Methodology

What we calculate

For each metropolitan statistical area (MSA), we estimate annual electricity demand and translate it into the number of reactors of various sizes that would be required to cover that demand at standard capacity-factor assumptions. We do the same translation for solar, wind, gas, and coal at typical fleet capacity factors. We attach capital cost bands to the reactor scenarios.

The headline equation is straightforward:

units_needed = annual_demand_MWh
              / (unit_MW × capacity_factor × 8,760)

Demand: state per-capita weighting (v2)

Translating state-level electricity data to MSA-level demand is the deepest methodology challenge on this site. The U.S. Energy Information Administration publishes monthly retail electricity sales at the state and utility level, but not at the MSA level — there is no clean public dataset that says "the Atlanta metro consumed X MWh in 2024."

For v2, each metro's annual demand is calculated as:

metro_demand = state_total_demand × (metro_pop / state_pop)

The state total comes from the most recent complete calendar year of EIA retail sales. Population values are from U.S. Census Vintage 2025 estimates. This captures state-level industrial mix variation that the v1 US-national-per-capita approach missed entirely: industrial states like Texas (~14 MWh/capita) come out higher; service-economy states like California (~7 MWh/capita) come out lower.

Residual error: state per-capita weighting still assumes a metro's electricity-intensity matches its state average. In reality, the Houston metro and rural Texas have very different demand profiles — Houston has more industry, rural Texas has more residential. We treat this as ±20% uncertainty on the metro demand figure for now, and disclose the methodology basis on every metro page's provenance footer.

For metros where state population is unavailable, the calculation falls back to U.S. national per-capita × metro population, with the fallback clearly labeled in the provenance footer.

Demand: utility-level aggregation (planned v3)

The most accurate method for MSA-level demand is utility-level aggregation via Form EIA-861. EIA-861 reports annual electricity sales by individual utility. Each utility's service territory can be mapped to counties, and counties roll up to MSAs. Summing utility sales weighted by service-territory overlap with each metro's counties produces a true MSA-level demand figure.

We have not yet shipped this because EIA-861 data is published as bulk Excel workbooks (~20 spreadsheets per annual release), not via a clean API. Building the full pipeline — Excel parsing, utility → county → MSA mapping, multi-MSA utility allocation — is a multi-day engineering effort. It is on the roadmap as v3.

Until v3 lands, MSA demand figures should be read as state-weighted first-order estimates. They are good enough for the order-of-magnitude translation to reactors that this site is designed to communicate, but not precise enough to inform a specific deployment decision.

Source comparisons (solar, wind, gas, coal)

Each metro page shows how many units of various non-nuclear sources would be needed to cover the same annual MWh as the nuclear scenarios. Counts use raw MWh equivalence at typical fleet capacity factors:

• Utility solar: ~25% capacity factor, ~6 acres per MW, ~400W panels
• Onshore wind: ~35% capacity factor, 3 MW turbines, ~100 acres per turbine including spacing
• Natural gas combined cycle: ~55% capacity factor, ~800 lb CO₂/MWh
• Coal: ~50% capacity factor, ~2,200 lb CO₂/MWh

Firm-equivalent counts for solar and wind

The raw MWh-match counts for solar and wind represent how many units you'd need to generate the same annual energy as a nuclear plant. They do not represent how many units you'd need to provide the same firm power.

To deliver round-the-clock firm power equivalent to a baseload nuclear plant, variable renewables need additional nameplate capacity for oversizing, paired energy storage to shift output across hours, and curtailment headroom. For each renewable card we show a second count range labeled "firm equivalent (with storage)" that captures this.

The multipliers come from NREL Standard Scenarios and Lazard LCOE+LCOS portfolio modeling for renewable+storage deployments designed to match baseload firm output:

• Utility solar with storage: 2.5–4.0× raw count
• Onshore wind with storage: 1.8–3.0× raw count

These are illustrative national ranges, not site-specific designs. A real renewable+storage portfolio for a specific metro would also benefit from geographic diversification, demand response, and existing transmission — none of which the simplified model captures. Treat the firm-equivalent counts as a more honest upper-bound estimate for a like-for-like comparison with firm baseload.

Grid mix and avoided emissions

Each metro page shows the most recent annual generation mix for the metro's electrical grid. We use EPA eGRID 2023 subregion data when available — the gold-standard EPA reference for grid emissions and resource mix. eGRID divides the U.S. into 26 subregions that match actual electrical grid boundaries (PJM, ERCOT, MISO, CAISO, etc.) better than state lines, since electricity flows across state boundaries but not across grid boundaries.

Each of our top-25 metros is mapped to its primary eGRID subregion based on central-city location. A few metros span multiple subregions (notably New York, which touches NYCW, NYLI, NYUP, and RFCE) — for v1 we use the dominant subregion and disclose the simplification. State-level EIA data is retained as a fallback for completeness.

The avoided-emissions estimate is shown as a range using EPA's two published rates:

• Low bound — eGRID Total Output rate: the subregion's average emission rate across all generation. Reflects the average emissions of currently dispatched fossil generation.
• High bound — eGRID Non-baseload rate: the subregion's marginal emission rate — what gets displacedfirst when clean energy is added, typically gas peakers and other higher-emission marginal generation. This is conceptually similar to the AVERT marginal rate but eGRID-published per-subregion.

Both bounds are scenario estimates assuming nuclear fully replaces the fossil portion of generation — a thought experiment, not a deployment forecast. The range itself communicates that "avoided emissions" is not a single number; it depends on the marginal-vs-average accounting framework chosen.

For metros without eGRID coverage (none in our current top-25, but provided for future expansion), we fall back to a hand-blended fossil rate (coal: 2,200, gas: 800, oil: 1,700 lb CO₂/MWh) for the low bound and the EPA AVERT national-weighted marginal rate (1,562 lb CO₂/MWh) for the high bound.

Reactor cost bands

Capital cost figures are anchored to the National Renewable Energy Laboratory's 2024 Annual Technology Baseline (ATB), which gives reference overnight capital costs of roughly $7,500/kWe for large reactors and $8,000/kWe for 300 MWe SMRs. We widen the bands significantly to reflect first-of-a-kind delivery risk: the recent Vogtle 3 & 4 build came in around $13,000–15,000/kWe, and the NuScale UAMPS Carbon Free Power Project was cancelled in late 2023 after costs rose substantially above initial estimates.

Microreactor figures are the most uncertain category. No commercial unit is operating in the United States as of this writing.

What we do not claim

• These figures are not forecasts or project proposals.
• They are not permitting determinations or siting recommendations.
• They do not account for transmission, storage, demand response, or the reliability profile of competing technologies (this matters most for the solar/wind comparisons).
• They do not reflect community acceptance, fuel supply, water availability, or the many other factors that determine whether a reactor — or any plant — can actually be built somewhere.

Sources

• U.S. Census Bureau, Population Estimates Program, Vintage 2025
• U.S. Energy Information Administration, retail sales API
• U.S. Energy Information Administration, electric power operational data (state generation by fuel)
• EPA eGRID2023 — subregion CO₂ emission rates (Total Output and Non-baseload) and resource mix; the primary source for grid mix and avoided emissions on metro pages
• NREL Annual Technology Baseline 2024
• NREL Standard Scenarios 2024 — capacity credit values used to inform firm-equivalent multipliers (NREL/TP-6A40-89587)
• EPA AVoided Emissions and geneRation Tool (AVERT) v4.3, April 2024 — marginal CO₂ emission rate for high-bound avoided emissions
• Lazard LCOE+LCOS analysis — informed renewable+storage multiplier ranges
• IAEA ARIS, NEA SMR Dashboard, NRC filings, vendor disclosures (reactor specifications)
• Power source archetypes (solar, wind, gas, coal): typical fleet values from public NREL, EIA, and EPA references; cataloged in lib/power-sources.ts.
• Metro boundary geometry: U.S. Census Bureau TIGER/Line 2020 — Core Based Statistical Areas. Polygons simplified via Douglas–Peucker (~0.012° tolerance) and served as static GeoJSON from /public/metros/.
• Metro / CBSA FIPS lookup: SimpleMaps US Metros (used to cross-reference our metro slugs against the CBSAFP codes in the TIGER shapefile). Free database used under their basic license — credited on every page footer.