When the Right Answer
Is the Utility

Energy planning often treats utility power as the baseline to escape. That is a mistake. The grid is one candidate path, with real strengths, real exposures, and a service process that must be evidenced. When it can meet the load on a usable schedule and its remaining risks are acceptable, building more equipment can destroy value rather than create it.

Section 01The grid has to earn the recommendation

A good grid case is not “keep doing what we do now.” It is a positive finding. The owner has a defined operating need. The expected load has a defensible time profile. The serving utility has provided project-specific evidence about the service path. The commercial terms have been read in the applicable rate and service documents. The remaining outage exposure fits the operation, or a smaller backup measure addresses it. Under those conditions, utility service can be the simplest complete answer.

This finding matters because the no-build option carries benefits that are easy to omit from a technology model. The owner avoids a new fuel supply, prime mover, emissions path, equipment yard, maintenance organization, replacement reserve, operating interface, and construction program. It preserves space and management attention. It can also preserve the option to add a narrower asset later if the load, tariff, or service evidence changes.

The grid still has to survive the same test as every other path. A rate sheet is not evidence of available capacity. A service application is not an energization date. An annual reliability average is not a promise to one facility. A utility estimate is not a final cost. A complete study separates each fact, labels its date and authority, and records what would make the conclusion change.

PG&E's current project resources say every project can have its own timeline, cost, and resource needs. Its application guidance says connection time varies with project complexity and the number of projects in progress, and directs applicants to the assigned job owner for project-specific information.¹² That is the proper starting posture: the grid path is investigated, not assumed.

The grid is not the absence of a project. It is a service path that must be documented well enough to choose.

Section 02Start with the decision, then fix the load

The owner first needs to say what must be powered and why. A growth project, equipment replacement, lease decision, resilience requirement, operating-cost concern, and emissions objective can point to different answers even at the same meter. Without that decision statement, a comparison rewards whichever technology has the most polished proposal.

Next, establish the load that the service must carry. Separate measured interval demand from connected equipment, a future operating forecast, and optional loads. Identify the hours that set the requirement. Test whether those hours occur together and whether the operation can reduce, shift, sequence, or temporarily interrupt any of them. The result may be lower than the first planning request, or it may reveal that the future service need is larger than recent bills suggest.

Demand-side measures are not a consolation prize. DOE describes grid-interactive efficient buildings as combining efficiency, controls, communication, and flexible loads so energy use can be reduced, shifted, or modulated while preserving the needs of occupants.⁵ At an industrial or commercial site, the same logic can apply to process staging, charging schedules, thermal storage, pumping, HVAC, and other controllable loads. The case for these measures is that they may reduce the required service and every downstream asset. The case against is operational: some loads cannot move, controls can add complexity, and a theoretical reduction has no value unless it is repeatable at the hours that matter.

A grid recommendation should therefore be based on the post-measure load, not the unchallenged request. It should also retain the original demand forecast as a scenario. If the project only works when a flexible process behaves perfectly every day, that condition belongs in the decision record rather than disappearing into an average.

Section 03Ask the utility for the right evidence

The useful utility answer connects a defined load, location, service arrangement, scope, cost basis, and schedule. In California investor-owned utility territory, CPUC explains that Rule 15 governs extensions of distribution lines and Rule 16 governs service lines connecting distribution facilities to customer meters. The applicable responsibilities and costs depend on the actual extension and tariff facts.³ The owner should not compress those facts into a single phrase such as “the utility will handle it.”

For a new or upgraded service, the evidence packet should identify the submitted load and voltage, utility point of service, facilities expected on each side of the boundary, applicant work, utility work, dependencies, allowances or contributions where applicable, study or design status, and the document's validity date. Record whether the schedule is a planning range, a milestone forecast, or a contractual commitment. Those terms are not interchangeable.

The current PG&E project page also identifies its 2026 Greenbook as the utility requirements resource for establishing electric or gas service to new or remodeled premises.¹ A design manual can define requirements, but it does not settle a particular feeder, transformer, easement, inspection, material, or construction issue. Project-specific correspondence and the controlling tariff must close those gaps.

Good grid evidence can still contain unknowns. The honest conclusion may be “utility service is preferred if the utility confirms this scope by this date.” That is a conditional decision, not a false certainty. It tells the owner what to resolve next and avoids buying an on-site system merely because the service answer is incomplete.

Section 04Match reliability evidence to the operation

Reliability should be measured at the level relevant to the decision. CPUC's annual reports define system average outage duration, frequency, and restoration-duration measures and distinguish recorded distribution reliability from resource adequacy at the broader system level.⁴ These metrics are useful context. They do not describe the exact outage history of one circuit, the behavior of one facility's internal distribution, or the cost of one interrupted process.

A grid-favorable site usually has one of three characteristics. Its historical service is acceptable for the operation. Its outage consequence is low enough to manage through operating procedures. Or its critical load can be protected with a bounded measure that is smaller than a full independent power plant. The owner should quantify the critical function, ride-through time, restart sequence, and consequence of failure before selecting equipment.

The opposite case is equally important. If a short interruption creates a large safety, product, data, or restart consequence, annual system averages may hide the decision-driving exposure. If the site cannot tolerate the utility's expected service characteristics, the grid may remain the normal supply while storage, standby generation, or a microgrid serves the resilience function. That is a hybrid finding, not proof that all energy should move behind the meter.

No technology eliminates every outage mode. On-site equipment has planned maintenance, forced outages, fuel constraints, auxiliaries, controls, and internal distribution dependencies. Utility service has external network and restoration dependencies. Reliability analysis should compare complete architectures against the same critical-load requirement rather than comparing the grid's observed interruptions with a generator's nameplate.

Section 05Compare the full burden, not just energy cost

A grid case can be economically strong even when an on-site model shows a lower modeled energy cost. The owner must compare the complete decisions. For utility service, include applicable rates, demand structure, service work, customer-side electrical work, timing exposure, and any bounded backup measure. For an on-site path, include development, equipment, interconnection, site work, fuel or charging energy, maintenance, auxiliaries, controls, compliance, insurance requirements, replacement, outage coverage, and management attention.

The comparison period and discount basis must be common. Sunk costs stay separate from future costs. A proposal allowance is not a firm quote. Tax or incentive treatment should be included only when current eligibility and owner-specific use are established; this paper makes no incentive assumption. Residual value and replacement should be modeled explicitly or set to zero with a written reason.

Optionality has value even when it does not fit neatly in a single present-value result. Choosing the grid can preserve land, avoid locking the facility into a fuel and maintenance structure, and allow a later technology choice based on better operating evidence. The case against waiting is that a future service constraint or cost change can narrow choices. The study should state the decision window and the trigger for re-opening the analysis.

An illustrative decision record might compare three scenarios: existing or upgraded utility service, utility service plus a bounded resilience asset, and a larger on-site supply architecture. These are illustrative categories, not a recommendation for any site. Each scenario should use the same load, operating horizon, outage consequence, escalation convention, and confidence labels.

Section 06Give every credible path its honest case

DOE's current on-site energy resource lists storage, combined heat and power, fuel cells, solar, thermal technologies, and other options, and notes that CHP prime movers can include reciprocating engines, gas turbines, microturbines, fuel cells, and steam turbines.⁷ Technology neutrality does not mean treating these paths as identical. It means applying the same operating need and complete-boundary test to each one.

DOE notes that storage shifts energy rather than producing it and that energy is lost in conversion and retrieval; it also describes the timing mismatch that can exist between solar production and electricity needs.⁶ CPUC states that Rule 21 covers interconnection, operating, and metering requirements for generation and storage connected to investor-owned utility systems, including non-export facilities.⁸ These are examples of complete-path facts. A device benefit does not remove its energy source, operating boundary, or interconnection obligations.

For linear generators, a 2024 California Energy Commission report provides public demonstration evidence for one specific project and technology configuration.⁹ A demonstration is evidence, not a universal bankability or performance conclusion. The owner still needs current configuration, service, emissions, fuel, cost, and operating evidence for the proposed system.

Section 07Keep resilience proportional to the critical load

Many sites reject a good grid case because they frame resilience as all or nothing. The better sequence is to define the critical load, required operating mode, acceptable interruption, and restoration objective. A facility may need seconds of ride-through for controls, hours for a protected process, an orderly shutdown, or sustained island operation. Those are different systems.

A small uninterruptible supply, selective storage, standby source, redundant feed, process buffer, or operating procedure may protect the real consequence while the utility continues to serve normal energy. Conversely, a site that must sustain complex loads through extended outages may need a complete microgrid architecture. DOE's microgrid development checklist treats planning, design, procurement, and implementation as a sequence of tasks and reference points, reflecting that a microgrid is a system rather than a single asset.¹⁰

The case for a bounded resilience layer is focus. Capital goes to the loads whose interruption matters. The case against is boundary risk: omitted auxiliaries, controls, fuel, cooling, communications, or restart loads can make a small system incomplete. The critical-load list and one-line diagram must agree before the owner relies on the result.

The grid can therefore be the right energy answer while not being the whole resilience answer. That distinction prevents two common errors: oversizing an on-site plant to cover a narrow outage consequence, and assuming normal utility service automatically meets a critical operating requirement.

Section 08Use conditions, not conviction

A recommendation should state the conditions that support it. For example: the utility path is preferred while the documented service scope can serve the post-measure load, the expected milestone fits the business decision, the lifecycle comparison remains favorable, and the residual outage exposure stays within the owner's stated tolerance. Each condition needs an evidence owner and review date.

Re-open the decision when a condition moves. Common triggers include a material load change, revised utility scope, missed service milestone, new facility control, changed operating criticality, a firm technology proposal, a fuel constraint, an equipment replacement event, or a rate change large enough to alter the comparison. The trigger should be defined before the organization becomes attached to an answer.

Some facts expire faster than others. Interval data may become unrepresentative after an operating change. A utility design or cost estimate can have a stated validity period. A vendor budget can age quickly. Site control and business authority can change with a lease or transaction. The study should not pretend that a good grid case is permanent. It should make the next review predictable.

Conditions also protect against false urgency. If the utility has not answered a material question, the next action can be to obtain that answer rather than immediately procuring another path. If the business deadline cannot wait, the study should identify the bridge requirement separately from the long-term supply choice.

Section 09The grid-case evidence packet

The final record should be short enough for leadership to use and complete enough for a reviewer to challenge. It should not bury the recommendation beneath a catalog of technologies. A disciplined grid-case packet contains the following:

1. The decision and dateState what operating or capital choice must be made, by whom, and when delay changes the outcome.
2. The post-measure load basisShow measured load, forecast changes, coincidence, critical load, and every flexibility assumption with source dates.
3. The utility evidenceAttach or cite the service request, submitted load, point of service, scope, responsibilities, cost basis, milestones, conditions, and validity period.
4. The operating-exposure testCompare site-relevant outage evidence and internal distribution limits with the consequence the owner is willing to carry.
5. The complete path comparisonGive utility, demand-side, solar, storage, firm generation, emerging technology, hybrid, and no-project cases the same load and boundary.
6. The resilience boundaryDefine critical loads, ride-through, islanding or shutdown behavior, restart sequence, and the smallest credible protection measure.
7. The change triggersName the facts that would reopen the decision and assign a date and owner for each review.

A recommendation is strong when another decision-maker can see what is known, what is conditional, what was rejected, and what would reverse the result. If the grid wins only because the alternatives were modeled incompletely, the study has failed. If an on-site path wins only because the utility was treated as an unverified delay, the study has failed in the other direction.

Section 10Choosing less equipment can be the harder decision

Building equipment looks active. Choosing documented utility service can look passive, even when it is the product of better analysis. The owner's job is not to maximize installed assets. It is to secure the operating outcome with the best complete path and keep future choices open where uncertainty remains.

The utility answer is strongest when five findings align: the post-measure load is defensible; service capability and responsibilities are documented; the schedule fits the business decision; reliability exposure is acceptable or bounded by a smaller measure; and lifecycle burden is lower than the credible alternatives. None of those findings is established by a slogan, a rate sheet, or a vendor comparison alone.

Sometimes the result will be utility service plus efficiency. Sometimes it will be utility service plus storage or standby capability. Sometimes the evidence will justify firm on-site generation, a hybrid microgrid, or a temporary bridge. And sometimes the right answer will be to stop. A technology-neutral study earns trust by allowing every one of those conclusions.

The grid should neither win by default nor lose by fashion. It should win when the documents, load, operating consequence, economics, and change triggers describe a better owner decision than building more machinery.