GridCARE Raises $64 Million to Cut AI Data Center Grid Connection Waits From Years to Months
- Data Centers Grid Connection & Reliability News
Marin Holt
- 0
- 7 minutes read
- Series A: $64 million
- Lead investor: Sutter Hill Ventures
- First capacity (Portland General Electric): 80 MW in 2026, scaling to 400 MW by 2029
- Identified pipeline: 2 GW across 12 markets
GridCARE, a Stanford-founded startup that uses physics-based AI to find unused capacity on the existing power grid, has closed an oversubscribed $64 million Series A led by Sutter Hill Ventures. The May 2026 round, backed by Kleiner Perkins partner John Doerr, National Grid Partners and Future Energy Ventures, follows a $13.5 million seed in 2025 and goes after the biggest bottleneck in the AI buildout: the multi-year wait to connect a new data center to the grid. The company says its software shortens that wait from years to months, and it has already proven the approach with Portland General Electric, identifying 80 MW of capacity available near Hillsboro, Oregon in 2026 and a path to more than 400 MW by 2029.
How It Works
GridCARE’s Energize platform reads billions of data points (utility planning models, interconnection queues, permits, rate schedules and extreme-weather records) to map where the grid has headroom that standard interconnection studies miss. Instead of waiting for a utility to build new transmission, developers use that map to site data centers where power is already available. DataCenterDynamics reported the system evaluates grid conditions in real time to compress interconnection timelines that routinely run three to seven years in congested regions such as PJM.
Early Proof Points
Beyond the Portland General Electric work, GridCARE identified 650 MW of connection capacity on National Grid’s New York network for large, flexible loads. The company now reports a pipeline of 2 GW of new AI compute across 12 markets and says it has unlocked more than $10 billion in economic value for data center developers by bringing capacity online ahead of schedule. The founding team carries deep grid credentials: CEO Amit Narayan earlier built AutoGrid, which Schneider Electric acquired; co-founder Arun Majumdar is a former director of ARPA-E and Google energy lead; and CTO Ram Rajagopal and co-founder Liang Min both come out of Stanford’s grid-research labs.
Critical Perspective
GridCARE’s $64 million Series A is a bet that software can find spare headroom on a grid that data centers are already straining, with a first 80 MW slated for 2026 on Portland General Electric and a path to 400 MW by 2029. The harder constraint is physical, not informational: Lawrence Berkeley National Laboratory counts more than 2,600 GW of generation and storage stuck in U.S. interconnection queues, with a typical project now waiting about five years from request to operation. Identifying latent capacity does nothing if the transformers, substations, and transmission upgrades needed to use it carry their own multi-year lead times and supply-chain backlogs. Portland General Electric has itself flagged load growth outrunning its system, so the 80 MW GridCARE aims to place there competes with the utility’s own queue. The company points to a 2 GW pipeline across 12 markets, but a pipeline is not energized megawatts. When the spare capacity GridCARE identifies still requires hardware upgrades that take years to build, how much of that 2 GW can actually be connected on the timeline its data-center customers expect?
Why It Matters
For AI operators the binding constraint is no longer chips or capital; it is grid access. Interconnection queues now hold years of backlog, and new transmission takes a decade to permit and build. Software that surfaces existing headroom is faster and far cheaper than steel in the ground, and it shifts the advantage from a utility’s construction schedule to data-driven capacity discovery. If GridCARE’s years-to-months result holds across its 12 markets, it changes where the next wave of data centers gets built and how fast the grid absorbs them, without the rate-base cost increases that large new loads usually push onto existing customers.
