Energy intelligence for large sites
Hour-by-hour energy decisions for data centers, factories and producers
We put your meter data, bills and equipment next to market prices, weather and grid rules. You see where money leaks out of the energy bill, and you get an hourly plan for engines, batteries, flexible load and grid purchases.
- Read-only access to your data
- Your team approves every action
- Texas and Northern Italy
Replay: reference AI campus, Texas
16 May 2025
Gas engines70.0 MW, 7 of 7 on+7.0 MW cooling from heat
Site load63.8 MW served6.4 MW compute curtailed
Batteryidle, 0% charged10 MW / 20 MWh
The hub price reached $1,570/MWh at 19:00. Engines were already at full output, so the plan curtailed 6 MW of compute for that hour.
Prices: Source: ERCOT, hub average. Site, equipment, load and plan: modeled reference campus (80 MW IT, 60 MW grid connection, 7 x 10 MW gas engines, 10 MW / 20 MWh battery).
The platform
One screen for what your energy costs and what to do next
The console shows each site's consumption, on-site output, grid imports, prices and cost. Every figure says where it came from and whether it is measured or modeled.
01 See
Your energy position
Meters, bills, contracts and equipment in one view, next to ERCOT or Italian market prices, weather and the rules that apply to your site.
02 Plan
An hourly plan
When to run engines, charge or discharge the battery, curtail flexible load or buy from the grid. Solved again every hour against a 48-hour price forecast.
03 Act
Buy and sell
When buying power or certificates, or selling your output, makes sense, you describe what you need. We prepare the request and follow up with you.
Backtest
Tested on 476 days of real ERCOT prices
We replayed every day from 1 May 2025 to 23 August 2026, solving the plan again each hour, and compared the result with the operating rules sites use today. The prices are real. The sites are modeled.
Below the best common rule
2.5 to 10.3%
Energy cost including peak charges, five modeled sites
Days replayed
476
1 May 2025 to 23 Aug 2026
Plan horizon
48 h
Price forecast with P10 to P90 bands, solved again every hour
| Site (modeled) | International Batteries | Spark-spread rule | Baseload engines | Grid first | Saving vs rule |
|---|---|---|---|---|---|
| Reference AI campus, 100 MW | $32.53M | $33.75M | $36.50M | $42.10M | 3.6% |
| Reference steel mini-mill, 80 MW | $29.86M | $31.78M | $32.23M | $36.16M | 6.1% |
| West Texas AI campus A from public filings | $20.75M | $23.12M | $25.48M | $90.40M | 10.3% |
| West Texas AI campus B from public filings | $54.60M | $56.45M | $60.16M | $105.28M | 3.3% |
| West Texas AI campus C from public filings | $29.94M | $30.70M | $31.34M | $47.34M | 2.5% |
Energy cost over the 476 days, including ERCOT peak (4CP) transmission charges, with start and warm-up limits on every unit. The spark-spread rule runs engines when the grid price is above their fuel cost and is the best of the common rules we tested. Campuses A to C are modeled from public announcements with partly assumed equipment. They are priced at the hub; in practice they would see West zone prices. None of these figures are customer results.
Who it is for
Sites that use a lot of power, and the plants that make it
Each side gets its own view on the same data. Consumers see cost and the next operating choice. Producers see output, the price they capture and their certificates.
Data centers
Large, steady loads with backup engines and batteries
Avoid the summer peak charges, put idle backup equipment to work when it pays, and see what the next SB 6 order would mean for your site.
Data centersIndustry
Factories with engines, turbines or CHP
Decide each hour whether making power on site beats buying it, with heat demand, starts, permits and carbon in the same calculation.
IndustryProducers
Solar, wind, hydro, landfill gas and CHP plants
See what each megawatt-hour earns at the hour it is made, the discount your output shape costs you, and the value of your certificates.
ProducersTry it on your own data
A first look at your site from one meter file
Drop a 15-minute or hourly export from Smart Meter Texas, your utility portal or your building system. Your browser reads it and prices it against real ERCOT hours. Nothing is sent until you choose to send it.
- Consumption and peak. Total MWh, average and peak MW, load factor, and gaps in the data.
- Your load in ERCOT's summer peaks. The four hours that set next year's transmission charge, and what that charge comes to.
- Energy at real-time prices. Your hours priced at the 2025 hub price, hour by hour.
- The dearest hours. How much of your energy cost sits in the 50 most expensive hours of the year.
How we work
Advice you can check before you act on it
- Read-only, advisory
- We read your data. We do not switch engines, batteries or breakers. People on your site approve each action.
- A source on every number
- Each figure links to its source and says whether it is measured, modeled or estimated.
- Backtest before advice
- We replay your history against your current operating rule before we recommend anything new.
- Your data stays yours
- We use it to model your site and reply to you. Ask us and we delete it.
Get started
Send one meter file and your equipment list
That is enough for a first model of your site. A person on our team reads every submission and replies by email.