How it works
Every year, for 25 years, the model asks one question: what’s the cheapest mix of power sources that still keeps the lights on, every hour?
These six sources compete every year — the model picks how much of each to build.
Where the data comes from
The model is grounded in six years of real US electricity data (2020–2025) — about 4.1 million hourly readings: six sources, every hour, across 13 regions, over six years.
The knobs you can turn
Four big levers, plus per-source cost dials, shape every run:
A price on carbon pollution that climbs every year, making gas and coal steadily more expensive.
A price per death — coal's air pollution and accidents cost lives. It works exactly like the CO₂ price, on a different harm.
How much more electricity the region needs each year.
The cost of borrowing to build new plants — this can make or break nuclear, which takes decades to pay off.
The carbon and mortality prices work the same way — two harms the market ignores, each given a dollar figure. How mortality pricing works.
Why price mortality?
Every source of power costs lives — mining and drilling accidents, and far more, the air pollution that shortens lives downwind. The market never charges for it, and the sources are wildly unequal: per unit of energy, coal is more than 1,000× deadlier than solar — three orders of magnitude.
deaths per TWh — central estimate
That gap is why a mortality price hits coal hardest while barely touching solar or wind. See the full risk ladder →
What happens every simulated year
Every source loses a slice of its capacity each year — older, longer-lived plants (like nuclear) lose less than younger ones.
The region needs more electricity than last year, by a set growth rate.
For each source, it picks how much new capacity to add — whatever combination keeps costs lowest.
Cheapest power goes first. Leftover power charges the battery; shortfalls draw it down. Anything still unmet is an outage.
Building, running, and any outages all cost money. The optimizer keeps adjusting what is built until it finds the cheapest hours over one year.
Want the math behind each step? See the methodology page.
How fast can the grid actually change?
The optimizer can’t rebuild the grid overnight. Each source can grow by only so much of its fleet per year — so change is bounded by how fast we can actually build, which is why results play out over years, not all at once.
max new capacity added per year, as a share of the fleet — based on historical build rates
A few terms you'll see
MW is how big a power plant is, the maximum power it can possibly produce. MWh is how much energy it actually produces. A big plant that rarely runs still makes little MWh.
How often a source actually runs, as a percent of the time. Nuclear: almost always. Solar: only when the sun's out.
Electricity demand that went unmet. Keeping the cost for this high keeps Outages near zero.
Ready to explore? Check out the library of real runs, or compare two scenarios side by side.