Global AI Data Center Permit Clears Weld County, Despite a Deep Trust Deficit

The Global AI data center permit won unanimous approval in Weld County, despite hours of opposition and the company’s earlier construction violations. Commissioners voted 5-0 on September 9, 2026, after an extended public hearing about noise, water, electricity, and transparency.

The decision allows Global AI (Colorado), LLC, to operate a data center inside a vacant industrial facility near Windsor, Colorado. It does not authorize the company’s entire proposed campus or remove the need for construction, utility, safety, and environmental permits.

That distinction defines the real story. Weld County approved a limited land use after adding conditions, while residents questioned whether future oversight could contain a much larger project.

Global AI has discussed an eventual capacity far beyond the initial 27.5-megawatt phase. The county must now enforce its conditions while evaluating any later expansion separately.

What the Global AI Data Center Permit Actually Approves

The vote settled whether Global AI can use this specific property as a data center, not whether every part of its broader plan can proceed.

The Weld County Board of County Commissioners approved Use by Special Review permit USR26-0019. A Use by Special Review permit, commonly called a USR, applies additional review to land uses with greater potential community effects.

The permit covers a data center inside an existing facility in the county’s light industrial zone. The property sits in an unincorporated area near Windsor and has supported industrial operations since the 1960s.

Global AI purchased the 438-acre former Carestream Health campus, previously operated by Eastman Kodak, in 2025. Its current data center proposal covers approximately 41 acres of that larger property.

Reusing the vacant manufacturing facility became a central argument for approval. Commissioner Coordinator Lynette Peppler emphasized that the project would not remove agricultural land from production.

She called the proposal “a really good reuse of resources,” according to the county’s permit decision. That finding helped distinguish the project from greenfield campuses built on previously undeveloped land.

The approval followed a separate September 1 review by the Weld County Planning Commission. That commission voted 7-0 to recommend approval before the county board issued its final land-use decision.

The commissioners nevertheless attached conditions after questioning Global AI and hearing extensive public testimony. Those conditions require closed-loop cooling, continuous noise monitoring, and a decommissioning plan backed by financial security.

A closed-loop cooling system recirculates its cooling fluid instead of continuously consuming water through evaporation. Global AI says this configuration will keep the facility’s water demand low.

The permit also subjects the operation to a 65 dB(C) noise limit at the property line. The C-weighted measurement gives more weight to low-frequency sound, which often drives complaints around industrial equipment.

County rules require evidence that water use satisfies state requirements. Global AI must also obtain a “will serve” letter showing that an electricity provider can serve the approved operation.

The first phase is expected to employ approximately 20 people. That number is modest compared with the workforce once associated with the former Kodak operation.

Commissioner Chair Scott James acknowledged that difference during the hearing. He argued that the project would still generate indirect economic activity beyond its permanent on-site workforce.

None of these findings gives Global AI unrestricted authority over the full campus. The company still needs relevant building, electrical, mechanical, fire, utility, and environmental approvals.

Any material change or later phase requires another county review. The Global AI data center permit is therefore an entry point, not a final authorization for a massive completed campus.

Why a 41-Acre Project Carries Statewide Stakes

The first phase is locally bounded, but its possible expansion would place Global AI among Colorado’s largest electricity users.

Global AI’s initial plan calls for approximately 27.5 megawatts of capacity. That figure describes a substantial industrial load, even before considering the company’s longer-term ambitions.

Previous reporting cited a company representative discussing expansion to as much as one gigawatt. One gigawatt equals 1,000 megawatts, more than 36 times the stated capacity of the first phase.

That scale is not part of the approved permit. However, it explains why nearby residents treated a limited zoning case as a much larger infrastructure decision.

A data center’s computing capacity depends on access to reliable electricity, cooling equipment, network connections, and backup systems. Each expansion can trigger new utility construction and additional regulatory questions.

Global AI would need formal amendments before pursuing the reported one-gigawatt plan. The company has not publicly provided a complete expansion schedule detailing when those amendments would arrive.

The gap between the approved phase and the possible endpoint places pressure on several institutions. Weld County must show that phased review provides meaningful control rather than incremental approval by default.

The electricity provider must determine whether it can serve new demand without compromising reliability or unfairly shifting infrastructure costs. Local governments must also evaluate water, roads, emergency services, and land-use compatibility.

Global AI faces a related challenge. It must demonstrate that the first phase can operate within the promised limits before asking residents to accept a larger footprint.

The approved location offers an advantage because it reuses industrial land and an existing building. Yet an established industrial designation does not answer every question about continuous data center operations.

Traditional manufacturing sites and data centers can create different patterns of power demand, low-frequency noise, cooling activity, and backup generation. Those effects depend on the equipment and operating schedule.

The company’s phase-one design uses air-cooled chillers and a closed-loop system. According to hearing coverage, commissioners made that cooling approach an explicit permit condition.

That condition reduces one prominent concern, but it does not eliminate the project’s other infrastructure requirements. Air cooling can shift attention from water consumption toward electricity use and mechanical noise.

The decision also creates a test for Weld County’s new data center rules. Global AI is the first applicant to pass through the county’s special-review process created for these facilities.

Earlier in 2026, officials delayed final consideration of Ordinance 2026-01 to allow more study and public input. At that time, the county said its zoning code did not specifically define or regulate data centers.

Residents were already raising concerns about water, infrastructure, noise, and surrounding land uses. The county’s ordinance review therefore preceded this specific final vote.

Approval transforms those standards from policy language into an enforcement obligation. Their credibility will depend on measurements, inspections, permit compliance, and transparent responses when limits are approached.

The Central Conflict Is Approval Versus Enforceable Trust

Global AI satisfied the county’s land-use criteria, but its earlier unpermitted work made future compliance the project’s defining test.

During the summer, Weld County issued three stop-work orders involving activity at the site. The county also issued a Notice of Violation and Order to Abate.

Officials said Global AI had installed chillers, piping, and data center-specific equipment without the necessary permits. The company initially appealed the orders before dropping its appeal.

County inspectors later confirmed that Global AI had complied and that no additional construction was occurring. That resolution removed an immediate procedural barrier, but it did not erase the trust problem.

Commissioner Pro-Tem Jason Maxey confronted the company about that history during the final hearing. He asked how commissioners could feel confident approving the application following recent noncompliance.

Global AI Chief Operating Officer Michael Cadigan said the dispute had been resolved. He also invited county officials to contact the company directly if future problems emerged.

Cadigan offered a more revealing acknowledgment when commissioners questioned Global AI’s limited outreach. He accepted that the company should have considered engaging residents earlier.

That admission matters because community engagement is not interchangeable with zoning compliance. A project can satisfy written criteria while leaving neighbors doubtful about future disclosures and operating behavior.

The county board ultimately concluded that the application met its standards. James summarized the governing logic plainly: when an application fits the county code, commissioners approve it.

That reasoning protects land-use decisions from becoming open-ended popularity contests. It also places greater weight on whether the code and attached conditions capture the project’s real effects.

Residents and community groups argued that the review remained too narrow. During the final hearing, public comment reportedly lasted two and a half hours.

Only three of 39 speakers supported the project, according to CPR’s hearing account. Opposition focused on water, electricity, noise, company experience, construction violations, and the perceived lack of transparency.

John DiMatteo, a Johnstown resident and senior software engineer, said profit incentives required strong public pressure. His concern was that the company would otherwise transfer costs or risks to the surrounding community.

Contractor Vance Ray offered the clearest supporting perspective. He described Global AI as a strong customer and argued that the project would create valuable work for local contractors.

Those positions reveal the central tradeoff. Supporters see an opportunity to revive an existing industrial property and attract AI infrastructure investment.

Critics see a developer asking for public trust before disclosing its full growth plan. The earlier stop-work orders gave that concern a concrete foundation.

The Global AI data center permit attempts to bridge this divide through conditions rather than delay or rejection. Continuous monitoring, further approvals, and a decommissioning bond are meant to convert promises into enforceable obligations.

Whether that bridge holds depends on access to information. Monitoring has limited public value if residents cannot understand the results or determine how violations are addressed.

Independent data center researcher Ahmed Saeed has argued that clear rules and transparent public accounting help communities evaluate these projects. His comments appeared in pre-vote reporting.

Transparency also needs to extend beyond permit documents. Residents need understandable information about actual electrical load, noise measurements, water use, backup generation, and requested modifications.

Global AI can strengthen its position by publishing those operational indicators voluntarily. Waiting until the next contested hearing would preserve the same information gap that shaped this approval.

Closed-Loop Cooling Does Not Close Every Resource Question

The cooling commitment addresses water consumption, but electricity, sound, backup generation, and expansion remain separate issues.

Global AI says its closed-loop system will consume less water annually than an average household. That is a company claim, and public reporting has not independently verified actual operating consumption.

Closed-loop cooling can materially reduce ongoing water demand compared with evaporative systems. However, the performance depends on the final design, operating conditions, maintenance, and any later equipment changes.

The permit condition is more important than the general technology description. It commits all phases covered by the current approval to the closed-loop approach rather than leaving cooling design entirely flexible.

Still, residents questioned whether later phases would seek exceptions or introduce different infrastructure. Patricia Garcia-Nelson of GreenLatinos emphasized that the present decision covers only an early phase.

She also raised concerns about potential discharge and the speed of the review. Her position was not opposition to every data center, but support for longer community scrutiny before approval.

The county says applicants must prove that water use satisfies state rules. Global AI is also pursuing annexation discussions with Greeley, which could affect its access to a long-term water supply.

Greeley approved a pre-annexation agreement giving the parties time to negotiate a final arrangement. That process is separate from Weld County’s special-use decision.

Electricity presents a different challenge. The company must provide a utility service letter before proceeding with the relevant development.

That letter would indicate service availability for the approved project. It would not automatically resolve the system effects of a much larger future expansion.

Power questions grow more consequential as computing capacity increases. Servers operate continuously, while cooling and power-conversion equipment add overhead beyond the computing load itself.

Backup diesel generators introduce another layer of uncertainty. Residents and planning commissioners raised concerns about their emissions and noise, especially during testing or grid outages.

The final configuration and permitting of backup systems will determine their practical effect. The zoning vote alone does not establish every technical specification.

Noise has become one of the clearest enforceable metrics. Weld County applies a 65 dB(C) limit at the property boundary and required continuous on-site monitoring.

Low-frequency sound can travel differently from ordinary workplace noise. Residents may perceive persistent humming or vibration even when conventional sound measurements appear moderate.

That makes the measurement method important. Regulators need consistent collection locations, calibrated equipment, defined reporting periods, and procedures for investigating complaints.

The county’s condition requires data to be available upon request. Public confidence would improve if officials regularly disclosed summaries instead of waiting for a dispute.

Decommissioning presents a longer-term risk. Specialized facilities can leave costly electrical, mechanical, and security infrastructure if an operator exits or technology demand shifts.

Commissioners required Global AI to submit a decommissioning plan within six months. The company must also provide a bond, letter of credit, or equivalent financial security for estimated closure costs.

That requirement protects taxpayers from some abandonment costs. Its effectiveness will depend on whether the security amount reflects real removal and restoration expenses.

The approved first phase therefore has several measurable promises. Water use should remain limited, noise must stay within established boundaries, and financial security must cover eventual decommissioning.

These are not abstract aspirations. They are the benchmarks against which the Weld County data center approval should be evaluated.

Local Opposition Is Becoming an AI Infrastructure Constraint

Weld County’s vote shows that available land and electricity no longer guarantee an easy path for new AI capacity.

Data center developers are searching for large properties, transmission access, favorable tax conditions, and communities open to industrial investment. Northern Colorado offers several of those attributes.

Yet AI infrastructure increasingly competes with housing growth, agricultural priorities, industrial users, and household electricity needs. That competition turns technical construction proposals into political decisions.

Near Global AI’s site, neighboring interests pointed to planned residential development west of the property. They argued that a large data center could conflict with thousands of future homes.

The current site has a long industrial history, which strengthens Global AI’s land-use case. At the same time, nearby development plans complicate the claim that past industrial use settles future compatibility.

Commissioners chose conditional approval over rejection. Their decision suggests that adaptive reuse carries significant weight when officials compare data center benefits with land consumption.

That approach could influence other developers. Converting vacant factories may face less resistance than building large campuses on farmland or open space.

However, reuse does not eliminate disputes over electricity. A former manufacturing property can acquire a much larger and more constant power profile after conversion to AI computing.

The employment profile can also disappoint communities expecting a direct replacement for old manufacturing jobs. Weld County expects approximately 20 positions during the first phase.

Construction and electrical contractors can gain substantial project work. Permanent staffing, however, remains relatively limited because modern data centers are highly automated.

James openly acknowledged that the site would not recreate Kodak’s former employment level. He instead emphasized indirect activity and the infrastructure role of data centers.

That argument reflects a broader shift in local economic development. Governments increasingly evaluate data centers through tax revenue, capital investment, and infrastructure positioning rather than headcount alone.

Residents often assess them differently. They compare permanent jobs with power consumption, land use, noise, public subsidies, and potential utility costs.

The result is a political divide that does not follow familiar party lines. Communities across different regions and voting patterns have challenged data center projects over similar resource concerns.

Weld County’s process illustrates why developers cannot treat public communication as a late-stage formality. Technical compliance answers only part of the community’s concern.

People also want to know who ultimately owns the project, which customers will use it, how large it can become, and who pays for supporting infrastructure.

Global AI disclosed enough information to obtain the initial land-use approval. It has not publicly presented a complete roadmap for the potential one-gigawatt buildout.

That missing roadmap gives opponents room to frame phase one as the first step toward a predetermined outcome. Global AI and county officials say later phases will face additional review.

The practical strength of that promise will become visible when the first amendment arrives. Commissioners will then decide whether new growth receives genuinely independent scrutiny.

Other AI infrastructure developers should study this case closely. Early outreach, specific resource estimates, and a credible compliance history increasingly influence project timelines.

A developer that begins work before securing permits creates an avoidable disadvantage. It forces every later promise to compete with evidence that prior rules were misunderstood or ignored.

Global AI has secured its first major local approval. It has not secured the durable community consent needed for a much larger campus.

Three Signals Will Determine What Happens Next

The decisive evidence will come from permit compliance, utility commitments, and the scope of Global AI’s next expansion request.

The first signal is the company’s compliance record during construction. Global AI must obtain the remaining building, electrical, mechanical, fire, utility, and environmental approvals.

Inspectors should confirm that work matches the approved plans and conditions. Another stop-work order would weaken the argument that the summer violations were a resolved procedural mistake.

Clean inspections would support the opposite conclusion. They would show that county enforcement and company management can keep a complex conversion within the approved process.

The second signal is the project’s documented operating footprint. Noise monitoring, water records, cooling performance, and backup-generator permits will reveal whether the initial promises translate into measurable results.

The company’s claim of very low water use deserves particular attention once the facility operates. Actual consumption offers stronger evidence than an engineering estimate or hearing statement.

Continuous noise data will be equally important. Compliance with the 65 dB(C) property-line limit would strengthen the county’s conditional-approval model.

Persistent complaints or disputed measurements would weaken it. They would also test whether residents can obtain monitoring results quickly enough to evaluate their concerns.

The electricity service letter belongs within this second signal. Its scope should show which phase and load the utility has agreed to serve.

A commitment tied only to 27.5 megawatts should not be treated as validation of a one-gigawatt campus. Each later increase needs its own technical and public examination.

The third signal is Global AI’s next application. A request for substantially greater capacity will reveal whether the current project remains a contained industrial reuse.

It will also clarify whether the company intends to move quickly toward the four-phase plan discussed by critics. The scale and timing of that filing matter more than broad statements about long-term growth.

County officials have said future phases require additional review and approval. A transparent review with updated resource data would reinforce that commitment.

A compressed process or broad waiver request would deepen skepticism. It would suggest that phase-one approval created momentum beyond the limited decision commissioners described.

The decommissioning deadline offers an earlier checkpoint. Within six months, Global AI must submit a plan and financial security sufficient to cover estimated closure costs.

Residents should examine how those costs are calculated and updated. A nominal bond can satisfy paperwork without protecting the public from realistic restoration expenses.

The Global AI data center permit has answered one question: this particular use can occupy an existing industrial facility under stated conditions.

It has not answered how large Global AI ultimately intends to become, how expanded power will be supplied, or how monitoring information will reach the public.

For developers, the lesson is direct. AI infrastructure now requires a social and regulatory strategy alongside land, servers, cooling, and electricity.

For local officials, the next task is enforcement. The credibility of Weld County’s first special-review approval rests on what happens after the unanimous vote.

For residents, the most useful response is sustained attention to the measurable conditions. Watch the permits, service commitments, monitoring records, and next expansion application.

Those records will show whether this approval remains a controlled first phase or becomes the opening move toward Colorado’s largest data center.