Require Tier 4 emission standards, metered runtime and permit-before-operate enforcement for on-site combustion

The proposal

Four requirements, applied as air permit conditions:

1. A Tier 4 emissions baseline for new engines, rather than allowing the cheaper Tier II class. Virginia's HB 507 and revised DEQ guidance impose 0.60 g/hp-hr on new permits from 1 July 2026 (Data Center Knowledge).
2. Metered, reported and published runtime hours per unit, distinguishing testing, emergency and non-emergency operation.
3. A narrow definition of emergency that excludes scheduled utility maintenance and economic dispatch.
4. Permit before operation, with the permit count verified against installed equipment before commissioning.

Why it would work

The health cost estimates for this equipment are driven almost entirely by runtime, not by installed capacity. Modelled Northern Virginia costs are $190 to 260 million a year at actual runtime and $1.9 to 2.6 billion if generators ran at maximum permitted levels (UC Riverside), and the state's own legislative audit finds generators are currently under 4% of regional NOx precisely because they mostly sit idle. The whole question is therefore how many hours they run, which is a permit variable and nothing else.

This is why the definition of emergency matters more than any emission factor. September 2025 DEQ guidance would reclassify utility-scheduled outages announced within 14 days as emergencies, letting dirtier Tier II units run during planned maintenance, despite research finding most facilities experience zero to two minor outages a year lasting one to five hours (Virginia Mercury). A broad emergency definition converts standby equipment into part-time power plants without any new permit.

The fourth requirement addresses a failure that has already happened at scale. xAI installed 35 turbines at its Memphis site and received a permit for 15, with satellite imagery showing at least 24 running the day before the permit issued (SELC). A permit regime that cannot count installed units before they operate is not functioning.

Implementation path

State environmental agencies hold the authority. The sequence that works is: adopt the Tier 4 baseline for new permits, require telemetry-based runtime reporting as a permit condition, publish the reported hours, and fund inspection capacity sufficient to verify equipment counts at commissioning. Where "temporary" or "mobile" equipment exemptions exist, close them, since that is the category the largest unpermitted installations have used.

Trade-offs and limitations

• Retroactivity. Roughly 8,000 of Virginia's 9,000 data centre generators are already installed Tier II units. A new-permit baseline does nothing about them, and replacement is expensive; a runtime cap is the only near-term lever on the existing fleet.
• Tier 4 aftertreatment costs and complicates. Selective catalytic reduction adds capital cost, urea logistics and a failure mode that a standby unit may not exercise often enough to detect.
• Enforcement is the binding constraint, not the standard. The Memphis and Southaven cases involved equipment operating with no valid permit at all. A stricter standard on paper does not help if the agency lacks the capacity or willingness to inspect.
• It does not address the grid emissions of the electricity the facility consumes normally, which is a separate and much larger quantity.