An update to the power burn model in Inside the SynMax US Gas Demand Forecast. The full method is in the methodology note SynMax Electric Power Long Term Forecast: Methodology 

Executive Summary

Our previous power burn model extended a trend fitted to EIA history. The next two years will be driven by things a trend cannot see: data centers connecting, new solar and wind, and a diminished coal fleet competing with gas at forward prices. The new model builds burn from those drivers, region by region, daily, two years ahead.

It puts US power burn at about 37.6 Bcf/d over September 2027 to August 2028, up from about 36.6 Bcf/d over the last twelve months of EIA data. Load growth adds about 4.5 Bcf/d. New solar takes away about 2.4 Bcf/d and new wind about 0.7 Bcf/d. Gas winning share from coal adds about 0.6 Bcf/d.

Load growth is the most difficult part to forecast. Recent utilities' filings assume growth well above what most regions have delivered, much of it from data centers. Vulcan, SynMax's satellite tracking of power projects, shows how much is actually being built: about 55 GW of data-center capacity under construction across the Lower 48, against about 338 GW announced with no ground broken. That evidence sets how much of each region's filed growth we use.

How the Forecast Is Built

1. Demand and supply. For each region we forecast total electricity generation, then subtract what nuclear, hydro, wind, solar, storage, and other sources will supply. The rest must come from non-nuclear thermal generation (thermal generation for short).
2. The gas ceiling. We convert that thermal generation into gas at the fleet's heat rate: the most gas the region could burn if every required thermal megawatt-hour came from gas.
3. Gas against coal. A daily model sets how much of that ceiling gas captures, from weather, the calendar, and the cost of coal against gas.

Load Growth and Vulcan

We start from each region's weather-adjusted electricity demand over the last twelve months and grow it along the utilities' own forecasts: those filed with FERC, plus ERCOT's for Texas. However, the filings over-predict. Utilities have folded large new loads, mostly data centers, into their plans. Many of those projects are early and uncertain, so we check each filing against what is being built as tracked by Vulcan. Vulcan reviews about 2,700 US data-center sites by satellite, recording when land is cleared, when structures go up, and when construction finishes.

Renewables Growth

Wind and solar are forecast as capacity times typical output for the time of year, with announced projects discounted by how often such projects have been built on schedule. On that basis, solar capacity rises from about 170 GW to about 264 GW over the forecast, led by South Central, and wind from about 167 GW to about 185 GW. Nuclear and hydro follow their typical pattern of the last ten years, and smaller sources follow recent history.

Gas Versus Coal

Gas's daily share of thermal generation responds to weather, the calendar, and fuel prices. Plant by plant, we compare the cost of a megawatt-hour from coal with the cost from gas at the Henry Hub strip, so a move in the curve passes straight into the forecast.

With the strip $1/MMBtu lower, power burn is about 0.9 Bcf/d higher on average over the two years; with it $1 higher, about 0.8 Bcf/d lower.

Performance

We ran fair backtests, scoring forecasts against out-of-sample actuals, as of March 31 in 2024, 2025, and 2026. Each run used only the plant data, generator inventory, coal fleet, and fuel prices known on that date, with the weather that actually happened and on the utilities' filings as they then stood.

Monthly burn typically missed measured burn by 1.1 to 1.4 Bcf/d, much of which can be attributed to the difference between modeled and realized gas prices. On this year's filings the 2026 run is high, putting July to September 2026 at 48.6 Bcf/d against 47.6 Bcf/d measured. That is why we check the filings against what Vulcan sees being built rather than take them at face value.

Regional Views

South Central: strong load growth, much of it round-the-clock, largely offset by the largest solar build-out.
East: data-center growth in PJM, with a coal fleet that still competes when gas prices rise.
Midwest: most coal is still cheaper than gas, so its burn moves with the strip.
Mountain: the fastest load growth rate, partly offset by solar.
Pacific: almost no coal. Its risk is hydro, and its spring burn is squeezed by solar and snowmelt together.

What’s Next

Next, we are linking Vulcan’s site-by-site construction stages directly to each utility’s filing, and we will add more past-date re-runs as each new season of data arrives. Contact us at support@synmax.com for the regional settings, the re-runs, or the full methodology.

Data sources: EIA-923 plant-level generation and fuel use; EIA-860M monthly generator inventory; FERC Form 714 utility load forecasts; ERCOT Long-Term Load Forecast; Vulcan satellite-verified power project and data-center tracking; NOAA GEFS ensemble weather forecasts; the Henry Hub forward curve and regional basis; the EIA Short-Term Energy Outlook coal price path.