OpenAI’s Computing Chief Sachin Katti Steps Down; Uday Ruddarraju Takes Over

By Fatima Raza · Technology Desk · Updated October 15, 2026

OpenAI is losing the man in charge of its computing capacity. Sachin Katti, who heads the unit that secures the data-center power and chips behind ChatGPT, is leaving after roughly a year in the role, with his last day set for November 15, 2026, after which he will stay on as an adviser. Taking his place is Uday Ruddarraju, a former xAI executive who joined OpenAI in 2025 and will now run both compute and infrastructure, reporting to OpenAI President Greg Brockman. Employees were informed on Wednesday, October 14.

Who Sachin Katti is and why this exit matters

Katti was not a figurehead. When he joined OpenAI in late 2025, the company handed him the single most important operational problem in the AI industry: finding enough electricity and chips to train and serve the next generation of models. In plain terms, OpenAI’s ambition had outrun the available data-center capacity in the United States, and Katti’s job was to close that gap. He did it by signing a string of compute partnerships with practically everyone who sells computing power: AWS, Nvidia, Broadcom, AMD, Oracle, Microsoft, Google, Cerebras and CoreWeave.

The headline achievement of his tenure came in August 2026: a ten-gigawatt lease agreement with SB Energy, reportedly backed by a financial guarantee from Nvidia. To put that in perspective, a single gigawatt can power roughly 750,000 homes; ten gigawatts is the scale of a medium-sized country’s entire power demand. It is the kind of deal that defines an era, and it was Katti’s signature.

His departure follows that of another executive who had been overseeing the data-center build-out, who left over the summer. Two senior exits in quick succession from the same function is not a coincidence worth ignoring; it signals that the team building OpenAI’s physical infrastructure is being reshaped at exactly the moment the spending is about to explode.

Uday Ruddarraju: the xAI veteran taking over

Ruddarraju is an interesting pick because he comes from the other side of the AI talent war. He was an executive at xAI, Elon Musk’s AI company, before joining OpenAI in 2025. By the summer of 2026 he had become OpenAI’s CTO for compute, and now he consolidates both compute and infrastructure under one leader, reporting directly to President Greg Brockman.

That consolidation is the real story here. Under Katti, compute capacity was effectively its own empire; folding it together with infrastructure under Ruddarraju suggests OpenAI wants a single pair of hands on everything from chip supply deals to the actual buildings they go into. I have watched enough corporate restructurings to know that “combined org” announcements like this usually mean the company wants faster decisions and fewer turf wars, and given the numbers below, speed is exactly what OpenAI needs.

The deals Katti leaves behind

Katti’s team secured double-digit gigawatts of computing capacity across multiple vendors, a portfolio that reads like a who’s who of the chip and cloud industry. The partnerships with Nvidia, Broadcom and AMD cover the silicon side; AWS, Oracle, Microsoft and Google cover cloud capacity; Cerebras and CoreWeave are the specialist neocloud players that have become critical to AI labs. It is a deliberately diversified supplier base, which matters because no single company, not even Nvidia, can supply everything OpenAI needs.

The structure of these deals is also telling. The SB Energy lease came with an Nvidia financial guarantee, which is an unusual arrangement: the chipmaker is effectively underwriting the data-center builder’s risk because it wants OpenAI’s demand locked in for its own chips. These are not simple purchase orders; they are multi-party financial structures, and they tell you how much capital is now tied up in the AI build-out.

Why this is happening: the $750 billion race

Context matters more than the personnel news. OpenAI has raised its projected compute spending to roughly $750 billion through 2030, up from an earlier figure of around $600 billion. That is a staggering sum, larger than the GDP of most countries, and it puts OpenAI in a direct spending war with Anthropic. Both companies are also reportedly preparing for IPOs, which means every infrastructure decision is now being made with public-market scrutiny in mind.

When a company is about to spend three-quarters of a trillion dollars on computing, the person managing that spend is one of the most important hires in the world. Katti leaving now, just as the spending curve goes vertical, raises an obvious question: did he leave because the job was done, or because the next phase needed a different kind of leader? OpenAI’s answer, consolidating the function under Ruddarraju, suggests the latter: the deal-making phase is giving way to the build-and-operate phase.

What it means for Stargate

Stargate, OpenAI’s flagship data-center project, was Katti’s original brief when he joined. The project is the physical expression of that $750 billion ambition: massive data centers built specifically for training frontier AI models. With Katti out and another build-out executive gone since the summer, Stargate now rests on Ruddarraju’s shoulders. The deals are signed and the capacity is secured; the hard part, actually building and filling those data centers, is what remains.

For readers in Pakistan, this story matters in a practical way. The global compute race is what drives the price and availability of AI services everywhere, including the tools Pakistani developers and businesses use daily. When OpenAI secures ten more gigawatts, that is capacity that keeps ChatGPT fast and affordable; when the build-out stalls, everyone downstream feels it. And with the compute race heating up between OpenAI and Anthropic, the next two years of AI progress will be decided as much by power contracts as by algorithms.

Also read on TimelyUps

• Technology news on TimelyUps — our latest coverage of AI, chips, phones and the companies shaping the digital world.

The bottom line: OpenAI’s biggest construction project just changed foremen, and with $750 billion on the line, the next phase will be about building, not buying.