What does the DC Readiness score mean?

It is a 0–100 screening estimate that blends fourteen weighted factors into a single number. A higher score means a site screens better across power, speed-to-power, fiber, energy cost, and hazard. Speed-to-power is scored on the structural interconnection speed of the market, so Texas (ERCOT) scores near the top and PJM near the bottom. That market factor was set before ERCOT paused approvals to energize new datacenter and crypto-mining loads of 75 MW or more on August 3, 2026, pending reports due December 10, 2026, so treat Texas speed scores as pre-pause. It is a starting point for site selection, not a substitute for site-specific engineering, environmental, or interconnection studies.

Is the catalogued capacity actually available power?

No. 'Catalogued candidate capacity' is the sum of per-site available-capacity estimates — a theoretical aggregate, not deliverable power and not power available now. Actual deliverable capacity depends on interconnection studies and utility commitments.

Where does the data come from?

Public sources including FEMA's National Risk Index, FCC fiber data, EIA electricity rates, FERC interconnection queues, parcel and land-use records, and climate normals. Scores are derived estimates and refresh monthly.

How often is the dataset updated?

The aggregate dataset refreshes monthly. Every page shows a machine-readable 'Dataset updated' timestamp.

DC Readiness Methodology

GridCensus scores 147,272 candidate datacenter locations on a single 0–100 DC Readiness score. The score is a weighted blend of fourteen factors derived from public data sources. It is a screening tool to triage where to look first — not a site-specific engineering, environmental, or interconnection assessment.

Every source, how authoritative it is and when we last read it is listed on Data sources and standards. On the site dossier, click any dotted-underlined figure to see the record behind it.

Scoring weights

Factor	Weight	What it captures
Power availability	20%	Estimated deliverable capacity, substation proximity, voltage, and headroom.
Speed-to-power	20%	Structural interconnection speed of the market, plus measured queue wait, completion rate, and load growth. Berkeley Lab's queue data puts the median request-to-operation time for generation and storage projects that came online since 2020 at about 3.5 years in ERCOT and 4.7 in PJM. Texas scores high on this factor, but the score predates ERCOT's August 3, 2026 pause on energizing new datacenter and crypto-mining loads of 75 MW or more (see below).
Fiber	12%	Count of nearby fiber providers and internet-exchange proximity.
Energy cost	10%	Commercial/industrial electricity rates.
Water	7%	Water-stress index (cooling-water availability proxy).
Hazard	6%	FEMA National Risk Index and flood-zone exposure.
Buildability	7%	Slope, parcel size, and site-development friction.
Labor	4%	Construction-trades employment and wage data.
Existing datacenter	4%	Proximity to existing datacenter clusters and ecosystem.
Land	3%	Estimated land price per acre and parcel size.
Construction cost	3%	Regional construction cost index.
Gas pipeline	2%	Proximity to natural-gas pipeline for on-site generation.
Tax	1%	Presence of datacenter tax incentives in the jurisdiction.
Climate	1%	Cooling/heating degree days and mean temperature.

Texas note: Since August 3, 2026, ERCOT has paused approvals to energize new datacenter and crypto-mining loads of 75 MW or more until a verification audit and a community impact review are complete; both reports are due to the Public Utility Commission of Texas by December 10, 2026. ERCOT speed-to-power scores here were set before the pause and do not reflect it. ERCOT board update.

Queue-wait inputs: Lawrence Berkeley National Laboratory, Queued Up (interconnection requests through 2024), median time from request to commercial operation for projects that came online since 2020.

Substation headroom estimate

Every transmission substation (about 62,000) carries an estimate of how much new load it could take before a transmission line reaches its rating. It is built entirely from public data:

A model of the grid. HIFLD transmission lines (route, voltage, length, voltage-class rating) joined to the substations at their ends: about 70,000 nodes and 89,000 branches, split into the Eastern, Western and ERCOT interconnections by balancing authority.
Supply and demand at summer peak. Every operating power plant from EIA-860 (1,259 GW) placed at its nearest substation in its own interconnection; demand spread by 2024 county population and scaled to each interconnection's public peak; plants dispatched by technology.
A DC power flow. For each substation the model adds load, meets it from the generators with room to ramp, and finds the first line that would reach its rating. The engine reproduces PYPOWER's results exactly on Texas A&M's public synthetic 2,000-bus Texas grid (flows and sensitivities to machine precision; the limiting line lands at exactly 100% in brute-force checks).
Calibrated with real congestion. Lines the grid operators report as binding (PJM, MISO, SPP, ERCOT, NYISO, CAISO) are treated as near their limit. Checked against held-out constraints: substations near bottlenecks the model was not given score 23–48% lower than others at the same voltage, and even with no congestion data at all the network model alone scores them 12–35% lower.

What it is not: a utility determination. Public data has no line impedances, exact ratings, transformer limits, switching arrangements or contingency lists; those live in the planning models utilities and grid operators hold as CEII. Read the estimate as a ranking (tight, limited, moderate, room) that tells you which substations to ask the utility about. Free accounts see the band; Pro and Enterprise see the MW figure and the limiting line.

Honest limitations

Scores are screening estimates derived from public data, not verified, site-specific assessments.
Catalogued candidate capacity is a theoretical aggregate of per-site estimates — never "deliverable" or "available now."
Interconnection feasibility, environmental remediation, and utility commitments require dedicated studies beyond this screen.

FAQ

What does the DC Readiness score mean?: It is a 0–100 screening estimate that blends fourteen weighted factors into a single number. A higher score means a site screens better across power, speed-to-power, fiber, energy cost, and hazard. Speed-to-power is scored on the structural interconnection speed of the market, so Texas (ERCOT) scores near the top and PJM near the bottom. That market factor was set before ERCOT paused approvals to energize new datacenter and crypto-mining loads of 75 MW or more on August 3, 2026, pending reports due December 10, 2026, so treat Texas speed scores as pre-pause. It is a starting point for site selection, not a substitute for site-specific engineering, environmental, or interconnection studies.
Is the catalogued capacity actually available power?: No. 'Catalogued candidate capacity' is the sum of per-site available-capacity estimates — a theoretical aggregate, not deliverable power and not power available now. Actual deliverable capacity depends on interconnection studies and utility commitments.
Where does the data come from?: Public sources including FEMA's National Risk Index, FCC fiber data, EIA electricity rates, FERC interconnection queues, parcel and land-use records, and climate normals. Scores are derived estimates and refresh monthly.
How often is the dataset updated?: The aggregate dataset refreshes monthly. Every page shows a machine-readable 'Dataset updated' timestamp.

Dataset updated November 2026. Screening estimates derived from public data sources.