What is AI utility capacity due diligence? AI utility capacity due diligence is the use of models such as ChatGPT, Claude, and Gemini to read will-serve letters and municipal fee schedules, then reconcile the service a property actually has against what your intended use requires, before the diligence period expires. It is the part of the file that kills deals late, because power, water, and sewer constraints rarely appear in an offering memorandum. For the broader process, see our guide on AI commercial real estate due diligence.
Key Takeaways
- A will-serve letter confirms a utility will serve a specific parcel at a stated level. It is conditional and it expires, and AI surfaces both facts fast.
- Electrical capacity is now the single most critical constraint in industrial site selection, with developers in several markets reporting multiyear waits for utility power.
- Convert every capacity figure to one unit before comparing. AI turns amps at a service voltage into kVA so the seller's number and your load calculation match.
- Tap fees connect you to the main. Impact fees and system development charges fund system expansion. Both are capital and both belong in the acquisition budget.
- An unbudgeted $850,000 service upgrade on a $20 million purchase moves a 6.50% going-in cap rate to roughly 6.24%, about 26 basis points of erosion.
What Utility Capacity Diligence Actually Covers
Utility capacity diligence answers one question for each service: can this parcel receive the electric, water, sewer, and gas volume my intended use needs, at what cost, and on what timeline? It is distinct from utility cost analysis, which looks at operating statement rates. Capacity diligence looks at physical supply.
Each utility measures capacity differently, which is exactly why the review goes wrong. Electric service is expressed in amps at a service voltage, or in kVA at the transformer. Water and sewer come in gallons per day, fixture units, or equivalent dwelling units. Gas arrives as BTU per hour at a delivery pressure. A seller may hand you a document for each, and none will be in the units your architect used.
The stakes have moved sharply. NAIOP reported in its Spring 2026 analysis that electrical capacity has become the single most critical constraint in industrial site selection, with developers in multiple markets reporting multiyear delays before a utility can deliver power. In some markets, utilities will not begin a serious conversation without a committed tenant, reversing the traditional sequence of choosing a site and then arranging power.
How AI Reads a Will-Serve Letter
A will-serve letter is a written statement from a utility that it has capacity and will provide a defined level of service to a specific parcel. AI reads it well because the document is short, dense, and highly structured, and because the failure modes are consistent: buried conditions, quiet expirations, and capacity stated for a use that is not yours.
Load the letter into Claude or ChatGPT and ask it to extract seven fields into a table you control: issuing utility, parcel or service address, the exact capacity committed and its unit, every condition precedent, the expiration or reservation date, whether the commitment runs with the land or with the applicant, and every fee obligation referenced. Then ask a second, adversarial question: list every sentence in this letter that limits, conditions, or qualifies the commitment.
That second prompt is where the value is. Will-serve letters routinely commit capacity contingent on paying connection fees by a date, constructing offsite improvements, or maintaining a specific use. Many reserve capacity for 6 to 24 months and release it silently on expiration. If the letter was issued to the seller for a different use, the commitment may not transfer at all. Treat any AI extraction as a reading aid, then confirm material terms with the utility in writing.
Sizing the Power Gap Before You Sign
The most common electrical error in acquisition diligence is comparing two numbers that are not comparable. Existing service is usually described in amps. Tenant load calculations usually arrive in kVA or kW. AI closes that gap in seconds, and the arithmetic is worth understanding because it drives real money.
For three-phase service, kVA equals amps multiplied by volts multiplied by 1.732, divided by 1,000. Consider a 120,000 square foot industrial building with existing 800 amp, 480 volt three-phase service, under contract to a buyer whose light manufacturing tenant needs 2,000 amps at the same voltage:
- Existing service: 800 amps at 480 volts three-phase equals roughly 665 kVA.
- Required service: 2,000 amps at 480 volts three-phase equals roughly 1,663 kVA.
- The gap: about 998 kVA, which means a new transformer, likely a new primary feed, and possibly utility side work.
Now price it. If that upgrade is sized at $850,000 and the deal was underwritten without it, the effect on a $20 million purchase producing $1.3 million of NOI is direct. Going-in cap rate is NOI divided by purchase price, so $1.3 million divided by $20 million equals 6.50%. Add the upgrade to basis and $1.3 million divided by $20.85 million equals about 6.24%, roughly 26 basis points. Ask AI to run that comparison across three upgrade scenarios so you negotiate a credit against a range rather than a single guess.
Modeling Tap Fees, Impact Fees, and Connection Costs
Tap fees and impact fees are different charges, and conflating them is a frequent AI error. A tap fee, sometimes called a connection fee, pays for the physical connection to the existing main. An impact fee or system development charge pays for the capacity your project consumes in the wider system. A project can owe both. Neither is an operating expense; both are capital, and both belong in the acquisition budget alongside the Phase I environmental review.
Most municipalities assess water and sewer capacity charges per equivalent dwelling unit, a normalized measure where one EDU represents the load of a typical single family home. A 200 unit multifamily development in a jurisdiction charging $4,200 per EDU for sewer capacity carries $840,000 in sewer system development charges alone, before water, before the tap, and before any offsite work.
Published fee schedules are exactly the kind of document AI handles well. Feed the current municipal schedule to Claude or Gemini with your unit count, and ask for a line item estimate showing the rate, multiplier, and subtotal for each charge, with a page reference for every rate. Requiring the citation is what makes the output auditable. Then verify the largest line items yourself, because rates change mid-year and models will happily quote a superseded table.
Building the Utility Request Package with AI
The bottleneck in utility diligence is rarely analysis. It is correspondence. Each provider wants a different package, response times run weeks, and a 45 day diligence period disappears while you wait. AI compresses the drafting so the clock starts on day one:
- Day 1: Draft a capacity availability request to each of the four providers from your parcel data and intended use, in each utility's preferred format.
- Day 1: Generate a seller records request covering existing will-serve letters, prior capacity reservations, utility bills showing peak demand, and any moratorium correspondence.
- Days 2 to 5: Reconcile the seller's stated capacity against your load calculation and flag every unit mismatch.
- Ongoing: Extract conditional dates from returned letters into a diligence calendar, so a reservation expiring before closing becomes a negotiating item rather than a surprise.
Our AI due diligence checklist for CRE acquisitions covers where these items sit in the wider timeline, and AI environmental due diligence covers the Phase I workstream running in parallel. For powered land or data center product, start with AI data center due diligence and power capacity, where megawatt scale changes the analysis. The AI Consulting Network builds this workflow against operators' own acquisition checklists.
Where AI Stops and a Licensed Engineer Starts
AI reads documents, normalizes units, and drafts correspondence. It does not produce a reliable load calculation, size a transformer, or tell you what a utility will commit to next quarter. Those require a licensed engineer and a direct conversation with the provider.
Three limits deserve emphasis. Models will confidently quote fee schedules and interconnection timelines that are out of date, so every rate needs a primary source check. Capacity is also dynamic: a utility with headroom in March may have contracted it away to a larger user by August, which is why a stale will-serve letter is nearly worthless. And AI cannot read local politics, even though moratoria on sewer connections and large load interconnections are political decisions before they are engineering ones. Use AI for extraction, normalization, drafting, and tracking, and professionals for calculations and commitments. CRE investors who want help drawing that line can reach out to Avi Hacker, J.D. at The AI Consulting Network.
Frequently Asked Questions
Q: What is a will-serve letter in commercial real estate?
A: A will-serve letter is a written statement from a utility confirming it has capacity and will provide a defined level of service to a specific parcel. It is typically conditional on fee payment and construction milestones, and it carries an expiration date. It is a commitment with strings, not a guarantee.
Q: What is the difference between a tap fee and an impact fee?
A: A tap fee pays for the physical connection between your property and the existing utility main. An impact fee or system development charge pays for the share of system capacity your project consumes. A project can owe both, and both are capital costs rather than operating expenses.
Q: Can AI replace a civil or electrical engineer on utility diligence?
A: No. AI is reliable for extracting terms from letters, converting units, drafting requests, and tracking expirations. Load calculations, transformer sizing, and cost estimates require a licensed engineer, and capacity commitments require written confirmation from the utility.
Q: Does power availability really affect non industrial property?
A: Increasingly yes. CBRE's European Logistics Occupier Survey 2026, which polled 109 large warehouse occupiers, found power availability now ranks as a top building selection factor for over 44% of occupiers, more than double the share four years earlier. Any use adding EV charging, heavy HVAC, or automation meets the same constraint.