DRUPS
Data Center & FacilitiesA Diesel Rotary Uninterruptible Power Supply: a rotary system that stores energy in a spinning mass to hold power steady and ride through grid disturbances without a break in supply. Storing that bridge as kinetic energy rather than in chemical batteries also lightens the environmental load: there are no banks of lead or lithium cells to manufacture, cool, and replace on a cycle, and none of the hazardous waste they leave behind.
Eleveight AI keeps its DRUPS units in a dedicated building, where they condition the cluster's power and bridge grid interruptions; diesel generators, not the DRUPS, are the sustained backup.
Overview
A DRUPS keeps a heavy mass spinning and uses that stored momentum as an instant reserve of power. It combines the rotating energy store, a motor-generator, and a diesel engine into one continuity system. In normal operation, it cleans the incoming grid supply and keeps the mass at speed; the value of that spinning reserve appears the instant the grid falters, when it can carry the load with no gap at all while longer-term backup spins up.
How it works
Because the reserve is kinetic rather than chemical, a DRUPS switches over with no delay, unlike a static UPS that must transfer to batteries. Momentum converts straight back into electricity the moment mains power dips or drops, covering the few seconds a generator needs to start and take the load. For anything beyond those seconds, diesel generation carries the outage while the DRUPS continues to bridge and condition what passes through it.
Why it matters
GPU clusters do not tolerate power interruptions gracefully: a single dip can kill a training run that has been going for days and stress hardware that is expensive to replace. The DRUPS exists so that grid instability never reaches the compute floor. It is best understood as an energy-storage and ride-through layer, the thing that guarantees continuity in the moment, rather than as the backup that carries a long outage, which is the generators' job.
Use cases
- Uninterruptible power continuity for GPU clusters
- Ride-through during grid dips and switchover
- Power conditioning against voltage and frequency instability
- Bridging to diesel backup during extended outages