02 · ENTROPY
Entropy®
Physics-informed energy orchestration.
Entropy makes complex power solutions possible, turning a multi-source generation stack into a reliable, grid-interactive asset at gigawatt scale.
- PI-MARL
- SUB-SECOND CLOSED LOOP
- DIFFERENTIABLE DIGITAL TWIN
- ANY GENERATION MIX
- GIGAWATT SCALE
One Number: Power Usage Effectiveness
POWER USAGE EFFECTIVENESS · THE OPERATOR STANDARD
Entropy orchestrates end-to-end energy infrastructure across any generation mix (grid, natural gas, solar, wind, fuel cells, battery storage) against the volatile demand profile of AI workloads. It makes sites viable that a conventional energy approach could never support.
Built on physics-informed multi-agent reinforcement learning (PI-MARL), Entropy is the operational intelligence layer behind one number: PUE (power usage effectiveness). StrataIQ's site designs, including generation mix and battery-storage configuration, are its foundation from day one.
It enters production observationally and graduates to active orchestration as verified value accumulates.
ORCHESTRATES
- Grid
- Natural gas
- Solar
- Wind
- Fuel cells
- Battery storage
Signals in, coordinated action out.
Grid frequency, market price, compute load, operator commands and weather stream in as one live picture. The core forecasts 90 seconds ahead, the agents pick the action, and the constraint manifold checks it against hard physical and contractual bounds before anything moves.
What comes out is one coordinated response across every lever (battery dispatch, generator setpoints, grid bids, cooling setpoints, compute shaping), not five separate ones. Follow a signal through the loop below: each inbound value prints in the ledger with the actions it produced, on its channel.
INITIALIZING CONTROL LOOP…
SCRIPTED SEQUENCE: A FIXED, DETERMINISTIC EVENT LEDGER THAT PLAYS IDENTICALLY EVERY LOAD. REPRESENTATIVE OF LOOP BEHAVIOUR; NOT LIVE SITE TELEMETRY.
01
HOURS → DAYS
Market
Market position and grid obligations, planned against forecast demand.
02
MINUTES
Dispatch
Generation mix and battery storage committed against live AI workload demand.
03
SUB-SECOND
Asset control
Unit-level setpoints across the fleet, in a closed sub-second loop.
Each layer is bounded by a differentiable digital twin: the agents can only commit actions the physics permits, which is what lets a learned controller run a gigawatt-scale asset at all.
Every AI factory has an energy problem.
It begins the day the factory goes live: volatile workload demand against a complex generation mix, grid obligations, and cost pressure. Entropy turns that problem into an advantage: driving PUE, making sites viable that conventional energy approaches couldn't support, and turning factories into intelligent, grid-interactive assets.
Entropy doesn't just optimize an existing power setup. It makes sites viable that would otherwise never get built.
CONTROL LOOP
SUB-SEC
ASSET-LEVEL SETPOINTS
CONTROL LAYERS
3
MARKET · DISPATCH · ASSET
GENERATION MIX
ANY
6 SOURCE CLASSES
OPTIMIZES
PUE
POWER USAGE EFFECTIVENESS
INTERESTED?