The head of infrastructure at Elon Musk's xAI has departed.

Uday Ruddarraju spent a little over a year at the company, which also includes social media platform X.

Update: Ruddarraju has joined OpenAI.

"After an unforgettable ride, I’ve decided to move on from xAI and yesterday was my last day," Ruddarraju said on LinkedIn.

"When I first joined, I thought everyone was absolutely nuts for thinking we could deploy 100K GPUs in four months, especially without a fully functioning site. Watching us go and double that and most importantly successfully train Grok 3 made me incredibly proud... and very happy to be wrong."

At the company, Ruddarraju said that he led the infrastructure teams across compute, networking, storage, data platform, production infrastructure, research infrastructure, and data center back office for both xAI and X.

He was a member of the Grok 3 training team that ran training at scale across 100,000 GPUs, "the largest pre-training job known to date." The Colossus supercomputer has "more than 250,000 GPUs," he claimed.

Colossus was deployed in record time. To power the data center, xAI turned to gas turbines, deploying a reported 35 - above the permit for 15. The NAACP and the Southern Environmental Law Center (SELC) have filed a notice to sue over the air pollution.

At the same time, Musk has said that he plans to import a power plant from abroad for a new Memphis data center that could host up to one million GPUs.

"Jensen Huang was right, Elon and his teams are singular in what they can achieve," Ruddarraju said. "Grateful to have played a small part in shaping the future of AI Compute from the inside."

Prior to his time at xAI, Ruddarraju spent six years at Robinhood, most recently as head of infrastructure. He spent another six years at eBay as technical lead, cloud and platforms.

Former Robinhood traffic infrastructure engineer Wenlong Xiong joined xAI in April, as have other Robinhood employees like chief architect Mike Dalton, who has also joined OpenAI. Former eBay engineering manager Sreekanth Pothanis joined in November as team lead for supercompute, responsible for compute infra, storage, and traffic for research and inference.

A number of employees have also moved over from other Musk ventures. Former SpaceX Starship offshore program senior engineer Brent Mayo was appointed senior manager of site builds, while Liz Balke, Tesla's Dojo data center infrastructure lead, joined xAI as a mechanical engineer.

The company raised $10 billion in debt and equity this month, partly to fund the development of additional data center capacity. xAI is also using Oracle's compute capacity to train its next-generation Grok models and to run inference workloads.