About the data

Map data: EIA-923

The map shows electricity generation by state using Form EIA-923, the U.S. Energy Information Administration’s monthly power plant survey. EIA collects, checks, and publishes the data with a lag of roughly one to two months. The SURVEYED · <month> label identifies the latest month included.

“Other” includes fuels outside the six named categories, such as petroleum, geothermal, biomass, and waste heat. It also includes pumped-storage hydro, which can report negative generation when pumping water uphill uses more electricity than the plant generates. We keep negative values in the data and display them as zero in the charts.

The map shows electricity generated within each state. States also buy and sell power across their borders, so local generation does not tell you where all the electricity used in a state comes from. Washington D.C., for example, imports nearly all its power from the PJM grid. Those imports and exports are outside the scope of this map.

Solar includes rooftop systems and utility-scale solar farms. EIA estimates small-scale distributed solar output, known as DPV, because it does not meter every rooftop. The solar total therefore combines reported generation with estimates. The survey is available at eia.gov/electricity/data/eia923.

Historical trends

The Trends page shows monthly generation from EIA-923, with one line or band per fuel. The Rings and Long Run views are historical records only. The Smoothed + Projection view uses twelve-month averages of those records and then continues past the last survey month, so part of that chart is a projection rather than a record.

On that chart a ruled line marks where the survey ends. Everything to the left of it is surveyed generation. Everything to the right is drawn in lighter ink, under a DRAFTED · FROM ANNOUNCED PROJECTS label, using the same calculation the Atlas year slider uses — see How projections work below. Each projection year is placed at its December and the months between two years are drawn as a straight line between them, because a twelve-month average rises gradually as a plant starts running rather than jumping in one month.

The charts end at the latest published survey month. The most recent month or two is usually unavailable while EIA collects and checks the data.

Why the line starts in 2014

EIA’s monthly state-level generation data goes back to 2001. We start in 2014 so the solar figures include rooftop generation throughout the chart.

EIA collects generation data from utility-scale solar farms and estimates output from small rooftop and other distributed systems. It first published monthly state estimates for distributed solar in the December 2015 Electric Power Monthly, with data back to January 2014.

A chart starting in 2001 would show a jump in solar generation in January 2014 because rooftop solar was added to the data. Starting in 2014 avoids that break in coverage.

The solar line includes both rooftop solar and solar farms. Part of the total is EIA’s estimate of distributed generation.

How we calculate “Other”

We subtract generation from gas, coal, nuclear, wind, solar, and hydro from EIA’s total to calculate “Other.” This avoids double-counting the overlapping categories in EIA’s source data.

“Other” includes petroleum, geothermal, biomass, waste heat, and pumped-storage hydro. Pumped storage uses electricity to move water uphill, then releases it to generate power. If it consumes more power than it generates in a month, its reported generation is negative. That can bring the “Other” total below zero. We retain the negative value in the data and show zero on the chart.

The national total adds together the fifty states and the District of Columbia using the same records as the state charts.

How projections work

The year slider projects generation through 2031 using additions and retirements listed in Form EIA-860M. Starting with each state’s latest twelve months of generation, we add planned generation when a project’s scheduled year arrives. The calculation is megawatts × 8,760 hours × capacity factor × completion factor, converted to gigawatt-hours. We subtract planned retirements without a completion discount and keep each fuel’s total at or above zero. This calculation does not model electricity demand, weather, or prices.

Projected years carry the label DRAFTED · FROM ANNOUNCED PROJECTS. Projects that leave EIA’s planned list stop contributing to the next projection. We retain their records.

Completion factors

Projects can be delayed or dropped. We multiply each addition’s estimated generation by a completion factor based on its EIA-860M status. An unknown status uses the lowest factor, 0.40. These factors are assumptions used in the projection.

Typical capacity factors

A capacity factor compares actual generation with what a plant could produce running at full capacity all year. The table lists our national defaults. For nuclear, coal, wind, and hydro, we use a state-specific factor when the data passes the checks below.

State-specific factors. For nuclear, coal, wind, and hydro, we divide the state’s last twelve months of generation by the output its operating capacity could produce at full power over a year. Generation comes from EIA-923 and capacity from EIA-860M. If the data does not pass our checks, we use the national default. Each calculated factor records its sources and survey period.

Fuels that use national defaults. The source data does not support comparable state-specific factors for these categories:

• Solar. Generation includes estimated rooftop output, while the capacity inventory covers utility-scale plants. Using these together would overstate a state’s capacity factor.
• Natural gas. Combined-cycle plants and peaking plants share a fuel category in the generation data but operate at different rates. Their national defaults are 0.55 and 0.12, respectively.
• Other fuels. Geothermal, biomass, and petroleum share a generation category, although their capacity factors differ. We use technology-specific national defaults.

We use national defaults where a state has too little operating capacity for a fuel to give a representative result. State-specific factors also compare current capacity with the previous twelve months of generation. Recent additions can lower the calculated factor because they have not operated for a full year. We discard results above 1.0, which can indicate that the generation and capacity data cover different equipment.

Battery projects appear on the drawing board but do not add to projected generation: they store electricity already counted when it was generated. We also calculate a scenario with no completion discounts for comparison, but the charts show only the discounted projection.

Each state’s drawing board shows estimated annual generation from announced additions by 2031, alongside the amount included after completion discounts. These two figures show how much the completion assumptions affect the projection.

Project stages and capacity

Each state’s drawing board and the national board group projects by EIA-860M status: announced, permitted (approvals received or in progress), under construction, and testing. Standing means the generator appears in the operating inventory.

The board combines two construction stages. EIA distinguishes projects that are at most half complete from those further along. We use different completion factors for these stages, 0.85 and 0.92, and display both as “under construction.”

The board shows capacity in megawatts (MW). Generation is the electricity produced over time, measured here in gigawatt-hours (GWh). Plants with the same capacity can generate different amounts depending on the technology and how often they run. Batteries count toward capacity on the board and in the operating inventory, but do not add to projected generation.

A missing project does not confirm a cancellation. When a generator leaves EIA’s planned list, we check for the same identifier in the operating inventory. A match lets us mark it as standing. Without a match, we cannot tell why it left. The site therefore does not report cancellation or attrition rates.

Project date history begins on August 12, 2026. Before then, each weekly update replaced the previous record. We now keep those updates so we can compare planned dates over time. We need roughly six months of records before reporting delays. Completed projects are labeled with the period we have tracked.

The Spread

The Spread asks how each of seven traded power hubs compared with one national gas-cost benchmark for a reference gas plant. Explore The Spread →

For each hub and calendar month:

• Spark spread ($/MWh) = power price − heat rate × gas cost
• Breakeven gas price ($/MMBtu) = power price ÷ heat rate

A positive spread means that month’s power price paid more than the reference plant’s fuel at the national benchmark cost. The page labels this GRIDLINES ANALYSIS · <month>: it is our arithmetic over published numbers, neither a survey nor a projection.

Where each number comes from

• Power. Daily weighted-average on-peak prices for the next trading day at Mass Hub, PJM West, Indiana Hub, Mid-C, NP-15, SP-15 and Palo Verde, collected by the Intercontinental Exchange (ICE) and republished, with permission, by EIA on its wholesale electricity page. We average each calendar month’s prices by the day they were traded, drop any month with fewer than 10 trading days, and use a hub’s month only once that hub has traded in a later month. Power traded on a month’s last trading day is usually delivered on the first day of the next. We publish only those monthly averages and the spreads built from them, never the daily prices. This product features derived information and is not an official feed of ICE.
• Gas. The U.S. electric-power sector’s average cost of natural gas delivered to generators, from Form EIA-923 (EIA’s electric power operational data: U.S. total, electric power sector, natural gas, cost per MMBtu). It is weighted by the gas received, not burned, and includes delivery. EIA publishes the same U.S. figure in Electric Power Monthly Table 4.13.A. Every hub is compared with this one national number on purpose; it is not a local or gas-hub price.
• Heat rate. 7.548 MMBtu per MWh for a natural-gas combined cycle plant, from EIA Electric Power Annual Table 8.2 (data year 2024, released 2025-10-16). That table gives 10.999 MMBtu per MWh for a gas-turbine peaker; only the combined-cycle line is drawn. These are tested full-load rates averaged across the fleet, not operating rates of any plant.

Power and gas are only ever paired within the same calendar month. If that month’s gas cost is missing, withheld or zero, or the hub traded on too few days, the month is left as a gap rather than estimated.

Two hubs are much sparser than the rest. ICE’s daily on-peak trading at Indiana Hub and at NP-15 thinned sharply after 2018, so most of their months since then have fewer than 10 trading days and are left out rather than estimated. A lone month that clears the rule is drawn as a dot; no line is drawn across the gaps on either side of it.

What it does not tell you

1. This is a fuel margin, not a profit. It leaves out operating and maintenance cost, capital cost, capacity payments and ancillary revenue.
2. On-peak power only. These are next-trading-day on-peak power prices; nights and weekends are excluded from the power leg. The gas cost is a monthly U.S. survey average of fuel received by electric-power generators, including delivery. It is not a traded gas-hub price or the cost of gas burned in that month.
3. No carbon cost. Generators in RGGI states and California also buy allowances, which this leaves out.
4. A reference plant, not a real one. The heat rate is an EIA fleet average for the technology, not any particular plant’s.
5. A hub is a market, not the nation. Each power price is local; the fuel benchmark is the same national average for all seven hubs. Their different footprints make this a comparison, not an individual plant margin or an estimate of each hub’s local gas cost.
6. A common month only. EIA-923 fuel costs arrive later than power trades. A spread appears only after both inputs cover that month; the page names the latest common month and never combines different months.

News sources and ranking

The Clippings page gathers headlines and links to the original publishers. We store the headline, the description supplied by the feed, and the link, along with tags and AI-generated analysis described below. We do not fetch or reproduce full articles.

We follow roughly three dozen feeds covering natural gas, nuclear, hydro, coal, renewables, wholesale markets, grid operations, and independent analysis.

Source labels

Each source has a label describing its role: official for statistical agencies and regulators, analysis, academic, trade, independent, industry, or advocacy. This “lens” appears beside the source name to help you understand its perspective. It does not establish whether an article is accurate.

Source type affects the default ranking. Statistical agencies and research organizations receive a higher weight than trade publications; advocacy and industry sources receive a lower weight. Across the full range, this weighting has about the same effect as two days of recency.

How we use AI in the news feed

A language model uses each headline and feed description to write a short summary, update fuel and state tags, and assess headline exaggeration and detail. It also flags press releases, opinion, and sponsored content. The model does not read the full article. Summaries labeled in brief are AI-generated by Gridlines Energy.

AI ratings can be wrong. They assess how a headline and description are written, including the use of specific figures, projects, filings, and dates. They do not fact-check an article or establish whether its claims are true.

Ratings affect the order of the feed. They do not remove articles. Sponsored posts remain available and labeled, and articles awaiting analysis receive an average score. Choose Newest to sort by publication date without these weights.

Coverage and limitations

• Survey dates show the period covered by the data. Live grid conditions are outside the site’s scope.
• Plant maps show locations available in EIA’s records. Missing coordinates and map display limits are noted beside the map.
• Projections use announced projects and the assumptions listed above. They do not model demand, dispatch, prices, or weather.
• Energy figures, map patterns, and state summaries are calculated directly from EIA data using fixed rules. AI is used for news summaries, tags, and ranking only.
• News headlines link to the original publishers. Full articles remain on their sites.

Support Gridlines Energy

Voluntary tips help cover hosting, the database, and AI processing for the news feed. The Support link in the header leads to ways to help, including our Ko-fi page. The site is free to use, and tipping does not unlock accounts or extra features.

Ko-fi is a separate service with its own privacy practices. The tip link opens Ko-fi; we do not embed its scripts or widgets on this site.

Privacy

No accounts, no login, no cookies, and no tracking of any kind run on this site today. If we add analytics, we will limit them to cookieless page counts, with no fingerprinting, cross-site tracking, or third-party advertising or analytics scripts.