Energy-aware orchestration

Compute follows energy.

Sarv schedules work when power is cheaper, cleaner, and locally available. Energy isn't the product — it's the intelligence underneath everything the Grid does, improving the economics and the footprint of every job.

How scheduling works

The same job, done at the right time, in the right place.

Delay-tolerant compute has a superpower: it can wait. The Grid's scheduler reads live signals from every Vault and from the energy system, then places work where it costs least — in money and in carbon.

Price

Agile tariffs swing through the day. Work runs when electricity is cheapest — sometimes nearly free overnight.

Carbon

Grid carbon intensity varies hour to hour. The scheduler prefers low-carbon windows automatically.

Local generation

Vaults with solar or home batteries run jobs on their own clean power first — compute behind the meter.

Grid friendliness

A network that can pause or ramp thousands of small loads is, at scale, a flexible asset the energy system values.

What a Vault sees

An energy dashboard, built in.

Every Vault understands its own energy context — generation, storage, carbon, price — and the owner can see exactly when and why it works. Nothing about your Vault's schedule is a black box.

Illustrative example of a live Vault readout:

Energy signalExample*
Solar generationHigh
Battery73%
Grid carbonLow
Price8.4p / kWh
Scheduler decisionRUN

* Illustrative. The pilot measures real electricity cost per job — one of the numbers that decides whether the Grid's economics work. See the experiment →

Why it matters

AI's energy problem won't be solved by building more of the same.

~950 TWh
Projected data-centre electricity demand by 2030. Centralised AI scales its footprint with its ambitions.
Behind
the meter
Where Vaults run — using grid connections and local generation that already exist, instead of queuing for new ones.
Flexible
by default
Thousands of schedulable small loads can duck peak demand — the opposite of a data centre's always-on draw.

We're careful not to overclaim: a distributed network is not automatically greener — inefficient devices could make it worse. That's precisely why energy-aware orchestration sits at the core of the design, and why true electricity cost per job is a pilot measurement, not an assumption.

Join the network

Smarter for the planet.
Cheaper for everyone.

Join the Network