Project Kilby Emissions and AI Power Costs

Gas plant near data center illustrating Project Kilby Emissions

Project Kilby Emissions are the clearest environmental test attached to Chevron’s move into gas-fired power for AI infrastructure. On June 22, 2026, Chevron’s subsidiary Energy Forge One entered a 20-year power purchase agreement with Microsoft for a co-located natural gas-fired power plant in Reeves County, West Texas, built to serve a Microsoft AI and cloud data center.

The deal changed the technical boundary of data center energy procurement. Instead of buying only from the grid or signing contracts for remote generation, the plan centers on a large behind-the-meter power source tied directly to a compute campus. That structure may give the buyer more certainty over capacity, but it also concentrates emissions, water, control equipment, permitting, and maintenance questions around one site.

As of August 27, 2026, the project had not begun producing power. Chevron projected first power output in 2028, with a phased commercial build-out toward the late 2020s and a final investment decision expected by the end of 2026. The environmental analysis is therefore based on project estimates and reported design choices, not measured operating data.

Project Kilby Emissions Start With Scale

Capacity And Contract Structure

Project Kilby is planned at an estimated 2.67 gigawatts of capacity. That is unusually large for a co-located power and data center development, and the scale matters because environmental costs rise with operating hours, fuel burn, cooling demand, and the amount of backup or control equipment needed to keep the plant reliable.

The 20-year term also matters. A two-decade power contract can help data center planning because AI workloads require dependable electricity, yet the same duration can extend exposure to fossil fuel emissions if low-carbon alternatives become easier to procure during the contract period. A related Abacus News analysis of gas powered AI infrastructure at Project Kilby examined how the power plant, water controls, and phased schedule fit into the project design.

The behind-the-meter model does not remove environmental obligations. It changes where they show up. Grid planners may see less direct load on transmission systems, while local communities and regulators face a large combustion source tied to one compute development. For more insights into this regional energy development and its community effects, related articles can be found through WayLatino, a site within the same publishing network.

Project Kilby Emissions Estimate

Project Kilby Emissions are still estimates because the plant was not operating as of August 27, 2026. The Environmental Integrity Project estimated that Kilby could emit more than 13 million metric tons of carbon dioxide per year, about 3,200 tons of criteria air pollutants, and 278,000 pounds of hazardous air pollutants annually, according to an Environmental Integrity Project filing.

Those figures frame the project as more than a data center siting issue. Carbon dioxide affects greenhouse gas inventories, while criteria pollutants such as nitrogen oxides and sulfur dioxide are local and regional air-quality concerns. Hazardous air pollutants add another layer of review because their effects depend on compound type, concentration, exposure duration, and local dispersion conditions.

The same environmental concern extends to gas build-out in Texas. The research record notes projections that similar new natural gas power plants could increase Texas greenhouse gas emissions by about 13% over 2021 levels and release thousands of tons of smog-forming pollutants and volatile organic compounds. The exact effect would depend on which plants are built, how often they run, and what controls they use.

Controls Reduce Some Pollutants, Not The Carbon Issue

NOx Controls Have A Narrow Job

Chevron’s project design includes Selective Catalytic Reduction to reduce nitrogen oxides. That is relevant because nitrogen oxides contribute to smog formation and are a common focus in air permits for gas-fired plants. The presence of SCR does not, by itself, resolve the carbon dioxide estimate or the hazardous air pollutant estimate.

No carbon capture system has been committed or scheduled for the facility in the research record. That absence is significant because the largest reported number attached to the plant is the annual carbon dioxide estimate. A plant can use emissions controls for specific pollutants and still remain a major source of greenhouse gases.

For AI operators, Project Kilby Emissions shift a portion of the environmental burden from abstract electricity consumption to named infrastructure. That can improve transparency if permitting, monitoring, and reporting are public. It can also make trade-offs harder to avoid because the emissions source is directly linked to the compute campus it powers.

Water Claims Need Local Verification

Chevron has said Kilby will use non-potable, brackish groundwater for cooling rather than freshwater, and closed-loop cooling systems for the data center campus to minimize water use. Those design claims address one of the main concerns around large data centers in dry regions: the risk that cooling demand competes with municipal, agricultural, or ecological water needs.

The claim still needs site-level verification. Brackish groundwater use can reduce pressure on freshwater supplies, but the environmental effect depends on withdrawal volumes, aquifer conditions, water treatment requirements, disposal of concentrated waste streams, and long-term monitoring. The research record does not provide enough detail to quantify those factors.

Closed-loop cooling can lower water withdrawal compared with systems that discharge large volumes after use, but it does not remove all water and energy trade-offs. Pumps, heat rejection equipment, treatment systems, and maintenance schedules affect real performance. Without operating data, the safer assessment is that the design may reduce water stress relative to freshwater-intensive cooling, but the scale of the benefit remains uncertain.

How Chevron Targets Frame The Trade-Off

Intensity Targets Are Not Project-Level Caps

Chevron has reported 2028 Scope 1 and Scope 2 production-intensity targets of about 24 kilograms of CO2 equivalent per barrel of oil equivalent for oil and gas production, with methane intensity near 2 kilograms of CO2 equivalent per barrel and flaring intensity near 3 kilograms of CO2 equivalent per barrel, as described in a Chevron SEC filing.

Those corporate metrics are useful context, but they are not the same as a binding emissions cap for a single gas-fired power plant. Intensity targets can improve per-unit performance while absolute emissions still rise if total energy output grows. For Kilby, the key environmental question is not only whether Chevron reduces methane leaks or flaring intensity across its operations. It is also how much fuel the plant burns, how many hours it operates, and what emissions controls are required under permits.

AI Load Growth Moves The Cost Outside The Data Hall

The research record used for this analysis puts data centers at about 2% of total U.S. electricity use. It also cites Goldman Sachs’ projection that data center growth could drive an extra 3.3 billion cubic feet per day of natural gas demand by 2030. Globally, the research record puts data centers and data transmission at 1% to 1.5% of electricity use and cites an International Energy Agency expectation that the share would double by 2026.

These figures explain why gas-fired AI power projects have drawn scrutiny. The computing equipment inside the data hall is only one part of the system. If AI demand drives new combustion capacity, environmental costs move into turbines, pipelines, cooling systems, emissions controls, and local permitting files. That makes power sourcing a technical design decision rather than a procurement footnote.

Adoption Barriers Are Environmental And Operational

Technicians inspecting cooling and power equipment at an industrial facility

Permitting, Monitoring, And Maintenance

The main barriers are not limited to public opposition. A large gas-fired campus must manage air permits, water sourcing, emissions monitoring, equipment reliability, fuel supply, and maintenance intervals. Each item can affect uptime and environmental performance. A poorly maintained control system can reduce expected pollution benefits, while unexpected operating patterns can change annual emissions from planning estimates.

  • Air emissions: The reported estimates include carbon dioxide, criteria pollutants, and hazardous air pollutants, each subject to different review and monitoring needs.
  • Water use: Brackish groundwater and closed-loop cooling claims require local data before the actual water effect can be judged.
  • Contract duration: A 20-year agreement can stabilize power supply but may extend exposure to gas infrastructure risk.
  • Technology limits: SCR targets nitrogen oxides, while no carbon capture commitment is reported in the research record.

Affected parties include Microsoft as the electricity buyer, Chevron and Energy Forge One as the power project developers, local residents near the plant, Texas air and water regulators, and grid planners assessing how behind-the-meter generation changes regional demand. The risks differ for each group. Microsoft faces emissions-accounting and procurement scrutiny. Chevron faces execution, permitting, and operating risk. Local communities face air and water questions tied to a specific site.

What Project Kilby Emissions Mean For AI Infrastructure

A Measured Reading Of The Environmental Cost

Project Kilby Emissions do not prove that all AI data center growth will depend on new gas plants. They do show that some large AI campuses are being planned around dedicated fossil fuel infrastructure when grid capacity, reliability, and timing become binding constraints. That is the central technical change.

The evidence supports a cautious assessment. Kilby offers a clear power path for a very large data center load, but the environmental cost is material: a multi-gigawatt gas plant, annual carbon dioxide estimates above 13 million metric tons, thousands of tons of criteria pollutants, hazardous air pollutants, water-use questions, and no committed carbon capture system in the research record.

Because the project was not operating on August 27, 2026, final judgment should wait for permit terms, actual equipment configuration, operating hours, fuel supply details, water data, and measured emissions. The deal has already happened. The environmental performance record has not.

Related Articles