EIA Electric Power Annual

It is the yearbook the entire US power industry gets measured against once a year. The Electric Power Annual consolidates the Energy Information Administration's plant, generator and utility surveys into roughly 130 tables across twelve numbered sections, each edition pairing the latest complete data year with the prior one - the current edition sets 2024 against 2023. Its scope runs the whole value chain: national summary figures, fourteen tables of retail sales, revenue, prices and customer counts by sector and state, net generation tables 3.1.A through 3.27 by fuel and owner type, fourteen summer-capability tables including planned additions and retirements, fuel consumption, stocks, receipts and delivered cost, system performance with capacity factors, SO2/NOx/CO2 emissions, advanced metering and demand response, SAIDI/SAIFI/CAIDI reliability indices, and a territories section covering Puerto Rico, Guam, American Samoa, the Northern Mariana Islands and the Virgin Islands.

For anyone sizing the Independent Power Producers & Energy Traders market specifically, the decisive feature is the ownership dimension: electric utilities, independent power producers (nonregulated), commercial and industrial generators are broken out separately in the generation and capacity tables. One verified example from Table 1.1 shows the shape - coal utility-scale generation of 652,156 thousand MWh in 2024 against 675,115 in 2023, a 3.4 percent decline, with the Electric Power Sector subset reading 508,149. Get a sample of this dataset and we return rows shaped exactly like the dictionary below.

Rows read as dimension-plus-paired-values. From Table 1.1, national net generation by fuel:

Table 1.1            Coal                                         652156
Table 1.1            Petroleum Liquids                             11456
Table 1.1            Coal (EP Sector)                             508149

Three things to notice in the anatomy rather than the values. First, the paired-year layout: every summary row carries its current and prior edition side by side with a computed change column, so a year-over-year comparison needs no self-join. Second, the sector split arrives in the same table - the same fuel row repeats under Total (All Sectors) and again under the Electric Power Sector, which is exactly where IPP-versus-utility share gets computed. Third, the units are declared in the table header itself (thousand MWh here, cents/kWh in the price tables, tons in the emissions tables), so typed ingestion never guesses. Swap Coal for any of the twelve fuel categories - natural gas, nuclear, wind, solar PV, geothermal, biomass - and the identical row shape hands you that slice.

Seven documented dimensions reconstruct nearly every table. Fuel / Energy Source names what burned, split or shone - coal, petroleum liquids and coke, natural gas, other gases, nuclear, hydroelectric, wind, biomass, geothermal, solar PV and solar thermal. Sector carries the ownership cut that isolates IPP activity from utility activity, the distinction this industry page exists around. Paired year columns and their percentage change give every row its built-in trend, while the state and census-division dimension powers the geographic tables. Definitions above are verified against the published record.

Additional fields on request: the section-by-section table inventory, delivered coal cost and heat content by state and census division, capacity factors by producer type, advanced-metering penetration and the SAIDI/SAIFI/CAIDI reliability indices. Generator-level detail - one row per unit with nameplate specifics - deliberately lives in the companion EIA Form EIA-860 Annual Electric Generator Inventory; the two are designed to be joined.

Three chips summarize the footprint:

• Geography: national totals plus all fifty states and census divisions, and - unusually for commercial power data - the five populated US territories, whose grids get their own section.
• Temporal: annual editions archived back to 1994, two-volume through 2000 and consolidated since 2001, with multi-year trend tables inside each edition. Three decades of annual history on consistent definitions.
• Granularity: annual, cut by ownership sector, fuel, technology, state and facility type. It aggregates rather than itemizes - the plant-level grain sits in the Form 860/923 companions.

Set against the wider catalog - 1,744 datasets averaging 7.81 - this record scores 9/10, carried by breadth across twelve sections, three decades of archive depth and verified field definitions. Its one structural limit is cadence: an annual volume answers structural questions, not this-week questions, and the monthly lane belongs to the EIA Electric Power Monthly.

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel your team already works in and set the cadence to match the decision you are feeding - overnight flat files sized for research benches, a direct pipe into Snowflake, BigQuery or Redshift, or lookup calls for anything interactive. Cadence changes are a settings conversation, not a re-integration project.

Every delivery ships the field dictionary above unchanged plus sample rows for validation, flattened to one observation per table-row-sector-year so a decade-over-decade comparison is a filter rather than a spreadsheet merge. Name the sections, states and years you care about and the sample comes back shaped to them before any commitment.

• Power market investors and IPP strategists size the nonutility generation base directly from the ownership-sector split, tracking whether merchant capacity gained or lost share against regulated utilities year over year.
• Energy traders and analysts anchor long-horizon fuel-switching views in the generation tables - the 2024-versus-2023 coal decline against gas, nuclear, wind and solar rows is the transition debate in four numbers.
• Utility strategy and regulatory teams benchmark retail prices, revenues and customer counts by state and sector, and read planned capacity additions and retirements before they land in interconnection queues.
• Fuel suppliers and rail/logistics planners work from the consumption, stocks, receipts and delivered-cost tables, including coal cost and heat content by census division.
• ESG and climate researchers pull SO2, NOx and CO2 emissions alongside generation so intensity ratios come from one internally consistent volume.

For contrast inside the same industry: PJM Markets & Operations and CAISO OASIS carry the high-frequency market signals this annual volume deliberately abstracts away, and the Form EIA-860 generator inventory supplies the unit-level grain beneath these aggregates.

Investors and quant researchers get thirty years of annually consistent generation, capacity, price and emissions aggregates ready for structural models without cleaning. Market researchers and consultants anchor power-sector market sizing and due-diligence decks on official figures instead of vendor estimates - the ownership split alone justifies the pull. Data scientists and ML engineers use it as the trusted target variable and validation layer for anything trained on higher-frequency feeds. Journalists, academics and students cite the canonical answer to "where does US electricity come from" with territory coverage nobody else bothers with. Persona-by-persona detail lives on the industry hub at /industries/independent-power-producers-energy-traders.

Three things worth knowing upfront.

First, cadence versus structure. This is the annual volume: it settles structural questions - who owns what, which fuels won the year, where prices sit by state - and leaves intra-year movement to monthly and weekly feeds. If your question moves at market speed, pair this record rather than substituting it.

Second, aggregation is the point. Roughly 130 tables mean roughly 130 deliberate cuts, not raw records; the plant-and-generator grain lives in Form 860/923 companions. Teams that try to force unit-level analysis out of an aggregate yearbook end up disappointed, and we would rather scope the right companion honestly.

Third, scope the sample to your sections. Twelve sections do not all matter to every team - a trader wants generation and fuel cost, an ESG analyst wants emissions against generation, a regulator wants sales, prices and reliability. Name your sections and years and the sample arrives shaped to them, dictionary intact either way. Delivery cadence - daily, weekly, or hourly - gets decided after the sample validates.