Cutting time to power for AI.
Why we invested
Time-to-power has become the binding constraint on the AI buildout. Interconnecting new load to the grid in key U.S. markets now takes five to ten years, even as hyperscalers, AI labs, and the federal government race to bring gigawatts of new compute online. Building more transmission lines and substations cannot keep pace—and the grid that already exists runs at roughly 30% utilization on average, according to research from Stanford. GridCARE finds that stranded capacity and activates it. Sitting between hyperscalers and utilities, the company uses physics-based AI on detailed grid models to identify latent capacity and unlock it through managed flexibility: batteries, dynamic line ratings, and flexible interconnection agreements. The flagship deployment with Portland General Electric is bringing 80 MW online this year, scaling to 400 MW by 2029 and energizing customers up to eight years ahead of schedule. A second deployment with National Grid in New York is now live.
The founding team brings a rare combination of operating depth and academic leadership in this exact problem. CEO Amit Narayan previously built AutoGrid into an 8 GW global grid-flexibility platform across 20 countries before its acquisition by Schneider Electric. He is joined by three co-founders from Stanford: Dr. Ram Rajagopal, a leading AI-for-grid researcher, as CTO; Arun Majumdar, Dean of Stanford's Doerr School of Sustainability and former VP for Energy at Google; and Liang Min, Managing Director of Stanford's Bits & Watts Initiative. Together they pair the technical foundation with the utility, hyperscaler, and government relationships needed to define Power Acceleration as a new category for the AI era.
Why this work matters
AI is creating the largest demand shock the U.S. grid has seen in decades. Yet much of the grid we already have sits underused. Unlocking even a fraction of that latent capacity is the difference between AI workloads being built where the grid is ready versus delayed for years or pushed onto more carbon-intensive paths. GridCARE turns the existing grid into a faster, cheaper, cleaner foundation for the compute buildout—and creates the operating model that lets utilities and large loads coordinate at the speed the moment requires.