Data Centers in Texas: Peak Math, Success, Failures, and How to Keep Residents Off the Hook

AI Reponses to Questions Posed by Lewis McLain

Introduction

It was exciting to me when I joined the City of Garland in the early 1970s. Working in municipal government was not something I had considered when I received my BBA in Accounting. I never really wanted to be an accountant. My true love was Budgeting and Cost Accounting. The gift I really received was the introduction to Utility Rate Making. Garland not only had Water & Sewer Utilities, but the city also had an Electric Utility. I was also fortunate to work with excellent outside Rate Consultants. The big present wrapped with a nice bow was the concept of Peak Demand vs Average Demand in utility systems. From there, I realized the concept applied to roadways and many other aspects of municipal services. LFM

The Quick Math (so this posting makes sense)

Every discussion about data centers and electricity should begin with two simple metrics: load factor and peak demand.

  • Load factor (LF) = Average demand ÷ Peak demand.
  • Peaking factor (the inverse) = Peak ÷ Average = 1/LF.

Example (same annual energy, different load factors):
Suppose a data center averages 50 MW (megawatts or one million watts) of demand across the year. The perfect situation would be if there were businesses with a 100% load factor, meaning a business used the same amount of power every single hour (actually every minute) of the year.

  • At 50% LF, the peaking factor is 2.0. That means Peak = 100 MW.
  • At 75% LF, the peaking factor is 1.333. That means Peak ≈ 66.7 MW.

Takeaway: By raising the load factor from 50% to 75%, the required peak capacity falls by about 33% while delivering the same yearly energy.

And here’s why that matters: Texas utilities and ERCOT must size substations, feeders, and generation to meet the peak, not the average.

Homes conversion rule of thumb:

  • 1 MW ≈ 250 Texas homes at summer peak (based on ~4 kW per home).
  • 1 MW ≈ 625 homes on an annual-energy basis (average load ~1.6 kW per home).

So a 100 MW campus is the equivalent of a new mid-sized city landing on your grid overnight.


The Perfect Story and Outcome

Now picture the ideal case. A fast-growing tech firm proposes a 100 MW data campus in Texas. Instead of rushing, city leaders and the utility sit down with the company at the start and insist on clear answers. The questions are simple but critical:

  • What will your peak demand be, and how will you manage it during the state’s hottest afternoons?
  • Who pays for the new substation and feeders, and who carries the risk if you scale back or leave?
  • How do we ensure your taxable value stays meaningful even after your servers depreciate?
  • What tangible benefits will our community see, beyond the building itself?

On the grid:
The company commits to a high load factor and pledges to curtail 20–30 MW during ERCOT’s four summer peaks. The new substation and feeders are paid through contribution in aid of construction (CIAC), so residents will never face stranded costs like the costly investment itself.

On the finances:
Abatements are milestone-based—tied to actual MW energized, not just breaking ground. Valuation floors lock in a taxable base for servers and electrical gear, guaranteeing a predictable $5–10 million per year for schools, police, and parks.

On jobs and training:
The campus directly employs about 60 skilled staff for operations. But the developer also funds a community-college training pipeline in IT and electrical trades, seeding hundreds of local careers. The construction phase delivers hundreds of short-term jobs for two years.

On resources:
The data hall commits to water-efficient cooling, capped at a set gallon-per-MW threshold with quarterly reporting. A community benefit fund supplements fire protection and road upgrades near the campus.

On politics:
Hearings are calm because everything is transparent. Residents know in plain English that their bills won’t rise, because the project carries its own risk.

Outcome:
Five years later, the facility hums steadily, the schools are flush with additional tax revenue, and the city is recognized as a model for how to land high-tech investment without burdening households or small businesses.


What Could Go Wrong? (Case Narratives)

Of course, not every story ends this way. Around the country, major data-center projects have stumbled, been cancelled, or backfired in ways that offer hard lessons for Texas communities.

Corporate pullback after big promises — Microsoft

In 2025, Microsoft canceled or walked away from about 2,000 MW of planned data center capacity in the U.S. and Europe. Analysts cited oversupply compared with near-term demand. Utilities and communities that had already been preparing for those loads were left with planning costs and the risk of stranded substations.

Lesson for Texas: Even blue-chip firms are not risk-free. Cities must require CIAC, minimum bills, demand ratchets, and parent guarantees so residents aren’t forced to backfill the shortfall if plans change.


Court voids approvals after years of work — Prince William County, Virginia

In August 2025, a Virginia judge voided the rezonings for the “Digital Gateway” project—37 data centers on 1,700 acres—citing legal defects in notice and hearings. Years of planning collapsed overnight.

Lesson for Texas: Keep zoning and notice airtight. Add regulatory failure clauses in agreements so if courts unwind approvals, the city isn’t on the hook.


Political rejection at the finish line — College Station, Texas

On September 11, 2025, the College Station City Council unanimously rejected a proposed 600 MW data campus after residents raised concerns about grid strain, noise, water use, and meager job counts. The rejection stopped the project before construction—but it revealed how quickly sentiment can flip.

Lesson for Texas: Require peak-hour commitments (4CP curtailment), publish MW timelines, and cap water usage. Transparency eases public concerns and avoids last-minute backlash.


Industry-wide pauses — Meta redesigns for AI

Between 2022 and 2024, Meta paused more than a dozen U.S. projects to redesign for artificial intelligence. Sites like Mesa, Arizona slipped years behind schedule. Communities banking on near-term tax revenue saw gaps in their budgets.

Lesson for Texas: Tie abatements to energized MW milestones. If load slips, abatements pause until actual demand materializes.


Subsidy blow-ups — Texas and beyond

By 2025, Texas’ data center sales-tax exemptions ballooned from $157 million to more than $1 billion per year in foregone revenue. Other states saw similar overruns as projects multiplied faster than expected.

Lesson for Texas: Model depreciation and appeals honestly. Use valuation floors in agreements, and don’t oversell the net gain at ribbon-cuttings.


Local backlash stalls projects — Central Texas

In Central Texas, residents have already forced pauses or redesigns of major projects, citing water stress, noise, and grid strain. CyrusOne and others adjusted timelines under pressure.

Lesson for Texas: Put MW forecasts, curtailment commitments, and water-use data in plain English. Opaqueness breeds opposition.


Who Pays When a Big Customer Leaves?

In Texas, fixed delivery costs don’t vanish if a large customer fails or exits. Unless safeguards are in place, those costs roll into the next rate case and land on residents and small businesses.

Protective tools include:

  • CIAC: Customer funds all dedicated substations/feeders.
  • Facilities charges: Monthly fees for customer-specific assets.
  • Contract demand and minimum bills: Revenue stability even if load shrinks.
  • Demand ratchets: If they ever peak high once, they pay a portion of that demand for future months.
  • Parent guarantees or letters of credit: Real money backing early-exit costs.
  • Peak-hour curtailment covenants: Written commitments to reduce load during ERCOT’s four summer peaks.

These tools are standard in Texas utility practice. The only mistake is failing to insist on them.


Bringing It Home to Collin & Denton (DFW)

The Dallas–Fort Worth market is growing fast: nearly 600 MW operating and another 600 MW under construction, almost all pre-leased. In Collin and Denton counties, just two or three large campuses can rival the load of an entire mid-size city.

That’s why development agreements must:

  • Stage energization in MW blocks,
  • Require 4CP curtailment reporting, and
  • Hard-wire CIAC plus facilities charges so no “stranded substation” ever lands on residents.

Conclusion: Planning With Eyes Wide Open

Data centers are the backbone of cloud computing, e-commerce, and artificial intelligence. For Texas, they promise billions in private investment and hundreds of millions in taxable value. But their true footprint is measured in megawatts, not headcount.

Handled well—with CIAC, ratchets, valuation floors, and peak-hour curtailment—they can be stable anchors of local finance. Handled poorly, they can leave residents paying for stranded substations, foregone tax revenue, and empty server halls.

The “perfect story” shows it can be done right. The failures across the country show what happens when it isn’t. For Texas cities, the path forward is clear: land the investment, but make the project carry the risk—not your ratepayers.


Contract terms cities and utilities should insist on (plug-and-play list)

  • CIAC for all dedicated facilities (feeders, substation bays, transformers).
  • Facilities charge (monthly) on any utility-owned dedicated equipment.
  • Contract demand with a minimum bill and demand ratchet.
  • Parent guarantee / letter of credit sized to cover early exit and decommissioning.
  • Peak-hour curtailment targets (spell out dates/hours and telemetry).
  • Milestone-based incentives (abatement pauses if MW milestones slip).
  • Valuation floors for server personal property and clear depreciation schedules.
  • Quarterly public reporting: MW online, curtailment at peaks, water usage if relevant.

DFW planning checklist (Collin & Denton emphasis)

  1. Get the MW ramp (Year 1–5), contract demand, and minimum bill in writing.
  2. Require CIAC + facilities charges so bespoke assets aren’t rate-based on everyone.
  3. Bake in peak-hour curtailment commitments (the four summer peaks).
  4. Tie local incentives to energized MW, not just building permits.
  5. Set valuation floors and independent appraisal rights.
  6. Secure credit support (parent guarantee or LOC) sized for the dedicated build.
  7. Publish quarterly progress (MW online and peak reductions) to keep trust with residents.

Sources (selected)

  • Corporate pullback: Microsoft cancellations ≈ 2,000 MW (TD Cowen). Reuters+1
  • Court reversal: Prince William “Digital Gateway” rezonings voided (Aug. 2025). Data Center Dynamics+1
  • Political rejection: College Station votes down 600 MW sale (Sept. 2025). Data Center Dynamics+1
  • Industry-wide pause/redesign: Meta paused >12 builds; Mesa AZ delay to 2025. Tech Funding News+1
  • Subsidy growth: Texas data-center tax costs > $1 B/yr; spikes across states. Good Jobs First+1
  • DFW market scale and pre-leasing: CBRE market profiles and releases (H1/H2 2024–2025). CBRE+2CBRE+2
  • Central-Texas pushback (CyrusOne pause noted): Austin American-Statesman review (Sept. 2025). Statesman

Leave a comment