Flexible Service: Trading Firmness
for Speed Under California's
New Tariffs

For as long as most operating executives have been alive, an electric service agreement carried one silent adjective: firm. California's newest connection offers delete that word in exchange for years off the energization date, and whether the trade is a bargain or a liability depends entirely on the load that signs it.

Section 01A faster date on an interruptible wire

The offer now on the table from California's largest utility, and increasingly from regulators nationally, is simple to state. The grid connects your new load sooner than the standard upgrade cycle would allow, and in exchange the load agrees to reduce or reshape its consumption during the hours when the local system is genuinely constrained. The industry calls the product flexible service, curtailable service, or a flexible interconnection. Whatever the label, its essence is a swap: the customer surrenders firmness, and the utility surrenders time.

The swap is possible because power systems are built for their rarest hours. Ranked from highest demand to lowest, the load duration curve of the U.S. organized markets from 2016 through 2024 shows that more than a tenth of the system existed to serve roughly 35 hours per year of extreme peak, and that for 98 percent of all hours more than a tenth of the system sat unused.^4,5 A new load that stands aside during those rare hours occupies capacity that, for the rest of the year, was already there. That is the entire engineering logic behind every program described in this paper.

A 2025 Duke University study put national numbers on that headroom. Across 22 balancing authorities serving about 95 percent of U.S. peak load, roughly 76 gigawatts of new large load could be integrated if each load accepted curtailment equal to one-quarter of one percent of its maximum annual consumption; 98 gigawatts at one-half of one percent; 126 gigawatts at one percent.^4,5 Numbers of that size explain why flexibility has moved from a demand-response afterthought to the centerpiece of interconnection policy in under two years. Federal regulators have joined the push: in June 2026 the Federal Energy Regulatory Commission directed the organized-market grid operators to justify or revise their rules for interconnecting large loads and co-located arrangements.⁶ The direction of travel, in Sacramento and in Washington, is one way: toward connection products that price flexibility explicitly.

Section 02The California instruments, read precisely

Three instruments define the current California menu, and they are not interchangeable.

Flex Connect operates at the distribution level. A participating site connects ahead of the long-term wires upgrades its service would ordinarily wait for, and a software platform, a distributed energy resource management system, coordinates the site's demand against the capacity actually available hour by hour. The utility reports more than half a dozen sites connected under the program, EV charging installations and grid-scale batteries among them, with connections accelerated by 18 to 24 months in many cases and more than 50 further sites committed.¹ Those early participants are worth noticing: they are load types whose entire operating model is discretion about when to draw.

T-Flex is the announced transmission-level counterpart, aimed at the largest new loads, data centers foremost. The published description extends the same logic upward: connect sooner, and adjust consumption during rare system constraints rather than waiting for the upgrade cycle to complete.¹ The public description is brief, which means the contract terms, the actual substance of this paper, are precisely what a prospective applicant should be probing.

Interim Electric Rule 30 buys speed with a different currency. In July 2025 the California Public Utilities Commission approved, on an interim basis, a uniform tariff pathway proposed by the utility for retail electric service at transmission voltages of 50 to 230 kilovolts, the service class of the largest campuses and computing facilities.^2,3 Decision 25-07-039 rewards a careful read, because its structure is unusual. Interim implementation is available only to applicants who pay as they go, through advances or actual-cost payments, and who voluntarily prefund up to 100 percent of specific transmission network upgrades. New transmission-level customers, not ratepayers, carry the initial cost of all transmission facilities.³

What the decision conspicuously does not settle is the money's return trip. Refunds of advances and contributions, interest on those amounts, and repayment of prefunded loans were all denied for the interim period and deferred to the proceeding's final decision. No interest accrues on advanced funds in the meantime, and agreements signed during the interim remain subject to whatever the final decision concludes on the deferred issues.³ An applicant signing today accepts a purchase price now and a rebate schedule to be determined later by a regulatory process it does not control. That is not a criticism of the Commission, which was explicit about the deferral and framed the structure as ratepayer protection.² It is a term of the deal, and it belongs in every board memo that recommends signing.

Read together, the instruments define two currencies for buying time: cash, under the interim transmission-service framework, and firmness, under the flexible-connection programs. Nothing prevents a large project from spending both. Everything in this paper argues for knowing, before signature, exactly how much of each is being spent and what it purchases.

Section 03The arithmetic of the curtailed hour

Curtailment exposure is usually quoted the way the research quotes it, as a share of annual energy. Half of one percent of a year is, by arithmetic, roughly 44 hours at full depth. Phrased that way it sounds like rounding error, and for the right load it is. The operator's translation runs the other direction: those hours do not arrive scattered and single. They cluster, in heat waves, on consecutive late afternoons, precisely when every other flexible customer on the same system is being asked for the same thing. A load that can give back one hour a week is a different animal from a load that must give back four hours on each of ten consecutive August evenings, and an annual percentage cannot tell the two apart.

Depth matters as much as duration. In the Duke analysis, 88 percent of the hours in which curtailment is required still retain at least half of the new load, and 60 percent retain three-quarters or more.⁵ Real constraint events mostly ask loads to shrink, not vanish. A facility that can shed its discretionary third on ninety minutes' notice may sail through a season of events that would cripple a facility that only knows how to be fully on or fully off. Contract language that limits curtailment to a stated capacity floor is a materially different product from language that permits interruption to zero, and the difference rarely shows up in the headline percentage.

The cost of a curtailed hour, meanwhile, has almost nothing to do with the energy not purchased. It is the value of what the site does with power: the contribution margin of lost output, scrap on interrupted batches, restart time and restart energy, thermal stress on equipment that dislikes cycling, service credits on breached commitments, overtime to catch up. For some loads that number is close to zero, because the work simply moves to another hour. For others it is a multiple of the entire annual power bill.

An illustration, with invented round numbers, and nothing more than an illustration: a plant clearing $30,000 of contribution margin per operating hour that expects 40 hours of full curtailment in a difficult year has roughly $1.2 million of annual margin exposed. If a flexible connection delivers service two years before the firm alternative would, those two years contain thousands of operating hours, and the trade is not close. Invert the operating profile: a thin-margin continuous process where any interruption scraps work in progress and imposes a half-day restart can lose more in those same 40 hours than the earlier date was worth. Same tariff, same percentage, opposite answers. The deciding variable is what an interrupted hour does inside the fence, and only the owner can price that.

One structural point completes the arithmetic. In classic demand-response programs, flexibility is a product the customer sells, and the utility pays for each performance. In a flexible connection, the compensation is the date itself, paid once, up front, in time rather than money, while the curtailment obligation continues for as long as the agreement runs. An owner evaluating the offer should amortize the value of the earlier date across every curtailment year of the contract, not treat the speed as a signing bonus and the curtailment as fine print.

The connection date is the headline. The product itself is written in the curtailment clauses.

Section 04Where flexibility is native, and where it is purchased

Load classes differ less in whether they can curtail than in what curtailment destroys. The honest sort is threefold: loads whose value survives being moved in time, loads that can buy that property with equipment, and loads for which the property does not exist at any sensible price.

Two patterns in that table deserve emphasis. First, the loads that fit best are the ones already visible in the early program results, charging installations and storage, loads whose operating model is built on discretion about when to draw.¹ Second, for the loads that fit worst, the fix is always some version of the same move: pair the flexible connection with on-site capability that carries the site through called events. That capability might be electrochemical storage, which operates without combustion or local emissions but holds finite hours and recharges through the same constrained service. It might be dispatchable on-site generation, whether engines, turbines, fuel cells, or linear generators, each carrying its own permit path, fuel exposure, maintenance calendar, and footprint, and each deserving the case for and against that this series makes elsewhere. Or it might be oversized process buffers, which are cheap until floor space becomes the scarce commodity. None of these is free, and none is universal.

The moment flexibility is purchased rather than native, the purchase belongs in the master comparison. The same storage that makes a curtailable connection tolerable could instead firm a standard limited service. The same generation that rides through curtailment events could, sized differently, self-supply the site and change the interconnection question entirely. Paper No. 001 of this series sets out the eight paths that comparison should span; flexible service is one column of that matrix, not a substitute for it.

Section 05The terms are the economics

Five contract dimensions decide what a flexible connection is actually worth, and none of them is the connection date that the summary page emphasizes.

• Trigger. Who may call an event, and against what standard: a defined and measurable system condition, a market signal, or operator judgment. Words like rare and infrequent are expectations. A written cap is a term. The two are not the same instrument.
• Depth. Interruption to zero, curtailment to a stated floor, or a dynamic limit set by the management platform. Who sets the limit, how often it can move, and the level below which it cannot go.
• Duration and frequency. Per-event maximums, annual maximums, consecutive-event treatment, seasonal concentration. If the agreement carries no numbers here, the operating plan cannot carry any either.
• Notice and control. Day-ahead scheduling supports moving production; minutes-level dispatch requires automation. Equally important is who executes the reduction, the customer's systems responding to a signal or the utility's platform enforcing a limit, and what telemetry the program requires. Failure to perform has a price, in penalties or in the service itself; it should be a written one.
• Exit and conversion. Whether a documented path to firm service exists, what triggers it, who pays for it, and what happens to the curtailment obligation if the upstream upgrades slip. A flexible connection with no conversion mechanism is not a bridge. It is a destination.

Where the speed is bought with cash rather than flexibility, the same discipline applies to the money. Under the interim transmission-service decision the questions are concrete: which payments are advances, which are contributions, which are prefunded loans, and the recognition, in writing, that refund and interest treatment for all three awaits a final Commission decision.³ Counsel, not enthusiasm, should mark that page of the agreement.

Section 06Seven questions to ask before signing

The list below compresses this paper into the meeting where it matters. Any organization can put these to a utility account team or a connection offer in an afternoon, and the quality of the answers is itself diligence data.

1. Who can call a curtailment, and against what written standard?Defined system conditions and measurable triggers can be planned around. Discretion cannot. Wherever the word rare appears without a number attached, ask for the number.
2. To what depth can the load be cut, and is there a floor?Interruption to zero and curtailment to a stated floor are different products sold under one name. Establish the floor, and establish who controls the dial between events.
3. How many hours, how many events, how clustered?Per-event and annual maximums, consecutive-day treatment, seasonal windows. The operating plan inherits whatever this clause leaves open.
4. How much notice arrives, and through what machinery?Day-ahead schedules let production move; minutes-level dispatch demands automation and telemetry. Know which is being signed, and whose systems execute the reduction.
5. What does a curtailed hour cost this specific operation?Priced in margin, scrap, restart, and breached commitments, not in avoided energy purchases. Model it at full depth and at half depth, for single events and consecutive ones.
6. What converts this to firm service, and when?A dated, funded conversion mechanism separates a bridge from a permanent condition. If the upstream upgrades slip, the obligation should not silently extend.
7. What money moves now, and on what terms does any of it return?Under the interim transmission-service framework, refunds, interest, and loan repayment were expressly deferred to a future regulatory decision. Signing means accepting that open item; a board should accept it knowingly.

Section 07A different product, not a discount

Flexible service is a genuine addition to the California menu, and for the right load it is the cheapest speed available. The physics is real, the early connections are real, and the regulatory momentum, state and federal, points one direction. Nothing in this paper argues otherwise.

But firm service sells certainty of supply, and flexible service sells earlier access with conditions attached. They are different products that happen to share wires, and the difference lives in clauses most connection summaries never quote. An owner who prices the curtailed hour honestly, reads the conversion terms skeptically, and weighs the offer against the full menu of paths, storage, self-supply, phasing, and simply waiting among them, may sign or may decline, and either way has made a power decision rather than accepted a sales pitch. The discipline, not any particular answer, is the point.