OpenAI's Head of Computing Sachin Katti to Depart Next Month
Sachin Katti, the head of computing capacity at OpenAI, will leave the company next month after approximately one year. He will transition to a consulting role after ending his full-time work on November 15, with Uday Ruddarraju taking over the overall leadership of the computing organization.
Katti stated on LinkedIn that he decided to leave to build new things, noting that the team has secured double-digit gigawatts of new computing capacity in Ohio, Portsmouth, Georgia, and projects like Trainium, with partners including NVIDIA, Broadcom, AMD, Oracle, Microsoft, Google, AWS, Cerebras, and CoreWeave. He also initiated GPT-Infra to apply OpenAI models to semiconductor and infrastructure-related tasks.
Katti joined OpenAI at the end of 2025, having previously served as Chief Technology and AI Officer at Intel for only about six months. His tasks included restructuring the slow-moving Stargate data center project and securing additional computing agreements with Amazon Web Services. The computing organization is responsible for planning the infrastructure needed for model training and operations, aiming for computing expenditures of about $600 billion to $750 billion by 2030.
Uday Ruddarraju joined OpenAI in July 2025, previously leading infrastructure engineering at xAI, and was promoted to Chief Technology Officer of computing this summer, reporting directly to President Greg Brockman. Katti expressed full confidence in Ruddarraju and the team. This year, OpenAI has seen product and business head Fidji Simo transition to a consultant role, along with the departures of Brad Lightcap and revenue head Denise Dresser.
Katti's direct leasing data center strategy resulted in a 10-gigawatt lease with SB Energy in August, backed by NVIDIA's financial guarantee. The Ohio project is approximately 8 gigawatts, while the Georgia project is the first site designed and led by OpenAI. The computing organization is experiencing continuous leadership changes, while the pace of investment continues to accelerate.
Computing capacity is shifting from third-party cloud leasing to direct data center leasing and long-term capital expenditures. Funds are flowing to suppliers like NVIDIA, Oracle, AWS, and self-built sites. Katti's departure will maintain some continuity in his consulting role, with Ruddarraju taking over commercial and technical execution. Beneficiaries of the infrastructure expansion include chip and data center suppliers, while frequent leadership changes put pressure on execution continuity.
Source: Public Information
ABAB AI Insight
Sachin Katti joined OpenAI in November 2025 after leaving Intel, where he had only been for about six months overseeing AI-related matters. During his time at OpenAI, he pushed for adjustments in the Stargate project and secured agreements with multiple cloud and chip suppliers, while also launching GPT-Infra to apply models to infrastructure design. This year, OpenAI has seen several executives leave or transition to consulting roles, including Fidji Simo, Brad Lightcap, and Denise Dresser.
Katti shifted the computing organization from reliance on third-party leasing to direct leasing of data center footprints, securing multi-gigawatt capacity. Uday Ruddarraju, who previously led infrastructure at xAI, now takes over overall computing leadership, reporting to Greg Brockman. Resource mobilization is reflected in project collaborations with NVIDIA and SB Energy, as well as feedback of some model capabilities into chip and system design.
Similar rapid executive turnover has occurred in early OpenAI and other AI labs' infrastructure teams. OpenAI is currently in a phase of expanding computing expenditures from tens of billions to hundreds of billions, with the infrastructure unit transitioning from project initiation to large-scale capital deployment and parallel construction across multiple locations.
This essentially represents a restructuring of the industry chain. AI labs are expanding computing from pure cloud leasing to self-built and long-term leased data centers, while bringing in former chip company executives to oversee commercial and technical integration. The mechanism is to shift capital expenditures for model training and inference from variable costs to long-term fixed investments through direct control of data center assets and supply chain agreements, while maintaining advisory continuity amid leadership changes to reduce execution disruptions.
ABAB News · Cognitive Law
Computing expansion relies on capital commitments, execution relies on leadership continuity
From cloud leasing to self-built is a permanent shift in cost structure
As executive turnover accelerates, advisory roles become a buffer layer