GridCARE closed an oversubscribed $64 million Series A on May 14, led by Sutter Hill Ventures with participation from John Doerr, National Grid Partners, Future Energy Ventures, Emerson Collective, and Stanford University. The Stanford Doerr School spinout, founded by CEO Amit Narayan, is selling a new category it calls "Power Acceleration" — a physics-based AI layer that finds dormant capacity on the existing grid and routes it to AI data center developers in months rather than the typical six-to-ten years.
The round is roughly five times the size of GridCARE's $13.5 million seed from May 2025, and it lands at a moment when power, not silicon, is the binding constraint on hyperscaler buildouts.
The number behind the round
GridCARE says its Energize platform has already unlocked more than 2 gigawatts of AI data center capacity across 12-plus markets and roughly $10 billion in economic value for developers. The mechanism: ingest billions of data points from utility interconnection queues, permitting workflows, rate cases, and weather feeds, then simulate what SVP of Business Development Alaina Bookstein described as "a quadrillion scenarios" against grid physics to surface conservative-dispatch headroom that operators rarely monetize.
Stanford's own analysis, cited by the company, puts U.S. grid utilization at about 30 percent. The other 70 percent sits behind reliability margins that physics-aware modeling can chip away at without new poles and wires.
The Portland test case
The most concrete proof point is a deal announced with Portland General Electric in October 2025. GridCARE structured flexible commitments that will release 80 MW of capacity to AI data center developers in 2026, with a path to more than 400 MW in Hillsboro, Oregon. The model swaps firm interconnection requests for time-shifted, curtailable load — which utilities can underwrite far faster than greenfield substations.
Why this round matters now
The five largest hyperscalers have committed north of $660 billion of 2026 capex, but timelines for power delivery are pushing site selection from a land-first to a power-first decision. Oracle, Meta, and the OpenAI-Stargate ecosystem have all signaled multi-gigawatt power gaps in announced builds. CoreWeave, Crusoe, and IREN have publicly tied capacity expansions to interconnection availability rather than chip allocations.
GridCARE's pitch is that the cheapest gigawatt is the one already sitting in the existing grid. If the Energize model holds outside Hillsboro — and Sutter Hill plus a National Grid affiliate clearly think it does — the strategic question for AI infra teams shifts from "where can we build?" to "which utility footprint will let our model unlock the most stranded capacity?"
What changes for builders
For enterprise AI buyers negotiating colocation contracts, GridCARE's footprint is a near-term signal of which markets have hidden runway. For utilities, it's a template that turns flexibility into bookable revenue without waiting on FERC. And for any operator pricing a multi-year AI factory PPA today, a credible Power Acceleration vendor is now a line item the CFO has to evaluate.
