Utilities & Grid Professionals
CEIgrid’s architecture treats the grid as a stability and planning problem, not a dumping ground for new load. Each 500 kW block is dispatchable and distributed — paired directly with compute — so the system can grow without creating new congestion hotspots.
Core grid idea
When compute is colocated with its own dispatchable generation, the grid sees a controlled interface: a predictable import/export profile rather than a large, sudden, concentrated load. The system can be operated “grid-light” — and still provide value to the grid when export is available.
Why distributed 500 kW blocks are different
- No single 100–500 MW step-load: reduces transformer, feeder, and substation stress
- Fewer congestion hotspots: load is geographically spread across rural corridors
- Faster interconnection: small blocks can use standardized studies and repeatable designs
- Dispatchable export: surplus power can be injected when compute demand is below capacity
“No spikes” export logic
Each POD can maintain headroom for compute spikes. When spikes are absent, the POD can export surplus power under a controlled ramp rate (setpoint + smoothing) so injection remains stable and non-disruptive. Think of it like distributed solar — but dispatchable and controllable.
Interconnection / protection (pilot-ready)
- Standardized interconnect package per POD (protection, metering, isolation)
- Defined ramp-rate limits and export caps per feeder
- Telemetry for state estimation and operational visibility
- Repeatable commissioning and compliance checks
What grid operators gain
- Predictable planning: modular growth vs. one mega-load event
- Reduced need for transmission upgrades driven by AI load
- Potential ancillary value (dispatchable export, local resilience)
- Operational clarity: PODs can be aggregated as a controllable fleet
Next step: pilot scope
A practical pilot starts with a small fleet (e.g., 10–50 PODs) on selected feeders to validate export behavior, ramp control, telemetry, and protection settings — then scales with repeatable templates.