OpenAI has consolidated compute and infrastructure under Uday Ruddarraju, who led infrastructure at Elon Musk's xAI until July 2025. Sachin Katti, who oversaw infrastructure and the $30B Georgia build, will leave the role within weeks and remain as an adviser. The change matters because one manager now controls both computing capacity and data centre construction at a company planning very large capital spending.
Ruddarraju takes over compute and data centres
Bloomberg reported the management change on Wednesday. Ruddarraju joined OpenAI in July 2025 and became chief technology officer for compute this July, while Katti arrived from Intel late last year after serving as chief technology and AI officer. Business leaders reporting to Katti keep their jobs, and Ruddarraju said he expects a smooth handover. He added that Katti had maintained the commercial partnerships well and that he must now spend more time on the commercial side.
The appointment resolves an unnamed hire from August. When OpenAI confirmed that its data centre chief had left, it said a compute CTO had already been hired but did not disclose the name. That hire was Ruddarraju. At xAI, he worked on the Grok 3 training run across 100,000 GPUs. Katti said he was proud of what had been built, expressed confidence in Ruddarraju and the team, and said he looked forward to building something new.
Katti took over Stargate in March, according to Data Center Dynamics, and oversaw the $30B Georgia site, the first OpenAI designed and led itself. He also oversaw an 8-gigawatt project in Ohio with SB Energy and Nvidia, where Nvidia guaranteed up to $105B in lease and power payments, according to Axios citing a securities filing. Since July he had also run GPT-Infra, which applies OpenAI models to tasks such as semiconductor design. That effort now moves to chip chief Richard Ho, who said this month he does not fully trust the models' design changes.
What consolidation means for AI buyers
For companies buying AI capacity, a single owner of compute and infrastructure can shorten decisions on expansion, partnerships, and equipment priorities. Large customers with heavy training or inference needs may see clearer accountability for delivery dates and commercial terms. Smaller firms will feel the effect indirectly, through API stability, quota policy, and the pace at which new models reach production rather than through direct talks on data centres.
The limits are visible in staffing and geography. Peter Hoeschele, described as central to Stargate, left for Meta earlier this year with two colleagues, identified as Keith Heyde and Shamez Hemani. OpenAI paused Stargate UK in April, citing British electricity prices more than four times US levels and unresolved copyright rules. Its one announced European site is a 230-megawatt data centre in Norway from July 2025, while Ohio alone is about 35 times larger.
The marker to watch is progress on the US buildout against the $856B compute figure shown to investors. Georgia will test whether OpenAI can design and lead a $30B site itself, while Ohio will test the structure with SB Energy and Nvidia's payment guarantees. If construction milestones, power contracts, and leadership stability hold, the consolidated model is working for enterprise customers.
