HyperBase is developing the American Macrogrid, an AI‑orchestrated network of distributed energy assets that clears interconnection queues for hyperscale compute. By using a physics‑informed intelligence system to dispatch generation, storage, and load in real time, it aims to reduce the typical five‑to‑seven‑year wait for new capacity, with an active pipeline of over 0 GW of projects.
Funding
Funding not disclosed
Founders
Product
Problem
U.S. interconnection queues hold over 2.6 TW of generation and storage projects, creating multi‑year delays for new power connections. This bottleneck limits the ability of hyperscale data centers and AI workloads to secure reliable, on‑demand electricity.
Solution
HyperBase builds an AI‑orchestrated American Macrogrid that aggregates distributed generation, battery and thermal storage, dispatchable load, and hyperscale compute assets into a single coordinated network. Their three‑layer, physics‑informed intelligence system—named Entropy—observes real‑time asset states, models dispatch under AC power‑flow and operational constraints, and issues set‑points that match compute demand with available clean energy. By treating the distributed fleet as a virtual power plant, the platform clears the interconnection queue, delivering power to AI workloads with minimal latency. The solution includes a digital twin that visualizes the clearing loop and provides telemetry for each component, enabling operators to optimize FLOPs per megawatt‑hour. HyperBase licenses the technology per capacity unit, allowing hyperscale customers to scale orchestration as they expand.
Target Audience
Primary customers are hyperscale data center operators and AI compute providers that require large, reliable power supplies, as well as utilities and large industrial firms seeking to monetize distributed energy resources.
Features
- Three‑layer intelligence (observation, physics‑based modeling, dispatch) that runs a top‑down/bottom‑up clearing loop
- Multi‑agent, physics‑informed reinforcement learning (PI‑MARL) that respects AC power flow, ramp limits, and state‑of‑charge constraints
- Real‑time coordination of solar, wind, thermal, gas generation, battery and thermal storage, and dispatchable industrial load
- Digital twin interface showing live telemetry, utilization metrics, and clearing loop status for each asset
- Patent‑pending architecture that integrates distributed assets into a virtual power plant for hyperscale compute
- Licensing model based on orchestrated capacity (per 100 MW) to align cost with scale