Your electrical position gets examined. The water contract and the development agreement get signed.

Current as of September 14, 2026. Executed agreements cited here were read in full from public records. Revision log at the foot of this article.

Part of a five-article series on phased Texas data center campus development and the commitments sized to a buildout that may not arrive.

• Part one. The exposure: why a compliant tenant is the expensive one
• Part two. Where the exposure bites: security, tariffs, hedges and covenants
• Part three. Sizing the shared works against an uncertain phase 2
• Part four. Public commitments sized to full buildout
• Part five. Recovery once the phase has already stalled

The short version, for the capital committee.

---

The scrutiny is inverted. A campus team will model the substation position to four decimal places, negotiate the interconnection agreement line by line, and take the water service agreement, the Chapter 380 agreement and the district assessment more or less as presented. Each of those was sized against the same buildout number as the substation. Each carries its own consequence if that number does not arrive. And because they are documented as fees, covenants and assessments rather than as assets, none of them surfaces when somebody reviews the capital plan for oversizing.

This article walks the public and municipal side of a phased campus, using executed Texas agreements rather than model forms, and ends with the one instrument that is genuinely worse than anything on the electrical side.

A companion piece, You underwrote the tenant's credit and the tenant's power. You did not underwrite the tenant's lender, covers the tenant-credit half of the same problem: what happens when a compliant tenant declines phase 2 and leaves the shared works with no funded beneficiary.

Impact fees, where the asymmetry is statutory

Texas Local Government Code Section 395.025 lists when a political subdivision must refund an impact fee, and every trigger runs to the subdivision's failure to deliver: existing facilities available and service denied, fees collected when service was unavailable followed by a failure to commence construction within two years, or service not arriving within a reasonable period. There is a catch-all for fees not spent within ten years.

Read the list again and notice what is absent. Nothing entitles a developer to a refund because its own later phases never arrived. Houston states plainly that water and wastewater impact fees are not refundable for any reason, including failure to obtain a building permit or complete the project. Attorney General opinion GA-0797 reached the same conclusion, finding no refund requirement outside Section 395.025 and specifically that a project going dormant is not, standing alone, a trigger.

Two qualifications keep that in proportion. Collection under Section 395.016 is tied to plat recordation, connection to the system, or issuance of a building permit or certificate of occupancy, so a phase that is never built frequently never triggers the fee at all. The sunk exposure attaches to capacity you actually connected, not to capacity you once drew on a plan. And the largest recent Texas data center agreements negotiated these fees down or away entirely, as Temple did.

One trap is worth knowing before you structure the land. Section 395.025(e) directs refunds to the record owner of the property at the time the refund is paid, not to whoever paid the fee. If a development entity pays and the land later moves to a partner, a REIT or a tenant taking fee title, the refund right moves with the land. Cheap to handle at the outset, impossible to fix later.

Reservation fees sit alongside impact fees rather than replacing them, and only one of the two usually appears in a capital plan. Section 395.019(3) contemplates an owner voluntarily asking a provider to reserve capacity for future development under a written agreement. In practice a developer pays Chapter 395 impact fees for the units it connects, pays a separate annual fee on the living unit equivalents it is holding but not using, and the agreement provides that unpaid reservation fees are an event of default after which those units leave reserved status. The impact fee buys capacity at connection. The reservation fee rents the option in the meantime, is expressly non-refundable, and can be forfeited on a fixed clock.

Water, where the evidence contradicts the assumption

The reflexive assumption is that reserved water capacity carries a standing charge. Some Texas providers do charge for held capacity. LCRA prices firm water reserved for future use at $82.50 per acre-foot per year against $165 for water actually used, exactly half rate, indefinitely. The West Travis County Public Utility Agency levies an annual, expressly non-refundable reservation fee on unused living unit equivalents and can terminate a non-standard service agreement if those fees go unpaid.

But the executed data center water agreements in Texas do not look like that, and the pattern is the opposite of a take-or-pay.

El Paso Water's agreement with the Meta affiliate states that the customer "shall not be obligated to use or take any portion of the Capacity Allocations, or to commence or complete construction of any portion of the Project," and separately bars reservation fees, capacity fees and impact fees by name. It goes further than any of the others: the customer may terminate at any time and for any reason on notice, may assign without the utility's prior consent to a future owner of the property, and the utility cannot reduce the capacity allocation without the customer's consent, which may be withheld in the customer's sole discretion.

Temple's executed utility service agreement for the Rowan project contains no take-or-pay, no reservation fee and no minimum monthly charge, and records that the company and its tenants "will not be responsible for paying impact fees for water or wastewater service." Amarillo's agreement with Fermi prices water at twice the in-city rate and caps delivery, but obliges the developer to pay only for water actually taken.

Three agreements is a small sample and none of them is yours. The useful conclusion is not that water reservation risk is absent. It is that this exposure is entirely a function of what your specific provider's tariff and your specific agreement say, and that the anchor tenants with the most leverage have been negotiating it away. If you are not that tenant, establish whether your provider is an agency that charges for unused capacity or a municipal utility that does not, because the difference is a real annual number and it will not be obvious from the term sheet.

Worth noting for anyone modelling a wind-down: none of the three agreements contains a release-back mechanism, so capacity returns only through customer-initiated termination or a consented reduction. Where those rights exist, as at El Paso, water is the easiest stranded commitment on a campus to shed. That is the mirror image of the electrical position, which cannot be reassigned at all.

Incentive agreements, where the drafting varies most

Texas municipal practice treats a clawback as what keeps a Chapter 380 grant on the right side of the constitutional prohibition on gifts of public funds, and Chapter 312 requires recapture where an owner fails to make the improvements. Those provisions bite hard when they are drafted hard. Dallas's agreement with a Digital Realty entity makes recapture on a jobs shortfall equal to every dollar of abated tax from the start of the abatement period plus any grant already paid.

Do not assume yours reads that way, because across the executed Texas data center agreements the range is enormous. It runs from that full-benefit recapture, to recapture limited only to the years actually in default, to an explicit cap at incentives already paid, to nothing at all.

Two examples mark the far end. Equinix's Chapter 381 agreement with Dallas County provides that an event of default "shall not entitle County to reclaim or receive a repayment right with respect to any Tax Abatement for any year before the Event of Default occurs," and that for a failure to meet a condition the county's sole remedy is termination, "not damages or specific performance or any other remedy." Missing the jobs requirement there produces no repayment at all. A Hutto agreement recites a ten billion dollar investment as an intention, then states that nothing obliges the owner to construct any buildings and that failing to build them is not a breach. The buildings are conditions precedent to eligibility, not covenants.

Phase scoping is the pattern that matters most for a phased campus, and it is close to universal. A Compass agreement in Ellis County applies its terms to each phase independently, so a default on phase 3 has no effect on phase 1. Temple's agreement for the Rowan campus is blunter still: although the company may develop the land in multiple phases over a period of years, the agreement and all of the parties' obligations "shall only apply to the Phase 1 Development." The same agreement provides that nothing in it requires the company to achieve its employment commitment, which is instead a condition precedent to receiving the infrastructure reimbursement.

Read one way, that is protective drafting, and it means the headline campus-wide investment number in the press release is almost never the enforceable commitment. Read the other way, it means the public incentive that made your pro forma work is scoped to phase 1 and shrinks alongside the buildout that is not arriving. Either way, the number to plan against is the one in the operative covenant, not the one in the recital or the announcement.

One drafting detail cuts against the harshest clauses, and it is worth raising with counsel rather than relying on. Most of these agreements label recapture as liquidated damages, reciting that actual damages would be speculative and difficult to determine. Texas applies a two-part test and will take a retrospective look at whether the stipulated sum greatly exceeds the actual harm, striking it as an unenforceable penalty where it does. A clause converting a modest jobs shortfall into repayment of a decade of abatement is exactly the disparity that invites that challenge. That is a reason to negotiate a cap or a proportional formula at signing, not a reason to assume a court will rescue you later.

Where a public commitment becomes your default

Most of what is above is sunk cost or absent cost. This part is different, and it is the reason the public side deserves the same attention as the substation.

Where a clawback is drafted with teeth, it triggers on your own failure to reach a committed number. If a tenant quietly declines to expand, and your eligibility thresholds were sized to the full campus rather than to phase 1, that tenant's decision converts into your breach of a public agreement. The tenant remains fully compliant. You are the one in default, with a public body, on a number you could not control.

That also runs back into the cost of holding the asset. An abatement conditioned on full-buildout investment or jobs is most likely to lapse in precisely the scenario where you are already carrying unused capacity. So the property tax line on an oversized position does not simply persist, it can arrive retroactively as recapture at the moment your coverage ratio is worst. Two independent-looking risks are the same risk arriving together.

Check two things specifically: whether your agreement's eligibility thresholds are scoped to phase 1 or the full campus, and whether recapture reaches prior years or only the years in default. Temple and Equinix sit at the protective end of both questions. Dallas does not.

How public and electrical exposure actually differ

It is worth being precise here, because the intuitive answer is wrong in both directions.

Municipal exposure is mostly sunk and land-based: fees paid on capacity connected, assessments levied on land at full-buildout sizing, obligations that do not come back but frequently do not grow if you simply stop. Electrical exposure is forward-looking and capacity-based: security posted against contracted peak, CIAC that is not recoverable, a non-utilization clause that recalculates against realized load. In absolute dollars per project, and in how explicitly it is conditioned on continuing to perform, the electrical side is frequently the larger of the two.

What the public commitments add is three things the electrical position does not have. The counterparty is often public, so the remedy is statutory and there may be no commercial conversation to have. The obligation is documented as a fee or a covenant rather than as an asset, so it never surfaces in an oversizing review. And where a clawback has teeth, it triggers on your own shortfall rather than on a counterparty's.

The one instrument that is genuinely worse

A public improvement district assessment deserves singling out, because it is the strongest failure-to-expand obligation on a campus and it sits on the developer rather than the tenant.

The assessment is levied against land sized to full buildout. It is a first and prior lien. It is a personal liability of the owner. It runs with the land, and it survives tax foreclosure as to installments not yet due. Offering documents say the quiet part directly: if the development cannot be completed as planned, the developer or a subsequent owner remains responsible for the full assessment, notwithstanding that it was sized assuming full development.

Districts cut both ways, and the distinction is worth getting right. Texas MUD bond issuance is gated by TCEQ on realized development, including completion of the financed facilities and a threshold share of the projected value of improvements actually built, with the developer reimbursed phase by phase. The developer's exposure there is front-loaded carrying cost rather than stranded public debt. But a district that stalls does not shrink its debt service. The tax rate rises on whoever is on the roll, and a MUD's lien has priority over mortgages.

---

The next action

Pull every third-party commitment on the campus that is not the interconnection agreement. The water and wastewater supply contract and any reserved capacity in it. The Chapter 380 or 381 development agreement and any abatement. The district assessment and its service and assessment plan. The carrier IRU. The land takedown schedule.

For each, record four fields:

What buildout number it was sized against. Usually the announcement number rather than the funded one.

Whether its obligations are scoped to phase 1 or to the full campus. This single field determines whether a tenant's decision not to expand becomes your default.

The release or termination mechanism, if one exists. El Paso-style at-will termination is rare and valuable. Most agreements have nothing.

The clawback, and whether it reaches prior years. The range across executed Texas agreements runs from everything to nothing, so the answer is in your document and nowhere else.

The page that comes back will not be balanced. Most campuses find the electrical position was examined carefully and the water contract, the development agreement and the takedown schedule were signed against the same optimistic buildout with nobody asking the phase 2 question at all.

---

Revision log

September 14, 2026, revision 1. Initial publication. This article separates the municipal and public-commitment material from the tenant-credit article, where it first appeared, so each argument can be followed on its own. The evidence is unchanged; the framing is standalone.

---