North Texas Wants the AI Economy. Fort Worth Is Asking What the Infrastructure Will Cost

In Brief

• The binding constraint on AI infrastructure in North Texas has moved from hardware availability to physical and civic capacity: electrical interconnection, water and cooling design, land use, noise, and the durability of community consent.
• Fort Worth is working through a specific set of proposed standards covering zoning, noise, water and economic development incentives, with City Council action scheduled for August 11, 2026. The City has stated it is not pursuing a moratorium, and has published its reasoning.
• The transferable lesson for executives is sequencing. Organizations that plan capacity around utility interconnection timelines and local approval processes get capacity sooner than those that treat compute as a procurement decision.

Executive Summary

The public conversation about artificial intelligence is largely conducted in software terms — models, tokens, capabilities, adoption. The conversation that determines whether any of it can scale in North Texas is conducted in different units entirely: megawatts, gallons, acres, decibels and months of interconnection queue.

Fort Worth is currently working through that second conversation in public, and doing so with unusual specificity. The City has published proposed amendments to its zoning, noise and water regulations, along with changes to how it evaluates economic development incentives for data center projects. The proposals are concrete: a 250-foot building setback from residential districts and uses, standby generators kept 300 feet from residential and fully screened, a 72-hour ambient noise study before development within 250 feet of residential, a requirement that new data centers use closed-loop cooling, wastewater pretreatment permitting, and a $500 million minimum initial investment with an identified end user before incentives are considered.

Read as local news, this is a zoning story. Read as infrastructure analysis, it is something more useful: a detailed, publicly documented account of what a growing American city concludes it must regulate when large compute arrives, and what falls outside its authority entirely.

That second category matters as much as the first. Fort Worth's own material is explicit that the City's authority covers four areas — zoning, noise, water utility requirements and City economic development incentives — and that grid access, interconnection and reliability sit with the Public Utility Commission of Texas and ERCOT, statewide water planning with the Texas Water Development Board, and air quality with the TCEQ. A city can shape where a facility sits, how loud it is and how it uses municipal water. It cannot determine whether the facility gets power.

For executives across Dallas–Fort Worth, this is not primarily a story about data centers. It is a story about what has to be true before AI capacity exists, and about the difference between a region that has capacity and an organization that can obtain it.

Direct Answer

Why is AI infrastructure becoming a power, water and land-use question rather than a computing question? Because the computing hardware is available for purchase while the physical conditions it requires are not. Large AI facilities need substantial, continuous electrical service, and in Texas that service depends on interconnection processes governed by ERCOT and the Public Utility Commission rather than by any municipality or developer. They need cooling, which is a water-design question with materially different consumption profiles depending on the approach. They need land positioned to satisfy setback, noise and screening requirements that cities are actively revising. And they need the durable consent of the communities that host them, which is not a permit but is functionally a prerequisite. The result is that timelines are now set by utilities and local governments rather than by equipment lead times — CBRE's own market commentary notes that grid interconnection timelines in Dallas–Fort Worth have not materially improved over the past six to twelve months, and that extended rezoning and entitlement timelines are a factor site selectors weigh when choosing between DFW and West Texas. The practical implication for any organization planning significant AI capacity is that the critical path runs through infrastructure and public process, and should be planned as such rather than discovered late.

Executive Summary Table

Why Fort Worth's Debate Matters Beyond Fort Worth

Fort Worth is not a marginal participant in this market. The City reports four existing data center developments spanning ten buildings across Council Districts 5 and 10, with a combined footprint of more than three million square feet. Individual facilities range from 209,000 square feet to a 1.5-million-square-foot campus. The first was built in 2000; the most recent completed construction in 2025. One more facility is under construction in District 7, four additional projects have been proposed across Districts 3, 6, 7 and 8, and two more have been proposed in the City's extraterritorial jurisdiction.

This is a city with two decades of operating experience with the asset class, currently absorbing a substantial pipeline. Its conclusions carry weight precisely because they are not reflexive.

The regional context is equally concrete. CBRE reports Dallas–Fort Worth became North America's third-largest colocation market in the first quarter of 2026, with inventory up 43.7% year over year to 1,249.4 megawatts and a record 716.7 megawatts under construction that was 88% preleased. The market's vacancy rate fell to a record-low 1.8%.

The number that best explains the current civic conversation, though, is a different one: despite adding 379.9 megawatts of inventory year over year, DFW's available supply rose by just 1.4 megawatts. Nearly everything built is spoken for before it opens. That is a market in which demand pressure on land, power and water will continue regardless of how any single city votes — which is exactly why the terms of development are being negotiated now.

AI Has a Physical Footprint

The abstraction that causes the most planning error is the idea that compute is delivered rather than built. Every unit of AI capacity corresponds to equipment in a building on land, drawing power from a grid and rejecting heat into either air or water.

Electrical service is the governing constraint. Large facilities require substantial continuous load, and connecting that load requires transmission planning and interconnection approval. In Texas this is explicitly outside municipal control. Fort Worth's own material states plainly that the City has no direct authority over whether or how a data center connects to the power grid. Texas addressed the broader question at the state level: Senate Bill 6, passed during the 89th Legislature and effective June 20, 2025, established a framework for managing the growth of large electricity consumers, addressing grid reliability, cost allocation, transparency and the protection of residential customers during grid emergencies. In the DFW market specifically, CBRE notes that Oncor has been conducting data center load cluster studies to determine a path toward grid interconnection.

Cooling and water vary enormously by design, and the variance is the point. Fort Worth published estimates normalized to a 100-acre example: a facility using closed-loop cooling averages roughly 300,000 gallons to one million gallons per month, while one using evaporative cooling averages four to six million gallons per month. For scale, the City notes that a single-family residential development on 100 acres uses approximately two million gallons per month. A closed-loop facility therefore uses meaningfully less municipal water than housing on equivalent land; an evaporative facility uses two to three times more. This is why the City's proposal to require closed-loop cooling for new data centers is substantive rather than symbolic — it is a design mandate that changes the water profile by roughly an order of magnitude.

The supply context behind that proposal is long-range. Tarrant Regional Water District is the sole and exclusive provider of raw water to Fort Worth, under a contract that prioritizes supply for Fort Worth alongside Arlington, Mansfield and the Trinity River Authority. The State Water Plan projects TRWD will need additional supply over the next fifty years, with the $2.3 billion Integrated Pipeline project as a cornerstone and indirect reuse projects at Mary's Creek and Cedar Creek Reservoir in design.

Land, noise and screening are where abstraction meets a neighbor's back fence. Fort Worth's proposed zoning amendments would establish data-center-specific standards including a 250-foot building setback from residential districts and uses, standby generators kept at least 300 feet from residential and fully screened, rooftop cooling equipment placed behind an acoustic barrier one and a half times its height, a 50-foot landscape bufferyard with a screening fence along residential, and a continuous row of evergreen trees at the site perimeter adjoining residential. The proposals would also prohibit cryptocurrency mining.

On noise, the current position is that industrial zoning districts are exempt from maximum sound levels. The proposed amendments would require, within 250 feet of residential, a 72-hour pre-development ambient noise study measured at the property line, then establish maximum sound levels not to exceed 5 dBA above that pre-development ambient, with mitigation at the owner or operator's expense if exceeded. Standby generator testing would be restricted to weekdays between 8 a.m. and 5 p.m., with emergency operations exempt.

That noise construction deserves attention because it is a genuinely thoughtful piece of regulatory design. Rather than setting a fixed decibel ceiling, it measures the site before development and holds the operator to a defined increment above what was already there. It is a standard that adapts to context, and it creates an obligation to establish a baseline — which is a discipline worth borrowing well outside the data center context.

The Infrastructure Costs That Do Not Appear in a Model Demonstration

A capability demonstration shows a result. It does not show the substation, the interconnection queue position, the water loading study, the wastewater pretreatment permit, the entitlement calendar or the three public meetings.

Fort Worth's proposed water regulations would require new data centers to obtain wastewater pretreatment permits as a Non-Significant Industrial User. The City's material describes what such permits allow it to require and enforce: chemical disclosure and Safety Data Sheet submission, spill prevention and slug control plans, secondary containment, monitoring and inspection access, discharge limitations, annual certifications, operator training, reporting requirements and emergency notification procedures.

Read that list as an operations executive rather than as a regulatory subject and something becomes clear: it is a description of a managed industrial facility with disclosure obligations, tested procedures and evidence requirements. That is a meaningfully different operating posture than "a building full of servers," and it carries staffing, documentation and compliance costs that belong in a capacity model.

The economic development side has been tightened in parallel. Fort Worth's proposed policy changes would require a project to demonstrate multiple viable sites under consideration, show substantial progress through the ERCOT interconnection process before final consideration, significantly exceed the policy investment minimum with a $500 million minimum initial investment and an identified end user, maintain a minimum business personal property value throughout the life of the agreement, and would limit grants or abatements to no more than 50% of tax revenue.

The ERCOT-progress requirement is the most instructive item in the entire package. It says, in effect, that the City will not underwrite a project whose fundamental constraint has not been addressed. Speculative announcements are not investment.

Economic Opportunity and Community Constraint

It would be a mistake to read any of this as opposition. Fort Worth's own published figures make the economic case in its own numbers: data centers have generated more than $83 million in gross property tax revenue for the city over the last five years. In 2024, gross property tax revenue from data centers equaled what the city collects from roughly 8,300 average single-family homes. Based on known in-progress development, the City projects that figure growing to the equivalent of more than 22,700 homes by 2030.

That is a substantial and durable public revenue base from a land use that consumes comparatively little in municipal services — no school enrollment, minimal traffic generation, low emergency call volume relative to footprint. Cities that host this infrastructure well are funding libraries, streets and fire stations with it.

The two facts sit together without contradiction. Data centers deliver real fiscal benefit. They also impose real, local, physical externalities on the specific people who live near them, and those people vote. A framework that acknowledges both is more durable than one that acknowledges only the side it prefers, because the alternative — benefit asserted, impact dismissed — is precisely what converts a routine approval into an organized opposition campaign.

This is also why the moratorium question deserves precision. The City has published its position directly: staff is not recommending a moratorium, because the City is already moving forward with regulations scheduled for consideration and potential adoption by August 11, and initiating a moratorium would add procedural steps and delay adoption by two to three months. The City's analysis notes that under the Texas Local Government Code a moratorium requires specific notice, public hearings and written findings, runs an initial 90 days with a possible 90-day extension, cannot regulate matters outside the City's authority or prohibit a lawful land use, and does not apply to vested projects, approved projects or those with zoning applications already submitted. By the City's timeline, the earliest a moratorium could take effect would be mid-to-late October, with regulations adopted at the end of the moratorium period in January 2027 — roughly five months later than the proposed August adoption.

That is a considered argument that a moratorium would slow the regulation rather than accelerate it. Executives following this issue should be careful not to import the assumption that regulation and moratorium are the same instrument. In this case the City's position is that they point in opposite directions.

The Metro Relay AI Infrastructure Accountability Test

Discussions about AI infrastructure tend to collapse into a binary — for or against — that serves no one making an actual decision. The following five questions are designed to be answerable by any party at the table: an operator, a large capacity buyer, a municipal leader, a utility, a developer or an enterprise executive evaluating a commitment. They do not presuppose an answer. They establish whether the parties know what they are agreeing to.

1. What capacity is required? Stated in physical units and over time, not in aspiration. Peak and continuous electrical load. Cooling approach and its water profile. Land area including setbacks and buffers. Expected growth and the trigger conditions for each expansion phase. A requirement that cannot be stated in these terms is not yet a plan.

2. Which public and private resources support it? Which grid, which interconnection pathway, which position in which queue. Which water system, under which contract, with what long-term supply plan behind it. Which roads during construction. Which emergency services. The question is not whether the resources exist but whether they have been allocated and by whom.

3. Which risks remain with the operator? Backup generation and its fuel supply. Redundancy. Failure modes and their consequences. The cost of mitigation if noise or discharge limits are exceeded. Risk that has not been assigned has not been managed; it has been deferred to whoever is standing closest when it materializes.

4. Which impacts are transferred to the community? Sound, at what level and at which hours. Traffic during construction and operation. Visual impact and screening. Water and wastewater loading. Effect on grid capacity available to other customers. Naming these plainly is not an admission of fault. It is the precondition for addressing them.

5. What evidence will demonstrate responsible operation? Not commitments — evidence. Monitoring data, and who can see it. Reporting cadence, and to whom. Independent verification of any claim material to approval. A complaint pathway with a defined response time. Fort Worth's proposed 72-hour pre-development ambient noise study is a clean example of evidence-first design: it creates a measurable baseline that makes a future dispute resolvable by data rather than by assertion.

The test's value is that it works in both directions. An operator that can answer all five is positioned to move faster through public process, because the questions a council will ask have already been answered. A municipality that asks all five is positioned to approve good projects confidently rather than delay everything defensively. And an enterprise executive who applies it to a capacity commitment learns quickly whether the counterparty has a project or a press release.

Common Misconceptions

"Data centers consume unlimited water." Consumption depends on cooling design and varies by roughly an order of magnitude between approaches. Fort Worth's own normalized estimates put closed-loop cooling below the municipal water use of residential development on equivalent land, and evaporative cooling well above it. The design choice is the variable, which is why it is now the regulatory target.

"A city can decide whether a data center gets power." It cannot. Grid access, interconnection and reliability are governed at the state level through ERCOT and the Public Utility Commission of Texas. Municipal authority covers zoning, noise, water utility requirements and local economic development incentives. Conflating these produces both misdirected advocacy and misdirected planning.

"Regulation means a region is closed for business." The proposals under discussion in Fort Worth set conditions, not prohibitions — with one narrow exception for cryptocurrency mining. Clear standards frequently accelerate development by making outcomes predictable, which is worth more to a developer with capital at risk than an ambiguous process with a theoretically lower bar.

"A moratorium is the fastest way to get regulations." Fort Worth's published analysis concludes the opposite in its specific circumstances: that a moratorium would add procedural steps and delay adoption of the regulations by two to three months.

"Announced capacity is available capacity." It is not. Announcement, entitlement, interconnection, construction and energization are distinct milestones separated by years. A press release is not a megawatt.

Leadership Considerations

For executives whose AI plans depend on capacity — whether purchased, leased or built — three considerations follow.

The first is timeline realism. If the regional constraint is interconnection and entitlement rather than equipment, then capacity plans should be built backward from utility and municipal milestones. CBRE's observation that DFW interconnection timelines have not materially improved in six to twelve months is not a market color detail; it is a planning input.

The second is that most organizations should not be in this business at all. The physical, regulatory and operational obligations described here are appropriate for operators and hyperscale buyers. For the substantial majority of mid-sized North Texas organizations, the correct answer is to consume capacity from providers who have solved these problems, and to spend internal effort on readiness to use AI rather than on the infrastructure to host it.

The third is that public trust behaves like a supply constraint. Communities that feel informed and heard approve projects faster than communities that feel managed. Any operator's engagement strategy is, functionally, part of its capacity strategy.

Metro Relay Perspective

The most useful reframing available right now is that AI infrastructure has become an accountability problem rather than a technology problem. The hardware works. The models work. What remains genuinely difficult is establishing, in public and in advance, who bears which cost and who can verify which claim.

Fort Worth's process is instructive not because its specific answers will be right for every jurisdiction, but because it is being conducted with published documentation, defined dates and stated reasoning. That is the standard the industry should want, for a practical reason: opaque processes generate opposition that is expensive to resolve, while transparent ones generate conditions that can be planned around.

We would also note the asymmetry in how this debate is usually framed. Operators are asked to prove their impacts are acceptable. Communities are rarely asked to state what evidence would satisfy them. The second question is as important as the first, and the Accountability Test is deliberately constructed so that both parties have to answer.

Strategic Recommendations

Plan capacity around interconnection and entitlement timelines rather than equipment availability, and treat utility engagement as a first-order workstream with a named owner rather than a procurement follow-up.

Specify cooling design early, because it determines water profile, siting flexibility and regulatory exposure simultaneously. It is the highest-leverage single decision in the physical design.

Engage municipal and community stakeholders before opposition organizes, with specific answers to the five accountability questions. Engagement after a project becomes controversial costs more and achieves less.

Build the evidence layer into the design rather than retrofitting it. Monitoring, baselines and reporting are far cheaper to design in than to add under a consent condition.

For the majority of organizations: buy capacity, do not build it. Direct the effort toward the internal readiness that determines whether purchased capacity produces value.

Track the August 11, 2026 Council action and the state legislative interim work on data centers, both of which will shape the conditions under which North Texas capacity is developed over the next several years.

Future Outlook

Three developments look reasonably likely across North Texas over the next several years, stated as expectations rather than certainties.

Municipal standards will converge and become more specific. Fort Worth's framework — setbacks, ambient-relative noise limits, cooling mandates, pretreatment permitting, tightened incentive thresholds — provides a template that neighboring jurisdictions in Tarrant, Denton, Collin and Dallas Counties will likely adapt rather than invent independently.

Interconnection will remain the binding constraint, and the projects that move will increasingly be those that addressed it first. Fort Worth's proposed requirement that incentive applicants demonstrate substantial ERCOT progress formalizes what the market already prices.

The state will legislate further. Fort Worth's own material notes that Texas House and Senate committees are studying data-center-related interim charges covering grid impacts, water supply, infrastructure, economic development and regulation, and anticipates legislative action in the next session. Senate Bill 6 established a framework for large loads; the interim work suggests refinement is coming.

Conclusion

The question in the title has a specific answer, and Fort Worth has done the work of writing it down: AI infrastructure costs setbacks and screening, an ambient noise baseline and a mitigation obligation, a cooling design that determines a water profile, a pretreatment permit with disclosure and containment requirements, a position in an interconnection queue that no city controls, and a $500 million threshold with an identified end user before public incentives enter the conversation.

None of that makes AI infrastructure a bad thing to host. Fort Worth's own revenue figures argue the opposite. What it makes AI infrastructure is a physical, civic undertaking with real costs that have to be assigned rather than assumed — and the assignment is more durable when it happens in public, in advance, with evidence attached.

For most executives across Dallas–Fort Worth, the practical takeaway is narrower and more immediate. The region is building substantial capacity. Your organization will almost certainly buy that capacity rather than build it. What determines whether it produces value is not proximity to a data center but whether your own processes, data and governance are ready to use it — which is a different problem, with a different solution, and one entirely within your control.

Key Takeaways

• The constraint on AI infrastructure in North Texas is physical and civic — interconnection, water, land, noise and consent — not hardware availability.
• Fort Worth has published specific proposed standards covering zoning, noise, water and incentives, with Council action scheduled for August 11, 2026.
• The City has stated it is not pursuing a moratorium, on the reasoning that a moratorium would delay the regulations by two to three months rather than accelerate them.
• Cooling design drives water consumption by roughly an order of magnitude, which is why closed-loop requirements are a substantive regulatory lever.
• Municipal authority covers zoning, noise, water utility requirements and local incentives. Grid access and interconnection sit with ERCOT and the PUCT.
• DFW added 379.9 MW of inventory year over year while available supply rose just 1.4 MW — nearly everything built is preleased.
• Use the AI Infrastructure Accountability Test to establish what capacity is required, which resources support it, which risks stay with the operator, which impacts transfer to the community, and what evidence will demonstrate responsible operation.
• Most organizations should buy capacity rather than build it, and invest their effort in the readiness to use it well.

Frequently Asked Questions

What infrastructure does a large AI data center actually require? Substantial continuous electrical service with an approved grid interconnection, a cooling system with a defined water consumption profile, land sufficient to satisfy setback and screening requirements, wastewater capacity and permitting, and — functionally, if not legally — durable community consent. The computing equipment is the most purchasable item on that list and the least likely to determine the schedule.

How much water does a data center use? It depends almost entirely on cooling design. Fort Worth published estimates normalized to a 100-acre example: closed-loop cooling averages roughly 300,000 gallons to one million gallons per month, while evaporative cooling averages four to six million. A single-family residential development on the same 100 acres uses approximately two million gallons per month. The design choice moves the figure by about an order of magnitude, which is why it has become a regulatory target rather than an engineering footnote.

What is closed-loop cooling and why does it matter? A cooling approach that recirculates the same water rather than consuming it through evaporation. It matters because it changes a facility's municipal water profile dramatically — enough that a closed-loop facility can use less water than housing on equivalent land. Fort Worth's proposal to require closed-loop cooling for new data centers is a substantive design mandate rather than a symbolic gesture.

Who decides whether a data center can connect to the power grid in Texas? Not the city. Grid access, interconnection and reliability are governed at the state level through ERCOT and the Public Utility Commission of Texas, with the transmission and distribution utility conducting the technical studies. Fort Worth's own material states plainly that the City has no direct authority over whether or how a data center connects to the grid. Municipal authority covers zoning, noise, water utility requirements and local economic development incentives.

What is Senate Bill 6? Legislation passed during the 89th Texas Legislature and effective June 20, 2025, establishing a framework for managing the growth of large electricity consumers. It addresses grid reliability, cost allocation, transparency, and the protection of residential customers during grid emergencies. It is the state-level answer to a question cities are frequently asked but cannot resolve.

What is Fort Worth proposing for data center regulation? A package covering zoning, noise and water. The zoning amendments would set data-center-specific standards including a 250-foot building setback from residential districts, standby generators at least 300 feet from residential and fully screened, acoustic barriers for rooftop cooling equipment, a 50-foot landscape bufferyard, and a prohibition on cryptocurrency mining. The noise provisions would require a 72-hour pre-development ambient study and cap sound at 5 dBA above that measured baseline. The water provisions would require closed-loop cooling for new facilities and a wastewater pretreatment permit. City Council consideration is scheduled for August 11, 2026.

Is Fort Worth pausing data center development? No. Staff is not recommending a moratorium. The City's published position is that a moratorium would add procedural steps under the Texas Local Government Code and delay adoption of the regulations by two to three months, pushing adoption from August 2026 to roughly January 2027. Anyone tracking this should verify current status directly, since Council action was scheduled after this analysis was prepared.

Why would a data center need a wastewater pretreatment permit? Because cooling systems discharge treated water, and permitting gives the utility enforceable authority over what enters the system. Fort Worth's material describes what such a permit allows it to require: chemical disclosure and Safety Data Sheets, spill prevention and slug control plans, secondary containment, monitoring and inspection access, discharge limits, annual certifications, operator training and emergency notification. Read operationally, that is the profile of a managed industrial facility, with the staffing and documentation costs that implies.

Do data centers actually generate meaningful tax revenue? Fort Worth's published figures say yes. Data centers have generated more than $83 million in gross property tax revenue for the city over the past five years. In 2024, that revenue equaled what the city collects from roughly 8,300 average single-family homes, projected to grow to the equivalent of more than 22,700 homes by 2030 based on known in-progress development — from a land use that consumes comparatively little in municipal services.

What should an enterprise executive take from a municipal infrastructure debate? That the critical path for AI capacity runs through utilities and public process rather than equipment procurement, and should be planned accordingly. For the large majority of organizations, the practical conclusion is to buy capacity rather than build it, evaluate counterparties using the five accountability questions, and direct internal effort toward the readiness to use capacity well — which is a separate problem, and the one they actually control.

AI Infrastructure Requires More Than Compute

Metro Relay helps North Texas organizations assess AI infrastructure requirements, operational dependencies, resilience, governance and evidence before major capacity commitments are made.

→ AI Infrastructure Readiness Review

Metro Relay is a technology advisory and implementation firm. We do not control utility planning, grid interconnection or municipal approvals, and we do not represent applicants before public bodies. What we do is help organizations understand what their AI plans actually require, what depends on parties outside their control, and what evidence they will need to demonstrate.