OpenAI Computing Chief Kati Departs, Rudaraju Takes Over Compute Strategy
nashnova research
OpenAI's head of compute capacity Sachin Katti will leave next month after roughly a year; former xAI executive Uday Ruddarraju takes over as the company raises its pre-2030 compute spending forecast to roughly $750 billion, putting infrastructure-team stability under scrutiny.
Who is leaving and who is stepping in?
Sachin Katti will depart OpenAI next month and stay on as an adviser. He served about one year; before OpenAI he was a Stanford professor and Intel executive.
His successor, Uday Ruddarraju, previously worked at Elon Musk's xAI. He was named OpenAI's compute CTO this summer and will report to president Greg Brockman.
This means → command of OpenAI's compute strategy shifts from an academic-turned-executive to someone who came from a direct competitor and is already embedded in the engineering operation.
What did Katti accomplish in one year?
His central mission was advancing the Stargate data-center program. The project had a rocky start — OpenAI failed to reach a single package deal with SoftBank and Oracle, and ultimately signed separate agreements with each.
He also secured additional compute contracts with Amazon Web Services (AWS) and pushed the company to restart building its own data centers rather than relying solely on leased capacity.
In August he closed a 10-gigawatt data-center lease with SB Energy, backed by a financial guarantee from Nvidia. In plain terms = 10 gigawatts is roughly the output of seven nuclear plants — an enormous lock-up of power.
Can the handover go smoothly?
Ruddarraju said he has been deeply involved in the infrastructure team's engineering, construction, and technical work, and has built working relationships with partner executives including Nvidia CEO Jensen Huang.
Katti's strategy and business leads will remain in their roles to maintain continuity.
The "GPT-Infra" project Katti led — using OpenAI models to aid semiconductor and infrastructure R&D — will move under chip-business head Richard Ho. This means → the reshuffle is not just a leadership swap but also a functional reorganization.
With $750 billion on the line, why does the team keep turning over?
OpenAI has raised its pre-2030 compute spending forecast from roughly $600 billion to roughly $750 billion, while competing fiercely with Anthropic for users; both companies are preparing IPOs.
Yet infrastructure-team turnover is a pattern: in August, data-center executive Chris Malone left; earlier, Stargate infrastructure lead Peter Hoeschele and two colleagues departed for Meta.
This reflects a deeper tension: spending is accelerating, but the team responsible for deploying that capital keeps changing hands — the money is moving faster, yet the people steering it keep rotating out.
市场有风险，内容仅供研究参考，不构成投资建议。