At peak capacity, it will consume 560 GWh per year—equivalent to the yearly consumption of 200,000 Uruguayan homes. This is more than any one locality in the country besides Montevideo, or about 5 percent of national electrical demand. Its perimeter is adjacent to the athletic fields of Lycée Français, the seats of the Lawn Tennis and Old Christians clubs, other businesses, the Colonia Nicolich neighborhood and the gated community of Las Higueritas.

Amenaza Roboto reviewed the case file in its entirety: The environmental impact study, the technical annexes, four rounds of inquiries from the Ministry of the Environment and the responses provided by Eleanor Applications with data and analysis from the consultants it hired. All of this formed the basis of the approval of the Prior Environmental Authorization (AAP), which permitted construction.

What follows is a reconstruction of what these files say, what they don’t say, and what was postponed until the day in which the data center becomes operative.

A data center doesn’t power off. Its servers process information continuously, and each second without energy can mean a loss of data and services. If the electrical grid fails, diesel generators activate to keep the servers on. At Teros, there are 27. Each measures over 8 meters long and 3 meters high, and weighs 35 tons, according to official technical specifications.

Among the contaminants that these motors release into the air, nitrogen dioxide (NO₂) stands out. An irritant, it’s the same gas that you would see coming out of the exhaust pipe of an old truck. It’s the only contaminant studied in the files that exceeds the Uruguayan legal limit during a power outage. The rest (particulate matter, sulfur dioxide, carbon monoxide) show normal values during emergency simulations.

The Environmental Impact Study carried out by the consulting firm EIA Estudio Ingeniería Ambiental includes a dispersion modeling analysis executed by Trinity Consultants, which simulated what would happen if all 27 generators were to activate at the same time due to a power outage. The concentration of nitrogen dioxide could reach 3,439 micrograms per cubic meter.

In Uruguay, the legal limit is 200.

The maximum tolerance level is 260.

The projected peak is 17 times the legal limit and 13 times maximum tolerance.

The risk isn't limited to emergency situations. During normal operations, the generators are turned on once a month as preventative maintenance, one at a time for about 30 to 45 minutes, depending on the generator's power output.

During those tests, the model yields a maximum concentration of 210 micrograms per cubic meter, above the legal limit of 200.

Regulations allow for up to 18 hours per year with values between 200 and 260, and the study concludes that this criterion is met. But the margin is minimal: The legal limit is exceeded during the project's regularly planned operations—with only one generator being tested.

Uruguay’s state-owned power company (UTE) has recorded 21 outages from 2001-2017 in the area. One of those occurred after an extratropical cyclone battered Montevideo and the south of the country in 2005. It lasted 65 hours and 17 minutes—almost three whole days.

The environmental report factored out that outage in its modeling. The consulting firm Estudio Ingeniería Ambiental argued that it stemmed from “a probability distribution different from that of typical outages” and that, by being a singular event, it wasn’t possible to model its probability.

The Ministry questioned the exclusion. It requested that “the exclusion of this event be reviewed” because the analysis should have considered “any contingent event, not just those that are representative of typical service interruptions.”

The consulting firm simulated outages of 7 to 11 hours, and assumed a conservative estimate of one 24-hour interruption per year. Neither scenario modeled the real-world 65-hour outage caused by the 2005 cyclone. The company maintained this exclusion throughout the entirety of the files.

The impact of 27 diesel generators running for days on end—like what occurred in 2005—was never calculated. It was concealed by an Emergency Action Plan that, at the time authorization was granted, had not been fully developed.

The generators for Project Teros have a Tier 2 certification, a U.S. Environmental Protection Agency (EPA) standard that dates to 2006. Since 2015, there has been a stricter standard, Tier 4, that reduces nitrogen oxide emissions by 90 percent.

With Tier 4 generators, the problem that dominates the entirety of the files would disappear.

The files explain why these were not chosen. Trinity Consultants, the firm hired to do the modeling, presented three arguments:

• Economic. An additional $16.3 million dollars, with a cost of reduced emissions three times the proposed budget.

• Technical. Specific challenges associated with emergency use, such as overheating the catalytic converter and the risk of automatic shutoff.

• Environmental. “The worst-case emissions calculated for the project still comply with the regulatory requirements and do not represent a significant environmental impact.”

An intermediate alternative (the selective catalytic reduction within Tier 2) was also rejected for its even higher cost: $26 million dollars.

In Trinity's analysis, the economic argument was the deciding factor. The technical and environmental arguments alone weren't enough to rule out Tier 4: The former references operational limitations, not impediments; the latter says that “it is not necessary.” Without the additional $16.3 million, Tier 4 would be a reasonable option.

The problem is that those $16.3 million dollars—and the claim that there isn’t a “significant environmental impact”—are based on an engine that was never installed.

The analysis of the generator technology was signed in 2021 for MTU motors. In 2024, Eleanor changed providers. It switched to the Kohler KD3500 due to supply chain issues. They are still Tier 2, but they are a different motor.

With the new motors, Trinity recalculated the model of the spread of contaminants in the air. There was a peak of 3.349 micrograms of nitrogen dioxide, 17 times the limit mentioned at the beginning of this article.

But the statement about the absence of significant environmental impact remained unaltered in the filings. It was written considering a type of motor that was never installed, and the air quality results for the motors that did get installed contradict this claim.

Something worse occurs with the economic argument: They didn’t go back to crunch the numbers. The $16.3 million that ruled out the Tier 4 machines were calculated using MTU prices from 2021. They didn’t go back to account for the Kohler machines. The decision to stay with Tier 2 is based on numbers that were never revisited or were never mentioned in the files.

The technical specifications for the Kohler generators—manufacturer documentation that supports Trinity’s analysis—were also not submitted to the approval process. The Ministry’s technical report explicitly states: “The technical data sheets supporting the emission profiles of the generators when operating at full capacity or at 75% of their rated capacity were not provided.”

Verification was postponed until the system goes live.

A fundamental problem underlies the entirety of the files. To evaluate the air in a place, it must first be measured. Uruguayan regulations, set by the National Directorate of Environmental Quality and Assessment (DINACEA), require three complete years of data with a minimum completion of 75 percent. Without this, there is no baseline.

For this area of Canelones, Trinity Consultants acknowledged that "none of the stations have at least three complete years of data with 75% completeness for NO₂, PM2.5, or PM10," the minimum threshold Trinity itself had established as the validity criterion.

The consulting firm looked to the closest motoring station: Barradas, in Montevideo, 7.7 kilometers away from the project.

But Barradas wasn’t enough: The completeness of the nitrogen dioxide data ranged from 0 to 17 percent for the years with available data. It could only be used for sulfur dioxide in one year.

Facing this insufficiency, Trinity Consultants supplemented itself with data from stations even farther away from Montevideo: Tres Cruces, Ciudad Vieja and Curva de Maroñas. The study itself recognizes that “although they may represent regional trends, the local impacts that they measure are not as representative of local trends at the site of the project.”

Not one of the four stations analyzed meets the criteria for three years of complete data for NO₂ or PM10; this criterion is met only for PM2.5, thanks for the Ciudad Vieja station. For carbon monoxide, the study itself points out that none of the stations in the study measure it. The models were run without that value.

Regulators approved the project with data that did not meet the minimum standard set by the consulting firm and the ministry itself.

Besides the air contamination and its effects on the local population, the files document another problem with regular operations: noise.

The acoustic study was realized by CSI Ingenieros. It evaluated the low frequency noise of the cooling equipment at night. These are low pitch sounds—vibrations between 63 and 125 Hz—that penetrate walls and cannot be muffled out by closing a window. Scientific literature links them to chronic sleep alterations, which raise the risk of high blood pressure, diabetes, autoimmune disease and neurological events.

The low-frequency analysis evaluated five critical residences located between 180 and 280 meters from the cooling equipment. Two to the west, on Fernando Otorgués street on the Colonia Nicolich side; two at the eastern limit of the property, on the Las Higueritas side; and one to the south, the house that the files themselves describe as the data center's "most vulnerable neighbor.

In an emergency, the situation worsens. With the 27 generators running, the nighttime noise at the point with the highest increase would go up by eight decibels: six times the acoustic energy.

The neighbor whom the files describe as "the most vulnerable," at 280 meters to the south, would be exposed to 72 decibels during a nighttime emergency, 12 decibels above the maximum limit.

The Ministry questioned the fact that no mitigating measures had been proposed. As a condition, it required "exhaustive and representative" monitoring during the first month of operation for each module, compliance within residential properties and a protocol for responding to complaints.

Normal operations will be monitored every 45 days. In emergency situations, which project the most vulnerable neighbor receiving 72 decibels, there are no planned measurements. UTE recorded 21 outages between 2001 and 2017, one of which lasted 65 hours.

The problem was put off until the operational phase.

The files reveal a regulatory system that functioned as best it could, despite structural imbalances.

On the company's side: international consulting firms, technical reports hundreds of pages long, legal teams, time. On the regulator's side: a technical team with tight deadlines. The formal process, from the submission of the study on October 26, 2023, until the Technical Report of April 30, 2024, which recommended granting the authorization, took six months. The resolution was signed one month later.

Within that frame of time, the Environmental Ministry secured meaningful corrections. It rejected the emergency threshold proposed by the company. It challenged the noise measurements. But it compromised on central points.

The Ministry approved the project with air quality data that the consulting firm itself recognized as insufficient; without technical specifications of the generators that the company installed; without justifying the choice of Tier 2 over Tier 4; and with substantive conditions deferred for the construction phase and the operational phase—technical specifications of the generators, a detailed emergency plan, an operational environmental management plan and noise monitoring.

Furthermore, the resolution does not stipulate consequences for noncompliance.

It relies on a revealing phrasing for what it does not mention: “It is not expected that the implementation of this project will result in unacceptable residual impacts.” The double negative does not affirm the absence of impacts. It affirms the expectation that, if there are impacts, they won’t be intolerable.

The framework of the study summarizes that logic: of 47 identified impacts, it classifies 39 as “low significance” and only two as “high.”

El agua subterránea no figura como factor ambiental evaluado. En Ciudad de la Costa y su área metropolitana, esto es notable: la napa está a 2,5 metros del sitio. Y en ese mismo sitio se almacenará combustible para 27 generadores, sobre un suelo que el propio estudio clasifica con "potencial expansivo alto" y "fuertes procesos de contracción-retracción".

Neither did the consultant estimate the carbon footprint of a project that will consume up to 560 GWh per year (5 percent of national electricity demand) in exchange for some 50 permanent positions of employment. The calculation was ultimately carried out by the ministry, outside of the impact study, and at the request of citizens during the public comment period. With the initial consumption agreed upon with UTE (420 GWh/year), that amounts to approximately 25,000 tons of CO₂ equivalent per year. The projection for full capacity does not appear in any document in the files.

The Prior Environmental Authorization is based, then, on data that doesn’t meet minimum requirements and on generators whose emissions technology dates to 2006, despite the existence of an alternative that reduces these emissions by 90 percent. What wasn’t resolved during the process to allow construction was put off for the operational stage.

Construction of the data center is almost finished. No one knows if there will be answers to the questions left open in the files. What is known is that, if there are, it will be with the data center in operation.