Air

The Paso del Norte airshed is a bowl. The Franklin Mountains wall it on the east, the Sierra de Juárez on the southwest, Mount Cristo Rey stands between them, and inside sit El Paso, Sunland Park, and Ciudad Juárez — roughly 2.7 million people sharing one body of air. Terrain traps it. The agencies use the word themselves.

Three jurisdictions, two countries, one air mass. Emissions cross those lines in both directions all day, and the monitors record whatever arrives — they cannot tell you whose it was. Everything that follows on this page is a consequence of that single physical fact colliding with a set of legal instruments that were each drawn around a jurisdiction.

Nonattainment is not a warning. It is the federal government’s formal finding that the air people breathe in a place fails a health-based standard. Part of Doña Ana County has carried that finding since July 1995, when the original “Sunland Park area” designation named Sunland Park, Santa Teresa, and La Union. The current designation — Marginal under the 2015 ozone standard of 70 parts per billion — took effect in August 2018. The area missed its August 2021 attainment deadline with a design value of 78 ppb.

Then it got better. This is the part of the story that rarely gets told, and it belongs here first: across the last four official measuring periods, the basin has very nearly recovered.

Ozone design values, parts per billion. The standard is 70. Source: EPA design-value workbooks.

Two things fall out of that table. The first is that the recovery is real and recent: Desert View, the basin’s worst monitor, dropped from 78 to 68 — below the standard. Every monitor in the area now reads clean except one. La Union, at 71 ppb, holds the entire area in violation by a single part per billion.

The second is quieter and it is the finding. The monitor closest to the proposed facility — Santa Teresa, about 3.3 miles away — has produced INVALID data for the last two consecutive measuring periods. It failed completeness. The nearest working monitor is Desert View, roughly nine miles east, and no monitor of any kind sits at the facility site. At the moment a 2,462-megawatt source is being permitted, the instrument nearest to it is the one not producing usable numbers.

To be precise about what the permit does and does not require: it requires a great deal of monitoring AT THE SOURCE. NMED’s draft directs the applicant to expand its monitoring plan beyond temperature and pressure to continuous tracking of startup and shutdown events, gas flow rate, electrical output in megawatts, exit sulfur concentration, and natural-gas leak rates at the fuel-cell units. That is a real instrument package, and it answers a real question: what is coming out of this facility.

It is a different question from what is in the air over Sunland Park. Source monitoring measures the plant. Ambient monitoring measures the bowl. No ambient monitoring requirement appears in the draft permit, which is ordinary for a construction permit and is not a lapse by the department — it is simply not the instrument this proceeding uses. The gap is structural, not anyone’s oversight.

Doña Ana is a two-nonattainment-area county: Anthony, New Mexico carries a separate PM-10 designation, classified Moderate, designated in 1991 and never redesignated — no maintenance plan was ever developed, and the area still exceeds. The missing-meter pattern appears elsewhere in this record →

Two mechanisms decide how the basin’s failure is accounted for, and neither of them asks about the basin.

The first is the international excuse. Under Clean Air Act §179B, the region argued it would have met the standard but for emissions from Ciudad Juárez. The inventory behind the argument is substantial: Juárez contributes an estimated 64% of regional nitrogen oxides and 67% of volatile organic compounds, and back-trajectory modeling puts Mexican-airspace influence on 80% of exceedance days at Desert View. EPA proposed to approve it in March 2023, which would relieve the area of mandatory reclassification to a stricter Moderate regime. That proposal was never finalized. It has sat pending for more than three years — while comparable §179B determinations elsewhere, including Phoenix-Mesa, were finalized in 2026. The area is neither formally excused nor bumped up. It waits.

The second is the line itself. A new source inside the nonattainment boundary faces a 100-ton-per-year major-source threshold and must buy offsets at 1.1 to 1. A new source outside it faces 250 tons and no offsets. Per NMED’s own modeling report, this facility sits about six miles west of the boundary.

Nothing here is a malfunction. Each instrument is doing exactly the job it was designed to do. The consequence is simply that existing exceedances are attributed across a border and new emissions are sited across a line — and the basin’s total load is nobody’s question. The monitors record the sum either way. The instruments only decide who has to answer for it.

“The air doesn’t stay over one area, it is all of ours. It’s part of the commons.”— Sen. Bill Soules, legislative listening session, August 11, 2026 · quote pending audio verification

Four things must be said before anything else, because they are true and because a record that buries them cannot be trusted with the rest. The facility is outside the nonattainment boundary — that is NMED’s own finding, not the applicant’s claim. The fuel-cell redesign is a genuine reduction in criteria pollutants: from caps of 249.97 and 248.90 tons of NOx a year across the two withdrawn applications to 37.2 in the current one, about thirteen times lower. The current application requests Title V major-source status outright — it asks for the heavier oversight regime rather than ducking it. And the largest criteria figure in the application is one the vendor itself disputes: the application records that the manufacturer “maintains that there are zero PM emissions from its equipment” and that testing “may exhibit positive sampling method bias” — and the applicant adopted the conservative particulate factor over its own supplier’s position. That is a genuine point in their favor, and this page says so plainly.

The two withdrawn applications requested caps of 249.97 and 248.90 tons — three-hundredths and one and a tenth of a ton under the 250-ton line at which federal review attaches. Their own tables put the East facility’s uncontrolled potential at 3,699.16 tons: the caps were built not from hardware but from dispatch promises — “only thirty (30) units will operate at 100% load”; “not more than five (5) turbines will operate simultaneously.” A cap that is a promise about how many turbines run is not a control technology. Both applications were withdrawn in April 2026, the day the department decided to hold a hearing. That story is told in full in the Friction Ledger →

From the application’s own summary tables — NMED Universal Air Quality Application, prepared by Trinity Consultants, 648 pages, cover letter April 27, 2026: Bloom Energy Server 6.5 units, 2,462.005 MW, 8,760 hours a year, 15% safety factor included. Cross-checked against NMED’s June 16 modeling review, which carries the same hourly rates.

And one number in that table crosses a line the argument has been looking past. 124.01 tons a year of VOC exceeds the 100-ton threshold at which a source inside a nonattainment area is major for Nonattainment New Source Review. The facility sits about six miles outside the boundary, so that review does not apply — but inside the line, this would be a major source for VOC, an ozone precursor, facing emissions offsets and Lowest Achievable Emission Rate. The boundary question this page keeps returning to is live on VOC, not on NOx.

What survives all of that is simple arithmetic. Thirty-seven tons of NOx, one hundred twenty-four tons of VOC, seventy-five tons of fine particulates, and a hundred sixty-one tons of carbon monoxide a year are additions to a basin sitting one part per billion from compliance after a thirty-year climb — and above them, ten million tons of CO2-equivalent. They are small measured against a threshold. They are not zero measured against a bowl. The thresholds ask about the source; the bowl does not care about the source.

The particulate number deserves its own sentence. At 75.4 tons a year, PM2.5 is the largest criteria pollutant in this application — twice the NOx that dominates the argument — and it is the pollutant for which the health literature has never established a safe threshold.

Sunland Park, Santa Teresa, La Union, Anthony. Border communities, largely Hispanic, and the county’s colonias — roughly 37 of them, home to about 13,000 people. These are the places named in the 1995 designation and the places the monitors sit in.

Ozone is a respiratory irritant; it inflames airways and worsens asthma, and children are the sensitive population by consensus of the health agencies. This record does not and will not claim that this facility will cause a particular number of cases — that claim cannot be made honestly from the documents. What can be said is the burden that already exists. In 2021, 16.3% of Doña Ana County teenagers reported having been diagnosed with asthma. Statewide, about 30,000 New Mexico children have it.

Set that against the sequence this same population has lived through: a water utility that bypassed three of four arsenic treatment plants for over a year, found in 2023 and settled in 2026 with all plants now compliant; ozone nonattainment since 1995; a separate particulate nonattainment area at Anthony; discolored tap water reported as recently as August 2026; and now a 2,462-megawatt gas-fired campus. Each of those is documented and none of them is the same problem. They land on the same people.

The federal instrument that would have counted exactly that accumulation — environmental-justice review in data-center permitting — was removed by executive order in July 2025, one month before the county adopted its bond resolution.

The school district serving these communities, Gadsden, receives $820,456 a year from the project’s payment in lieu of taxes — about seventy dollars per student. The full arithmetic is in the record →

The withdrawn applications answered one open question on this page: they state the distance from the facility to the nearest residence, school, or occupied structure as 1,500 meters — a fact from the applicant’s own filing, not this record’s measurement. One thing this section still does not have and will not guess at: the county’s pediatric asthma hospitalization rate, which sits behind an interactive state query and has not been pulled.

The hearing begins September 14, 2026, at the Sunland Park Multi-Generational Center — in person and on Webex, with Spanish interpretation. Technical testimony and prehearing motions were due August 24. The department must decide by November 23.

The regulatory regime this permit is judged under is unresolved in both directions. Because §179B was never finalized, the area is neither excused from reclassification nor bumped to the stricter Moderate rules — while its design values improve toward attainment from the other side. Today’s siting happens against a regime that could tighten or dissolve, and either resolution arrives after the permit is decided.

And this permit may not be the campus. Two facts sit on the record, and this page publishes both without resolving them. The applicant’s own cover letter, April 27, 2026: the two prior applications “are being withdrawn, and Yucca does not foresee submitting any additional construction permit applications.” And the deed record, November 12, 2025: a 702-acre parcel — larger than the entire publicly announced footprint — conveyed to an entity named SI SNM01Expansion, LLC, alongside the parcels held by the power and data-center entities. The commitment is narrow (construction permit applications) and soft (does not foresee). This record asserts no future emissions from that parcel; it records that the parcel exists, who holds it, and what the applicant has said.

Which is the precedent worth naming. If the basin’s total load is nobody’s question, then every future facility — and every future phase of this one — sites the same way: just outside whatever line exists, into the same air. The next application inherits this one’s geometry. The monitors will record the result either way.