About Util

A compute scheduling software application designed around timing and location as controllable variables.

We built Util to lower costs and carbon emissions for flexible compute workloads. The application helps users understand the timing tradeoffs that shape both economic and environmental outcomes.

Positioning

What Util is

Util is a recommendation and scheduling layer for large compute workloads. It automatically decides when a workload should run based on changing electricity prices, carbon intensity, and location.

Pricing
Using live electricity price data, Util pushes workloads toward cleaner windows while reducing user costs.
Carbon intensity
Access to constantly updating emissions data and Util's carbon predictions helps users compute in a cleaner way.
Location
With some regions offering cleaner and cheaper energy, Util shifts workloads toward the most optimal machine.
Why it matters

Timing and location change both economic and carbon outcomes

A lot of workloads are deadline-sensitive but not start-time-sensitive. That difference creates a meaningful optimization opportunity.

Electricity cost changes hourly
Running at 1:00 PM and 3:00 AM can produce meaningfully different cost outcomes.
Clean energy infrastructure
The same workload can land against very different emissions conditions depending on where it is run.
Flexibility is often already present
Training runs, overnight jobs, render pipelines, research simulations, and ETL workflows are often movable without harming delivery.

System overview

A visual summary of how Util evaluates workload timing and turns changing energy signals into clearer scheduling decisions.

Util system overview
Audience

Who Util is for

Util is for people who want a practical way to reason about when compute should happen without becoming energy-market specialists.

Independent researchers with overnight workloads
Teams running GPU or CPU batch jobs with flexible deadlines
Operators who care about both spend and grid impact
Long-term vision

From recommendations now to more automated systems later

The long-term direction is to make timing-aware compute a normal part of how flexible workloads get planned and launched.

From suggestions to automation
Start with recommendations users can inspect, then move toward increasingly automated scheduling where appropriate.
From single jobs to systems
Over time, the product can expand from individual workloads to fleets, regions, and more integrated execution environments.