Data Center Utility Agreement Review with AI: The Workflow Before Signing

Utility agreements now carry financing, schedule and exit risk. Developers need a structured review before commitments harden.

Data center utility agreement review has become a core development workflow. It is no longer a legal back-office task that happens after site selection. For large-load projects, the utility agreement can decide whether the deal is financeable.

The reason is the rise of large-load tariffs and special service rules. The Smart Electric Power Alliance reported that proposed and approved large-load tariffs and rules grew from 41 in July 2025 to 104 in July 2026 across more than 70 utilities. Edison Electric Institute's September 2026 large-load tariff tracker includes examples such as Montana-Dakota Utilities' high-density contracted demand response tariff, which applies to customers with at least 10 MW of high-density computer processing demand and a minimum 85% load factor.

Those terms are not boilerplate. They change the project model.

A developer may be asked to make minimum-payment commitments, post collateral, fund upgrades, accept exit fees or commit to operating flexibility. Those obligations can survive a tenant delay and collide with lease terms, debt covenants and construction phasing.

AI can review the documents faster. It cannot decide whether the risk is acceptable. That remains human judgment.

What the agreement review has to answer

A utility agreement review should produce an underwriting-grade answer to seven questions:

1. What load is being committed?

2. What does the developer pay if the project is delayed?

3. What does the developer pay if the project is canceled?

4. What collateral is required and when?

5. Who pays for transmission, substation or distribution upgrades?

6. What energization dates are binding?

7. Can the tenant lease absorb the same obligations?

If the team cannot answer those, it is not ready to sign.

Step 1: Build the document set

Start by collecting the full utility and regulatory file, not just the draft service agreement.

The review set should include:

• electric service agreement

• interconnection or load study documents

• line extension agreements

• contribution-in-aid-of-construction terms

• tariff sheets

• large-load riders

• demand response terms

• collateral requirements

• upgrade cost estimates

• utility commission filings

• correspondence that changes timing or obligations

• tenant lease provisions related to power, delay and pass-through costs

AI is strongest at intake. It can classify documents, identify missing exhibits and create a first-pass index in minutes. A human should confirm completeness because missing schedules often contain the obligations that matter most.

Step 2: Extract the commercial obligations

The next task is extraction. The model should pull every quantified obligation into a structured table:

• committed MW

• ramp schedule

• minimum demand charge

• minimum load factor

• deposit amount

• collateral form

• security posting date

• upgrade payment timing

• exit fee formula

• curtailment rights

• delay damages

• termination rights

• assignment restrictions

This is where AI saves real time. Utility agreements are repetitive, but the risk is buried in definitions, exhibits and tariff cross-references. A model can find those clauses quickly and link each extracted value back to source text.

The output should be a model-ready obligation table.

Step 3: Reconcile the agreement against the development model

The extracted obligations then need to be compared with the pro forma, construction schedule and tenant lease.

This is the control step. It asks whether the commercial model reflects the legal commitments.

Common mismatches include:

• utility deposits omitted from sources and uses

• minimum demand charges starting before rent commencement

• collateral requirements sitting at the wrong entity

• transmission upgrades treated as reimbursable when they are developer-funded

• exit fees excluded from downside cases

• energization milestones later than tenant delivery milestones

• demand response obligations inconsistent with tenant uptime requirements

AI can flag mismatches. The development, finance and legal team has to decide how to fix them.

Step 4: Score the risk before signature

A useful review ends with a risk score by category, not a pile of comments.

The scorecard should cover:

Schedule risk

Does the utility commitment align with the construction path? Are energization dates enforceable or best efforts? Are upgrade dependencies outside the utility's control?

Cost risk

Are all deposits, upgrade payments, minimum charges and collateral postings included in the model? Are they funded by equity, debt, tenant reimbursement or sponsor support?

Exit risk

If the tenant walks, how much exposure remains? If the project is delayed, when do costs start? If the load is reduced, does the tariff punish the project?

Financing risk

Will the lender treat the obligations as project debt, contingent liabilities or operating exposure? Are there covenants that restrict additional commitments?

Operating risk

Do demand response or curtailment terms conflict with service-level obligations? Can the facility actually operate under the flexibility promised?

Step 5: Assign human owners

The final workflow step is ownership. Every material issue needs a named decision-maker.

• Utility counsel owns enforceability.

• Development owns schedule impact.

• Finance owns model treatment.

• Leasing owns tenant pass-through language.

• Engineering owns technical feasibility.

• Executive leadership owns risk appetite.

AI can maintain the issue log and draft redlines. It should not be the approver.

What to automate and what not to automate

Automate extraction, cross-references, obligation tables, issue logs and comparisons against the model. Do not automate final judgment on legal enforceability, credit exposure, regulatory strategy or tenant negotiation.

That distinction matters. A model can flag that a tariff requires 85% of expected energy to be paid for even if usage falls short. It cannot decide whether the sponsor should accept that exposure.

Why this belongs before site control hardens

Review the utility agreement before land, tenant delivery milestones or financing assumptions harden. Once site control, lease commitments and procurement deposits are moving, utility terms become harder to renegotiate.

The better workflow is early, structured and repeatable:

1. collect the full utility file

2. extract every quantified obligation

3. reconcile against model, schedule and lease

4. score schedule, cost, exit, financing and operating risk

5. assign human owners

6. negotiate before signature

Large-load utility agreements are now development, financing and risk allocation documents. Treating them as standard utility paperwork is how good sites become bad deals.