Sachin Katti Departs OpenAI Compute Role to Launch New Venture
The former Intel CTO and OpenAI executive concludes his tenure overseeing massive infrastructure expansion.
1Executive Transition Announced
Sachin Katti, OpenAI’s vice president of compute strategy and GPT-Infra, announced his departure on October 14, 2026. His final day is scheduled for November 15, 2026, though he will remain as an advisor. Katti previously served as chief technology officer at Intel before joining the OpenAI leadership team to manage compute operations.
2Infrastructure Scale and Scope
During his tenure, Katti oversaw the orchestration of massive AI compute infrastructure buildouts. His teams secured double-digit gigawatts of new capacity across projects in Ohio, Georgia, and Portsmouth. These efforts involved complex partnerships with industry leaders including NVIDIA, Broadcom, AMD, Oracle, Microsoft, Google, AWS, Cerebras, and the startup CoreWeave.
3Strategic Compute Initiatives
Katti’s remit covered commercial strategy, execution, and GPT-Infra product efforts. He focused on widening capacity access, improving deployment economics, and maximizing investment returns. He championed a recursive opportunity where OpenAI’s models help design and optimize the very infrastructure that future systems rely on, spanning silicon, datacenters, and energy supply chains.
4Project Camellia Expansion
OpenAI’s Project Camellia in Effingham County, Georgia, represents a major long-term datacenter development. The project contracts 3.2 gigawatts of power from Georgia Power, phased between 2028 and 2032. OpenAI committed to community benefits, including $80 million in local funding and $71 million in Codex credits for eligible students within the state.
5Pike County Development
The PORTS-Pike Technology Campus in Ohio aims to secure 8 gigawatts-IT of capacity by 2032. SB Energy will build and operate the site under a 20-year lease, with NVIDIA investing $1.5 billion. The facility will exclusively host NVIDIA AI compute infrastructure, creating thousands of construction and long-term operating jobs for the region.
6Custom Silicon Performance
OpenAI recently reported results for Jalapeño, its first custom inference chip. Tested on the InferenceX benchmark, the chip delivered 1.5 to 1.9 times more AI work per watt than comparison systems. It achieved 1.7 to 3.6 times lower latency across models like GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T.
7Multigenerational Roadmap
Deployment of the Jalapeño chip is planned for late 2026 as the first generation of a multigenerational roadmap. OpenAI is currently deep in development for Gen 2, while Gen 3 designs are already taking shape. This strategy pairs deep industry partnerships with custom infrastructure informed directly by OpenAI’s own production workloads.
8Future Industry Outlook
Katti believes the industry is still in its early innings, noting that AI is fundamentally reshaping the entire computing stack. His departure marks the end of a rapid buildout phase where GPT-Infra evolved from an idea to a functional business in three months. He plans to build something new next.
Original article: Unite.AI — Sachin Katti Departs OpenAI Compute Role to Launch New Venture