is likely to understand what capacity decisions really mean: how fast hardware can be deployed, how site readiness affects schedules, and what it takes to keep model work on track.
The commercial side is part of the technical job
Ruddarraju's emphasis on commercial work stands out. AI infrastructure is often treated like it's mainly about acquiring more chips. In practice, deployment depends on a chain of agreements and physical resources. Data center construction requires sites and power; multi-company projects depend on workable commercial arrangements; and the timing of each dependency affects everything that follows.
The Georgia and Ohio projects show how big those obligations can get. The headline figures reported alongside them are substantial, but they still need careful reading. An announced project, a financing or payment commitment, and an operating data center are not the same thing. Big numbers don't automatically mean the site is built, the capacity is online, or costs are already fully incurred.
So the question isn't only whether OpenAI can secure compute. It's whether commitments can turn into reliable capacity on a schedule that matches product and research needs. That means balancing speed with cost, flexibility with long-term commitments, and big capacity plans with real-world power and construction timelines.
Ruddarraju inherits that execution challenge. His background may help with operational coordination, but results still depend on teams, partners, and factors beyond any single executive's control.
The scale and concentration of OpenAI's infrastructure plans
The infrastructure portfolio described in the source is heavily concentrated in the United States. The Ohio project is an eight-gigawatt effort involving SB Energy and Nvidia. The Georgia site is a $30 billion build that OpenAI designed and led itself. By comparison, OpenAI's announced European data center in Norway is described as 230 megawatts. Based on those figures, Ohio's stated capacity is about 35 times Norway's.
That comparison is useful, but it's not a like-for-like measure of capacity that's already operating. These projects are at different stages, and a planned or announced capacity number shouldn't be treated as available compute. The numbers show the direction and scale of commitments - not a guarantee that power is delivered or infrastructure is usable today.
The geographic imbalance still matters. OpenAI paused its UK Stargate plans in April, citing British electricity prices that were more than four times US levels and unresolved copyright rules. The UK government called the decision disappointing. The Guardian later reported that OpenAI had never visited the north-east England site it had announced.
Those details underline that infrastructure decisions are shaped by local conditions. Electricity prices affect long-term operating economics. Regulatory uncertainty can complicate investment decisions. And site selection needs real, on-the-ground work - not just public announcements. Even a project that looks attractive on paper can fail to progress if the business case, power supply, or local operating conditions don't hold up.
A US-heavy strategy has benefits and risks
Putting more investment in the United States may give OpenAI access to a deeper pipeline of projects and commercial partners, but it also increases exposure to a narrower set of energy markets and regional constraints. At this scale, large data centers need substantial power, and coordinating that demand with utilities, construction schedules, and local infrastructure is a major part of the job.
The reported $856 billion compute bill shown to investors underscores how large OpenAI's plans are. The source says most of that spending is expected to happen across the Atlantic. But investor-facing projections aren't the same as money already spent. Forecasts and commitments point to ambition, while actual capital deployment and operating capacity depend on execution over time.
For investors and business partners, the key question is not just how much capacity OpenAI intends to build. It's how efficiently it can turn spending into compute that's actually available when it's needed. Capacity that arrives late, costs more than expected, or can't be powered as planned may not deliver the strategic value people assume.
For customers, the infrastructure strategy matters because compute availability can affect how fast AI services roll out and what they cost. But the connection from a construction project to a user experience is indirect. A leadership change alone can't predict model releases, pricing, or reliability. It's one piece of the operational system behind those outcomes.
What the leadership change could mean for execution
Putting compute and data center operations under a single leader gives OpenAI a clearer coordination point. That can help when capacity planning, project delivery, and commercial decisions need to move together. It may also help connect short-term compute needs with longer-term infrastructure commitments instead of treating them as separate planning tracks.
There's also a real downside to consolidation. If too many decisions funnel through one role, the organization can slow down. The answer isn't necessarily to split responsibilities again - it's to make decision rights clear and give teams authority to handle routine issues without escalating everything. With existing business leaders expected to stay in place, a practical test will be whether they can keep projects moving while aligning with Ruddarraju's priorities.
The questions to watch
The reporting available so far doesn't show that the transition will automatically speed up projects or change their scope. Those outcomes shouldn't be assumed. Instead, there are concrete indicators that can show whether the new structure is working:
- Project delivery: Do announced sites move through development and become operational on credible timelines?
- Power access: Can OpenAI and its partners secure the electricity required for planned capacity at workable costs?
- Commercial coordination: Are agreements among OpenAI, infrastructure partners, and suppliers translating into clear responsibilities and delivery milestones?
- Geographic balance: Does OpenAI remain heavily focused on US projects, or does it expand its international footprint?
- Use of capital: Do investor-facing compute plans correspond to capacity that's deployed and useful, or mainly ambitious projections?
- Operational accountability: Can teams maintain continuity while reporting to a new leader?
These measures tell you more than the appointment itself. Leadership changes can signal intent and assign responsibility - but delivery, cost, and usable capacity show whether the underlying system is functioning.
There's also a broader point for AI companies. Infrastructure is now a strategic capability, not a back-office service. As models get bigger and products demand more compute, teams have to treat hardware, power, facilities, and commercial relationships as part of product planning. Reserving compute without a realistic way to operate it isn't a strategy - it's a bet with dependencies you still have to solve.
OpenAI's move places an executive with large-scale infrastructure experience in charge of those dependencies. That sounds like a fit. Whether it turns out to be the right choice will come down to execution, not the résumé.
Conclusion
OpenAI's decision to put compute and data center operations under Uday Ruddarraju reflects how central infrastructure has become for AI companies. Ruddarraju brings experience from a large GPU training operation, while Sachin Katti stays on as an adviser and the business leaders below him are expected to remain.
This appointment makes it clear who's responsible for a demanding portfolio - but it doesn't solve the harder questions: whether projects can actually be delivered on schedule, whether power will be available at sustainable cost, and whether investor-facing plans turn into usable operating capacity. Those are the metrics that matter more than the org chart.
If you're tracking AI infrastructure, watch for real milestones: project progress, power arrangements, and capacity that's actually brought online - not just headline investment figures. For the original reporting and details behind this leadership change, read The Next Web's report.