EPA eGRID data needs facility load for a data center carbon audit

EPA eGRID data needs facility load for a data center carbon audit. The Environmental Protection Agency publishes eGRID data and records its revisions; neither supplies a carbon-intensity value for a campus whose load, interconnection, or procurement is unknown. A credible assessment begins with the facility’s electricity use and the grid factor for the time and place of that use. It does not begin with a single national average, a press-release renewable claim, or a multiplier copied from another portfolio. A useful audit separates measured load, location-based emissions, market-based instruments, backup generation, water and construction impacts, and the uncertainty attached to each.

EPA’s eGRID release supplies U.S. generation and emissions data along with data notes and revisions. Those records make a regional historical estimate more inspectable, but they still do not supply a carbon-intensity value for a campus whose load, interconnection, or procurement is unknown. (EPA eGRID data notes and revisions)

Build the audit from four separate records

An emissions claim becomes inspectable when its inputs are kept separate:

Record	Minimum evidence	What it can show	What it cannot show alone
Facility load	Metered MWh, interval, and boundary	Electricity actually consumed inside the stated boundary	Which generators served each interval
Grid factor	Region, year, method, and data source	A location-based estimate using a transparent factor	A customer-specific contractual claim
Procurement	Renewable contract, certificate, retirement, and matching rules	What the operator procured under its accounting method	The physical generation mix at the meter in every hour
On-site generation	Fuel, runtime, emissions controls, and output	Direct emissions and resilience tradeoffs	Grid emissions avoided without a counterfactual study

EPA’s eGRID data supplies generation, emissions, emission rates, resource mix, and related data across U.S. geographic levels. Its revision record documents that even plant-level quantities and emission rates can change when underlying data is corrected. That makes it a valuable input for a historical location-based estimate, not a license to assign a precise footprint to a facility whose load, interconnection, and procurement are unknown. (EPA eGRID data notes and revisions)

Decide which question you are answering

Many contradictory carbon claims are answers to different questions. State the question before selecting a factor:

Question	Appropriate starting point	Common error
What emissions were associated with the electricity used at a site last year?	Metered MWh multiplied by a documented location-based annual factor	Presenting a national average as a local result
What did the company claim under a market-based accounting approach?	Contracts, certificates, retirement evidence, and the accounting rules used	Calling a contractual claim the physical hourly grid mix
What emissions change would a new load or efficiency project cause?	A marginal or dispatch-aware method with a defined counterfactual	Using an annual average as a forecast of incremental generation
How does a project affect a community?	Grid study, local air permits, water and land analysis, tax and rate design	Treating carbon intensity as the only material effect

EPA publishes different quantification approaches because the use case affects the method. Its page describes non-baseload factors, capacity-factor methods, historical hourly rates, and energy modeling for energy-efficiency and renewable-energy analyses. A buyer should not treat them as interchangeable simply because each produces a number in units of carbon dioxide per megawatt-hour. (EPA electricity-emissions quantification guidance)

Make the boundary explicit

Before comparing two campuses or a company target to a regional estimate, record:

The physical boundary: IT load only, whole facility, campus, or all leased capacity.
The time boundary: hourly, monthly, annual, historical, or projected.
The emissions boundary: operational electricity only, or also on-site fuel, construction, equipment manufacture, and water.
The accounting basis: location-based, market-based, marginal, or a scenario model.
The source and version for every factor, including data-year and revision.
Whether the number is metered, estimated from capacity, or inferred from a model.

This prevents a common false precision: multiplying a reported megawatt capacity by 8,760 hours and an average grid factor does not establish a campus’s actual energy use or emissions. Capacity is not utilization, and a factor can be annual, regional, historical, or marginal. A transparent estimate may still be useful, but it needs to say which assumptions it makes. EIA’s data-center outlook is likewise a scenario analysis of server electricity use, not a meter reading for any campus. (EIA data-center server energy analysis)

Treat procurement and physical delivery as distinct claims

Power-purchase agreements, renewable energy certificates, and clean-energy targets can be material to a company’s financial exposure and market-based reporting. They do not automatically answer the location-based question. Conversely, a location-based annual factor does not by itself establish whether a procurement contract caused new clean generation or reduced future emissions. GHG Protocol guidance distinguishes a location-based method using grid-average emission factors from a market-based method using contractual instruments; both labels should remain visible in a corporate inventory. (GHG Protocol Scope 2 guidance questions)

Keep both columns in an audit. If a company reports market-based electricity emissions, document the instrument, vintage, geography, retirement, and matching convention. If the aim is a physical-grid estimate, document the load location, time window, grid region, and factor. Do not subtract one from the other and label the remainder a universal “PPA gap.”

Include the constraints that carbon figures omit

Electricity emissions are only one part of infrastructure risk. A broader due-diligence review can separately request evidence about interconnection timing, on-site combustion, air permits, water source and discharge, transmission upgrades, backup-generator testing, community rate effects, and contractual responsibility for new infrastructure. Those questions are outside the electricity-emissions calculation and require their own site records.

GHG Protocol also records a public consultation on Scope 2 updates that ended in January 2026. Consultation is not, by itself, a replacement standard; check the published guidance and the accounting version used for the inventory. (GHG Protocol Scope 2 guidance and update status)

Grid data are historical and may be revised. That supports a rigorous, repeated audit, not a permanent conclusion that any particular region or fleet will stay above or below an average. Record the data year and revision alongside every estimate. (EPA eGRID data notes and revisions)

A compact due-diligence checklist

Before relying on a data-center carbon claim, ask for:

Interval and annual MWh, with the operational boundary.
The eGRID or other grid-region factor, data year, revision, and calculation method.
Separate direct-fuel emissions and generator runtime.
The accounting treatment and evidence for contracts or certificates.
The basis for any forecast, including utilization and interconnection assumptions.
A sensitivity range for load and factor uncertainty.
Local permits and community-impact evidence outside the carbon calculation.

The goal is not to discard every estimate. It is to make clear what the estimate describes, what it leaves out, and what evidence would change it. That is more useful for an investor, regulator, community, or infrastructure buyer than a single carbon headline carried from a different facility or method.