HARC Data Center Emissions Warning Puts Texas Air Quality Gains at Risk

HARC warns that Texas data center expansion threatens to reverse air pollution reductions, putting electricity supply decisions at the center of the AI buildout. The HARC data center emissions assessment describes a conditional future, rather than an unavoidable outcome. Its central challenge is concrete: expanding computing capacity while preserving gains in cleaner air.

The Houston Advanced Research Center compared different ways of supplying the electricity behind that expansion. The relevant conflict runs between the urgency of energizing new facilities and the environmental consequences of their power sources. A server campus can support digital services while depending on industrial equipment that burns fuel elsewhere or on its property.

That places pressure on developers, electricity suppliers, regulators, and corporate buyers choosing where their computing runs. Google and Microsoft publish clean electricity ambitions, but those ambitions answer only part of the question. Texas must also examine what equipment operates, which pollutants it releases, and who lives nearby.

What the HARC Data Center Emissions Assessment Actually Says

The warning concerns modeled emissions under different development scenarios, not a measured return to earlier air pollution levels.

HARC examined four electricity supply configurations, three demand projections, and five pollutants. Its data center emissions assessment compares possible outcomes with a baseline that excludes new data centers.

Under the projected grid mix scenario, nitrogen oxides would rise 136% by 2031, approaching their 2004 volume in the relevant sectors. Nitrogen oxides are combustion pollutants that contribute to ozone formation.

Volatile organic compounds, chemicals that can participate in ozone formation, would exceed their 2002 volume by 32%. Fine particulate matter, airborne particles small enough to enter deep into the lungs, would return to its 2006 volume.

Those comparisons concern emissions from power generation and industry. They should not be read as identical changes in statewide pollution concentrations or individual exposure.

Holding demand assumptions constant, HARC also found substantial differences between electricity supply configurations. Combustion could produce between two and 10 times the nitrogen oxide emissions of grid supply paired with solar and wind.

That distinction matters when interpreting the headline. A modeled increase under the projected grid mix should not be relabeled as an estimate for exclusively on-site gas generation. These are separate configurations with different assumptions.

The assessment therefore creates a decision problem, rather than establishing a countdown to a fixed environmental outcome. Developers and public officials still influence which modeled pathway becomes relevant.

For readers, the first question should concern the comparison being made. Is a claim discussing historical emissions, future emissions without new facilities, or a particular development scenario? Each reference point answers a different question.

A forecast can be informative without being a prediction of exactly what will happen. Its value comes from identifying consequences that decision-makers can still change.

That is also why a statewide headline cannot resolve a neighborhood dispute. Residents considering a proposed facility need information about its equipment, location, and operating conditions.

The same restraint applies to assigning responsibility. A sector-level scenario does not establish that a particular cloud company caused the projected increase.

The immediate news is that electricity procurement belongs inside the assessment of computing expansion. Evaluating buildings and servers alone leaves a central environmental dependency outside the frame.

Texas Data Center Pollution Creates a Local Health Question

The people exposed to electricity-related pollution are not necessarily the people buying the computing services.

The American Lung Association identifies data center pollution pathways through both electricity generation and backup equipment. Its factsheet connects combustion-related pollutants with respiratory and cardiovascular harms.

It also distinguishes regional power generation from generators installed at facilities. That separation matters because an electricity connection does not eliminate pollution associated with producing electricity.

A facility’s environmental footprint therefore extends beyond its fence. Depending on its supply, the associated emissions can occur at a distant power plant, nearby generators, or both.

For a community reviewing a proposal, this creates an information problem. A drawing of the campus does not necessarily show the infrastructure supplying its electricity.

Consider a hypothetical enterprise moving document processing into a new cloud region. The procurement team sees computing capacity, service availability, and contractual commitments.

A neighboring household sees the physical development supporting those services. Its questions concern equipment operation, air monitoring, and what happens when normal electricity supply becomes unavailable.

Both perspectives describe the same computing service. Their priorities differ because the benefits and potential burdens reach different people.

This is why Texas data center pollution deserves more specific reporting than a statewide total can provide. An emissions inventory measures releases; exposure concerns what people breathe.

Converting between those concepts requires additional information. Weather, atmospheric chemistry, source locations, and operating patterns affect how emissions translate into local conditions.

A projected emissions increase should therefore trigger further analysis without becoming an unsupported claim about individual illness. The available evidence does not justify assigning a precise health outcome to every proposed campus.

The distinction also works in the other direction. Uncertainty about a neighborhood’s eventual exposure does not make the underlying emissions irrelevant.

Local officials face the practical task of deciding what evidence must arrive before development proceeds. Waiting until a facility operates can leave communities debating impacts after major investment decisions have hardened.

Useful disclosure would connect the computing project with its electricity arrangements. Residents should be able to understand whether nearby equipment supplies routine operations, emergency backup, or another purpose.

Those categories affect the questions worth asking. Backup equipment needs scrutiny about testing and emergency use; routine generation raises questions about sustained operation.

For corporate customers, this becomes a reputational and operational issue. An infrastructure partner’s relationship with its host community can matter throughout the service contract.

Buyers cannot resolve every local dispute through procurement. They can, however, avoid treating environmental information as unrelated to the service they purchase.

The pressure falls on developers to explain their plans, suppliers to substantiate their claims, and public agencies to make comparisons understandable.

Data Center Power Choices Matter More Than a Clean Energy Label

A clean energy commitment needs a location, a timeframe, and an explanation of what happens when clean supply is unavailable.

Google provides a useful example of the distinctions buyers can examine. Its hourly carbon-free energy documentation describes a regional metric based on electricity supply and attributed clean generation.

The company says its goal is to match consumption with carbon-free energy in every region during every hour by 2030. Its published regional information gives customers a way to consider electricity characteristics alongside application requirements.

Microsoft’s electricity matching goal likewise describes matching electricity consumption with zero-carbon energy purchases by 2030. The company also reports separate measures for energy and water efficiency.

These disclosures concern related environmental dimensions, but they should not be treated as interchangeable evidence. Energy efficiency, clean electricity matching, water consumption, and local air emissions answer different questions.

An efficient building can still consume electricity produced through combustion. A facility with clean electricity purchases can still require backup equipment with its own emissions profile.

Likewise, a carbon-focused metric does not provide a complete inventory of pollutants affecting local air quality. Carbon dioxide accounting and nitrogen oxide control require different evidence.

This does not make corporate energy commitments meaningless. It establishes what additional information a buyer needs before drawing a broader conclusion.

Google’s regional approach offers one comparison point: where electricity is consumed matters. Microsoft’s separate efficiency reporting offers another: reducing operational overhead is distinct from changing the electricity source.

Neither company’s published commitment establishes the eventual emissions of every Texas project. Applying a corporate target to an individual facility requires evidence about that facility.

The commercial tension becomes clearer when a developer must choose an electricity arrangement. Capacity arriving sooner can help meet a customer’s deployment schedule.

Environmental performance may depend on different conditions, including the generating technology and its operating requirements. The procurement decision needs to account for both.

A hypothetical AI application illustrates the problem. Its interactive service may need immediate responses, while some evaluation or batch processing can tolerate scheduling flexibility.

That flexibility creates options worth examining. It does not establish that shifting workloads will automatically reduce emissions.

The effect depends on the electricity available when and where the work moves. Moving computing without checking those conditions can simply relocate the associated burden.

Data center power choices should therefore be described through operating evidence. A claim becomes more useful when it identifies the supply arrangement and the period it covers.

Buyers should also distinguish an announced project from an operating resource. A future clean electricity agreement cannot establish the current emissions of equipment running today.

The relevant question is whether the electricity plan supports the computing schedule under real conditions. Aspirations help define direction; operating records show what actually supplied the load.

The AI Buildout Puts Electricity Planning on the Critical Path

Computing capacity depends on an electricity system whose constraints cannot be solved through software deployment alone.

The Department of Energy’s electricity demand assessment reported that data centers used about 4.4% of US electricity in 2023. That assessment projected a range of 6.7% to 12% by 2028.

Those figures come from the department’s published 2024 assessment. They provide a dated national reference, rather than a current count of Texas projects or a facility-level forecast.

Their relevance is the scale of the planning question. Computing demand must be translated into physical supply, delivery infrastructure, and operating arrangements.

For an AI company, a deployment milestone can appear as a product decision. For its infrastructure provider, that milestone also depends on whether sufficient electricity reaches the servers.

A delay or change in that supply can alter the development plan. The consequences extend to customers who never negotiate directly with a utility.

This makes electricity sourcing part of service delivery. Enterprise buyers evaluating capacity commitments should understand the assumptions behind promised availability.

A provider might describe a campus’s maximum planned demand. That figure alone does not establish when the campus will operate or how much electricity it will consume.

Capacity describes a potential rate of use. Energy consumption reflects use over time, so the terms should not be substituted for each other.

The distinction matters when interpreting dramatic project announcements. A proposed campus, a connection request, a completed building, and an operating facility represent different stages.

Adding every announced maximum together can obscure those differences. Treating every proposal as irrelevant would also miss infrastructure decisions already moving forward.

A more useful approach tracks project maturity alongside power arrangements. That helps distinguish a speculative ambition from a commitment requiring near-term supply.

The same discipline applies to claims about AI efficiency. A more efficient model can reduce electricity needed for a particular task under comparable conditions.

It does not establish the direction of total electricity use. Deployment volume, workload complexity, and the number of tasks also matter.

Developers therefore need separate measures for task efficiency and aggregate consumption. Improvement in one measure should not be used as evidence about the other without supporting data.

Enterprise customers face a related choice. Evaluating only model quality and response speed leaves infrastructure dependencies outside the purchasing decision.

A service can meet an application benchmark while relying on an electricity arrangement that faces community opposition or unresolved development conditions. These issues belong in a broader assessment of delivery risk.

For knowledge workers using AI products, the connection is less visible but still relevant. The interface hides the physical system supporting each service.

The practical implication is to ask for evidence at the appropriate level. Product claims require product evidence; facility and electricity claims require infrastructure evidence.

What the Projections Cannot Establish

The HARC data center emissions warning supports scrutiny, but it does not settle every argument about development, exposure, or responsibility.

The central uncertainty concerns which facilities become operational and how their electricity demand is supplied. A scenario describes consequences under assumptions; it does not guarantee those assumptions will hold.

That distinction protects the analysis from opposite errors. Treating a high-emissions pathway as inevitable overstates the evidence, while dismissing all conditional modeling ignores its planning value.

The strongest skeptical question is whether the assumptions match the projects likely to proceed. That requires examining actual development decisions as they become clearer.

A project might change its schedule, scale, electricity source, or equipment. Each change can affect which comparison remains relevant.

Those possibilities should be tracked rather than invoked as a general reason to disregard the warning. Uncertainty becomes useful when it identifies evidence that would strengthen or weaken a conclusion.

Another limitation concerns the boundary of the claim. Sector-wide emissions estimates do not automatically describe conditions at a specific school, workplace, or home.

Local assessment requires evidence that connects sources with nearby conditions. Without that step, a precise neighborhood claim would exceed the available analysis.

The same boundary applies to individual companies. Industry expansion involves many operators, customers, suppliers, and infrastructure arrangements.

A study about the sector does not prove that Google, Microsoft, or another named company follows a particular modeled pathway. Company-specific conclusions require company-specific evidence.

Supporters of development can reasonably point to the need for reliable computing services. An electricity arrangement that cannot support the intended workload creates a real operational problem.

That argument establishes a requirement, rather than resolving the choice of supply. Reliability claims should explain the equipment and operating conditions needed to deliver it.

Critics should face a similar standard. A cleaner alternative must be evaluated for its actual availability and ability to serve the intended facility.

An option that exists in principle may not match a particular location or schedule. Establishing those constraints is more informative than assuming either universal availability or impossibility.

The analysis should also avoid treating all combustion equipment as identical. Fuel, generating technology, controls, and operating hours can change the relevant emissions profile.

Likewise, a renewable electricity claim needs enough detail to assess its relationship to consumption. A broad label cannot substitute for the underlying arrangement.

These qualifications sharpen the core concern. The public needs evidence that allows different electricity plans to be compared on consistent terms.

For readers evaluating Texas data center pollution, the most credible account will explain both the modeled risk and the limits of inference. Precision helps communities challenge weak proposals without overstating what the research establishes.

Three Signals Will Show Which Direction Texas Takes

The next useful evidence will connect proposed computing capacity with disclosed power arrangements and enforceable operating conditions.

The first signal is Texas’s forthcoming review of data center impacts. The Electric Reliability Council of Texas, the state’s grid operator, has published a data center review process covering developers pursuing grid connections.

ERCOT requested information from projects with demand of at least 25 megawatts. A megawatt measures electrical power, rather than electricity consumed over a period.

The submission deadline was October 12, 2026. ERCOT says it intends to work with the Public Utility Commission of Texas on a report published by December 10, 2026.

The requested information includes reliance on grid electricity, on-site generation, water use, cooling, public assistance, and community impacts. Those categories can help connect development plans with their supporting infrastructure.

The report’s usefulness will depend on the detail made public. A summary that distinguishes electricity arrangements would help readers assess which risks deserve closer examination.

If the review reveals extensive reliance on routine on-site combustion, it would strengthen concern about that development pathway. Detailed evidence of cleaner supply would support a different assessment.

The second signal is what project applications and permits specify. Readers should look for the generating equipment, authorized operation, and pollution-control requirements attached to individual facilities.

An announcement about a data center opening does not supply those details. Nor does a general promise that the project will meet applicable requirements.

The useful comparison concerns what a facility is authorized to operate and what it ultimately runs. Those records can test whether public descriptions match the infrastructure plan.

Evidence of sustained combustion requirements would reinforce the need for scrutiny. Evidence of lower-emitting arrangements, supported by operating conditions, would weaken claims that the worst pathway applies universally.

The third signal is whether corporate disclosures become more specific about actual supply. Regional reporting and facility information can help buyers evaluate electricity claims with greater confidence.

Google and Microsoft provide examples of metrics that distinguish aspects of environmental performance. The next question is how clearly customers can connect those metrics with the services they use.

A commitment becomes more informative when reporting explains its boundary, period, and relationship to consumption. Missing detail leaves customers unable to evaluate whether a broad claim applies to their deployment.

None of these signals alone will determine Texas’s future air quality. Together, they can narrow the gap between development announcements and observable operating decisions.

For developers and enterprise buyers, the immediate action is to examine the electricity assumptions behind the next capacity commitment. Ask what will supply the facility when it opens, how that supply changes over time, and which disclosures support the answer. Keep local air pollution separate from carbon accounting, while evaluating both. For readers following the HARC data center emissions warning, compare the December review with project records and subsequent operating disclosures. The question is whether the infrastructure being selected supports cleaner growth under actual conditions. Follow the equipment, the electricity supply, and the evidence that connects them.