Utility Letters: Getting the
Grid Answer in Writing

In most power projects that go wrong, there is a number that governed everything: a capacity figure, an energization date, an upgrade cost. In a remarkable share of them, that number never existed anywhere except in someone's memory of a phone call with the utility.

Section 01The weakest document in the file is a memory

The scene repeats across every sector we study. A planning meeting reaches the electrical question, and someone reports the grid answer: the utility said about two years; the planner told us the bank is at capacity; the account rep thought we could get part of the load served now. Site selection, lease commitments, equipment orders, and board schedules then arrange themselves around that sentence. Nobody in the room can produce the sentence itself. It lives in notes from a call, or in nobody's notes at all.

The verbal answer is weak for structural reasons, and it is worth being precise about them, because none of them involve bad faith. First, the person answering sees one slice of a shared and moving system. Capacity is consumed continuously by other customers' projects, and the queue ahead of a site changes month to month; an answer that was a fair reading in March can be stale by June with no one having misled anyone. Second, the individual on the phone is almost never authorized to commit the company. Utility obligations run through tariffs, filed forms, and executed agreements, not through conversations, and a careful employee who declines to improvise a commitment is doing the job correctly. Third, people rotate. The planner who remembers your project moves on, and the recollection moves with them. Fourth, your own request drifts: the load, the voltage, and the in-service date evolve between conversations, so the remembered answer often belongs to a question you are no longer asking.

Both failure modes are expensive. Read optimistically, the recollection becomes capital committed against a date nobody promised, discovered at the moment the site needed power. Read pessimistically, it becomes equipment purchased, fuel arranged, and permits pursued against a constraint that a formal request might have dissolved for a filing fee. The second error is the quieter of the two, because pessimism feels prudent even while it is buying an alternative the site may not have needed.

None of this makes the utility the villain of the story. It makes the conversation the wrong instrument. Commitments in this industry are made through documents, the documents form a ladder, and the ladder is climbable by anyone willing to ask correctly and to read what each rung actually says.

Section 02The ladder, from remark to contract

Every grid answer your organization holds sits somewhere on a ladder of increasing specificity, cost, and commitment. Two properties rise together as you climb: how much the document commits the utility, and how much it costs you, in fees, preparation, and elapsed time, to obtain. The management skill is matching the rung to the decision at hand. A screening decision can rest on the lower rungs. A capital decision cannot.

The names vary by utility, by state, and by whether the question is serving new load or interconnecting on-site equipment. The ladder does not vary. The first discipline is to ask, of every page in the project file: which rung is this. The second is to notice which rung the next decision actually requires, and to close the gap before the decision rather than after it.

Section 03The rungs that get misread

The purchased snapshot: cheap, fast, and honest about itself

The cleanest public illustration of a written screening answer comes from the generation-interconnection side of the house, and it is instructive for every other request. California's large utilities sell pre-application data reports under their generator-interconnection tariffs. At Pacific Gas and Electric Company, the current tariff form prices the standard package at $300 and, for that package on its own, commits delivery within ten business days of a complete request; richer packages carry higher posted fees and longer windows.¹ For that fee, an applicant receives the recorded capacity picture for the circuits and substation equipment most likely to serve a named location: total, allocated, queued, and available capacity; voltages; approximate circuit distance to the substation; line-section load estimates; protection equipment counts; and known constraints from power quality to secondary networks.¹

Just as instructive is what the form says about its own limits. The report reflects readily available, pre-existing records at the time of the request. The form cautions that the existence of available capacity does not imply a project of that size can interconnect without system impacts, that the distribution system is dynamic and subject to change, and that the data may be outdated by the time an actual interconnection request is filed.¹ That candor is not a defect. It is the correct labeling of a snapshot, and it models how every low-rung document should be read: the disclaimers are part of the answer. A new-load inquiry moves through a different door, the utility's service-planning process rather than an interconnection tariff, but the same logic governs the early answers it produces: written, dated, inexpensive, and explicitly provisional.

Ability-to-serve letters: the most over-read instrument in the field

Somewhere above the snapshot sits the letter. Different utilities and municipal providers issue them under different names: will-serve letters, ability-to-serve letters, service-availability letters, capacity-confirmation letters. Lenders ask for them. Site-selection committees collect them. Permitting agencies sometimes require them. They are genuinely useful, and they are also the most misquoted documents in the industry, because their titles tend to promise more than their sentences do.

Read one slowly and three features stand out. The load is described, and the statement applies to that description, not to the larger load the project later grew into. The commitment verb is usually intent, not reservation: the utility expects to be able to serve, subject to conditions. And the conditions clause carries the real content: subject to system conditions at the time of application, subject to completion of required upgrades, subject to the applicable tariff and rules. A letter built from those clauses is not a capacity reservation and not a date commitment. It is a statement of present ability and good-faith intent, valuable precisely for what it is, dangerous only when a financial model treats it as a contract.

The line between letter and contract, it should be said, runs through content rather than format. At the same California utility, the extension agreement that follows a completed service application is itself drafted as a letter: it states that the company will extend its facilities to provide the requested service to the project address, and that the letter, together with the tariff rules and standard forms it incorporates by reference, serves as the contract.² A letter can be a binding agreement, and a two-hundred-page study can bind nobody to anything. Rank every document by its operative sentences, never by its thickness or its letterhead.

A recollection is not a record, and a title is not a commitment. The grid answer that governs capital should be the one somebody signed.

Study results and executed agreements: the top of the ladder, still worth reading

Engineering studies, whether a service-method determination for new load or a system-impact study for interconnection, commit the analysis, not the outcome. They are the first documents on the ladder produced specifically for your project by the utility's engineers, and they deserve respect and skepticism in equal measure: respect, because the scope of upgrades and the cost figures are now project-specific and written; skepticism, because every study rests on a queue snapshot and an assumption set with an expiry date. The professional habit is to log four things the day a study arrives: its as-of date, the stated accuracy class of its cost figures, the queue and load assumptions embedded in it, and the events that would trigger a restudy.

Executed agreements sit at the top: line-extension and service-extension agreements under the distribution rules,³ special-facilities arrangements, and, for the largest loads, construction and interconnection agreements at transmission level. These carry mutual obligations: scope, cost responsibility, milestone schedules, security postings, termination rights. They are the rung on which capital can responsibly stand. They still reward close reading, because estimate-revision clauses, milestone obligations that run in both directions, and events the contract allocates to neither party all live in the text.

If one number shows how much this hierarchy matters at market scale, it is this: the utility's own investor reporting distinguishes conversation from contract. Pacific Gas and Electric Company's second-quarter 2026 earnings presentation, filed with the U.S. Securities and Exchange Commission, reports a data-center demand pipeline exceeding 12 gigawatts, and, within it, 490 megawatts holding executed interconnection construction agreements.⁴ The gap between those two figures is not a criticism of anyone. It is the ladder, printed in an investor deck: most demand is still talking, a fraction has signed, and the utility itself accounts for the difference.

Section 04How to ask, and how to be worth answering

Getting good paper is mostly a matter of asking through the right door, for a named instrument, with an answerable request. The informal conversation keeps its place: use it to learn which instruments exist for your situation and what each one requires. Then move the question itself into the formal channel, because applications and portal submissions create records, dates, and queue positions by default, while conversations create none of the three.

Name the instrument. Utilities publish their forms and fee schedules in their tariff books; the pre-application report described above is a posted form with a posted price.¹ Asking for the specific product, a data report, a written ability-to-serve statement for a described load, a service-method determination, an extension agreement, gets a materially better answer than asking whether someone could send something in writing. Where the right instrument is unclear, the correct question is equally specific: which written products exist under this tariff for a request of this description, and what does each one cost, deliver, and commit.

Make the request answerable in one pass. A one-page load letter does more for your timeline than a dozen calls: service address and parcel, requested capacity and voltage, target in-service date, a realistic ramp schedule, load factor and duty cycle as honestly as they are known, site-control status, and a single point of contact. Honesty in both directions is self-interested. An inflated capacity request distorts the utility's planning, can carry deposit and study-cost consequences under the applicable rules, and burns credibility that later phases will need. An understated one wastes the queue position it earned. Ranges, stated as ranges and updated in writing as they narrow, outperform false precision on every timescale.

Then interrogate every answer with the same three questions, in writing: what is this based on; as of when; and what would change it. Add a fourth for anything that will sit in the file for more than a quarter: when does this lapse, and what refreshes it. Finally, convert every meaningful conversation into shared paper with a confirmation note, two sentences sent the same day, recording your understanding and inviting correction. A confirmed note commits nobody. It simply moves the exchange from two private memories into one shared record, which is most of what this paper is asking for.

One more reason the discipline pays right now: the instruments themselves are in motion. In July 2025 the California Public Utilities Commission approved an interim framework, Electric Rule 30, for connecting large transmission-level loads in Pacific Gas and Electric Company's territory, conditioned on applicants funding the transmission work up front, with refund treatment deferred to a later decision.⁵ In June 2026 the Federal Energy Regulatory Commission directed the six regional grid operators to justify or reform their tariffs for integrating large loads.⁶ When the rules move this quickly, last year's quoted practice ages fast, and dated paper appreciates: it fixes what was true, and what was offered, at a moment the rules may since have left behind.

Section 05The seven lines a written answer needs

Whatever the rung, a decision-grade written answer contains seven lines. Use the list as a completeness check on any document the project intends to rely on. Whatever is missing is the next request.

1. The instrument, named, and the issuer, identified.Which tariff product or agreement this is, who issued it, and in what capacity. Paper that cannot be classified ranks as rung one, whatever it looks like.
2. The quantity, located.Capacity at a named point: a service point, circuit, bank, or substation. A number without a location is not yet an engineering answer.
3. The date, labeled.Target, estimate, or commitment, in the document's own words. The label is the content; an unlabeled date defaults to hope.
4. The conditions, enumerated.Every subject-to spelled out. Unstated conditions still exist; you simply lose the ability to manage them.
5. Cost responsibility, with its accuracy class.Who pays for what, whether each figure is an estimate or firm, and how deposits, advances, and refunds are treated under the tariff.
6. The as-of date and the expiry.What records the answer reflects, when it goes stale, and what refreshes it.
7. The next instrument, named.What converts this rung into the one above it: the application, study agreement, or contract that follows. A document that names its successor is a plan; one that does not is a souvenir.

A written answer carrying all seven lines can still change; the grid is shared, and every tariff reserves room for conditions to move. The difference is that it changes in daylight, against a record of what was said, by whom, as of when. That record is what turns a change into a negotiation instead of a surprise.

Section 06The benchmark duty runs in both directions

The reason this paper belongs in a technology-neutral series is that the written grid case is the benchmark against which every alternative is priced. On-site generation in any technology class, storage in any role, flexible or phased service, relocation, and deliberate waiting are all compared, explicitly or not, to what the grid would deliver and when. If the grid case in that comparison is a recollection, every line of the comparison inherits its error.

The error runs both ways, and each direction has a constituency. A pessimistic, undocumented grid case flatters every alternative: it can lead an organization into machinery, fuel exposure, air permits, and operating obligations acquired against a constraint that a filed application might have cleared. An optimistic, undocumented grid case flatters the incumbent path: it can hold an organization in place, unserved, until the deadline arrives with no capacity behind it, and the bridging that follows is bought at premium prices under time pressure. Neither error is a technology's fault. Both are documentation failures.

The same discipline therefore applies, with equal force, to the alternatives. Self-supply has its own ladder: a vendor's delivery quote and a brochure efficiency figure are its verbal rung; a signed equipment contract, an executed interconnection agreement for the on-site machine, a construction permit from the air district, and a fuel-supply arrangement are its executed rung. Storage is likewise sized and priced on documents rather than datasheets, or it is not yet sized and priced. An honest comparison sets a documented grid case against documented alternatives, rung for rung.

In study work we hold that comparison to a parity rule: no path is compared using a documentation rung lower than its competitors'. If the grid case rests on a study result and the generation case rests on a brochure, the generation case gets upgraded or the gap gets stated in writing. The rule sounds procedural. It decides projects, because most bad energy decisions are not bad engineering; they are mismatched documents standing on opposite sides of the same spreadsheet.

Section 07The file is the asset

Run this discipline for a full project cycle and the organization ends up holding something quietly valuable: a dated file in which every load-bearing answer is written, classified by rung, tagged with its basis and expiry, and traceable to its author. That file outlives staff turnover on both sides of the relationship. It shortens every later conversation with a lender, a board, an insurer, or an engineer of record, because the question of what the utility has actually said takes minutes to answer instead of weeks to reconstruct.

The operating rule underneath this entire paper fits in three clauses. If it matters, request it in writing. If it arrived verbally, confirm it in writing the same day. If it cannot be obtained in writing, treat it as unresolved and plan accordingly. Utilities are not offended by this discipline; it is how their own commitments are structured, and the applicant who works in instruments rather than impressions is, from the other side of the counter, the easiest kind to serve.