The Technology-
Neutral Test

The most expensive energy decision most organizations will make this decade is not which technology to buy. It is which path to power to take, and most owners make that decision with advice from parties who are compensated to sell exactly one answer.

Section 01The decision arrives before the technology

When an organization discovers that its site cannot get the electric capacity it needs, on the date it needs it, a decision process begins. In our experience that process almost never starts neutral. It starts with a phone call to whoever is nearest: a generator dealer, a solar installer, a battery integrator, an equipment manufacturer. Each of those parties is competent, and each of them earns money in only one way. The engine dealer's analysis will conclude in engines. The solar proposal will conclude in solar. The battery model will conclude in batteries.

None of this is dishonest. It is structural. A seller's feasibility study is a sales document, and the owner who treats it as a decision document has quietly outsourced the most consequential judgment in the project to the party with the largest stake in its outcome.

The alternative is a discipline, not a slogan: price every credible path to power for the specific site, with sourced and dated inputs, before any technology is selected. This paper describes that discipline. It sets out the eight paths that a large energy user should see priced side by side, and it closes with a seven-question test that any owner, board, or finance committee can apply to any energy recommendation, including ours.

Section 02What changed: the time-to-power era

For most of the last fifty years, the interval between requesting utility service and receiving it was a footnote in project planning. That era is over for large loads in California, and the clearest evidence comes from the utilities' own investor disclosures.

Pacific Gas and Electric Company's second-quarter 2026 earnings presentation, filed with the U.S. Securities and Exchange Commission, reports more than 12 gigawatts of data-center demand in its service pipeline. The same presentation shows 490 megawatts with executed interconnection construction agreements.¹ Those two numbers describe the era better than any editorial could: demand for new capacity has outrun the institutional machinery that delivers it, by roughly an order of magnitude, in the disclosure of the utility itself.

Regulators know it. In July 2025 the California Public Utilities Commission approved interim Electric Rule 30, a streamlined framework for connecting large transmission-level loads such as data centers in PG&E territory, with applicants funding transmission work up front.² In June 2026 the Federal Energy Regulatory Commission opened targeted proceedings to speed the integration of large loads onto the bulk power system nationally.³ And the utilities themselves are building a new product category: connection offers that trade firmness for speed, such as PG&E's Flex Connect program and its announced transmission-level counterpart, T-Flex, which connect flexible loads sooner in exchange for agreed curtailment behavior.⁴

Each of these developments is genuinely useful. Together they also mean something that owners should sit with: the menu of ways to power a site is longer and less standard than it has ever been. Full service, flexible service, phased service, bridged service, self-supply, hybrids of all five. A menu that long, moving that fast, is precisely the environment in which a single-technology recommendation should be treated as incomplete until proven otherwise.

Section 03The eight paths every large load should see priced

A defensible power decision starts with the full menu. In our study work the menu has eight entries. Not all eight survive first contact with a real site, but each one earns its elimination in writing, with a reason attached, rather than by never being considered.

A note on path five, because it is where neutrality dies first

On-site generation is not one option. It is a family of options with materially different profiles, and the honest comparison inside the family matters as much as the comparison across paths.

• Reciprocating engines are typically the lowest capital cost per unit of capacity among the common choices, with short delivery times and a deep service ecosystem. They are also combustion machines, and in California's stricter air districts new combustion capacity faces demanding control-technology requirements and, in some cases, offset obligations that can slow or foreclose the permit. Noise and scheduled outages belong in the model, not the footnotes.
• Gas turbines and microturbines are compact and produce high-grade exhaust heat that rewards sites with real thermal demand. Efficiency at small scale and at part load deserves scrutiny, and they remain permitted combustion equipment.
• Fuel cells are non-combustion electrochemical plants with near-zero criteria-pollutant emissions, which in several California districts translates into materially lighter air permitting. They run quietly at high electrical efficiency and pair well with heat recovery. They carry higher capital cost than combustion peers, their economics remain exposed to gas price, and stack-replacement schedules are a real lifecycle line item.
• Linear generators are fuel-flexible, modular, and fast-starting. As a newer class of equipment, the diligence burden shifts to fleet operating history and service depth, which a study should examine rather than assume in either direction.
• Solar with storage burns nothing and buys no fuel, and for daytime-weighted loads it can be decisive. For continuous 24/7 demand its capacity factor means it rarely stands alone; land or roof area becomes the binding constraint faster than most first models expect.

Every one of those sentences cuts in a different direction. That is the point. A study whose technology section reads like a brochure for any single row of that list was written by a seller, whatever its cover page says.

A seller's feasibility study is a sales document. The owner who treats it as a decision document has outsourced the judgment that matters most.

Section 04Neutrality is a discipline, not a slogan

Nearly every energy firm describes itself as technology-agnostic. The claim is cheap because it is unfalsifiable in a brochure. It becomes falsifiable in a work product, and there are three places to look.

Follow the compensation. If the party performing the analysis earns more when one particular technology wins, the analysis is a proposal. This is not an accusation of bad faith; it is how incentives work on everyone, including us. It is why we sell the study as the product, at a fixed fee, rather than giving analysis away to sell equipment behind it.

Look for the paths that make the analyst nothing. The no-project case, the relocation case, and the plain utility-service case earn a technology vendor exactly zero dollars. If they are missing from the document, the document was never a decision aid. In our framework they are mandatory rows precisely because they are the rows a seller cannot afford to include.

Check whether the requirements were written before the answer. A neutral process fixes the site's actual requirements first: the load's shape as measured, the deadline as dated, the constraint as documented, who controls the site and meter, and who can actually sign. Only then does it ask which paths satisfy them. A process that starts from a product and works backward will present the product's strengths as if they were the site's requirements. Owners can detect this in one move: ask which requirement, if it changed, would change the recommendation. If no answer comes, the recommendation was never derived from requirements.

Section 05The seven-question test

Any owner can run this test on any energy recommendation in thirty minutes, without an engineering degree. We publish it because it is as binding on us as on anyone; question two does not have a carve-out for the person writing this paper.

1. Were at least three genuinely different paths priced, including no-project?Three technologies from one family do not count. Grid, self-supply, and doing nothing are different paths. Engine A versus engine B is one path with two logos.
2. Does the analyst earn differently depending on which answer wins?If yes, you are reading a proposal. Proposals are legitimate documents; just never let one be the only analysis in the room.
3. Is the utility's own answer documented in writing?A dated capacity and timeline statement from the utility, not a recollection of a phone call. The written grid case is the benchmark every alternative must beat.
4. Does every load-bearing number carry a source and a date, and is every estimate labeled as one?Firm and estimated figures mixed without labels is the single most reliable marker of a sales model.
5. Was the load measured rather than assumed?Interval data, not nameplate ratings or annual totals. The load's shape decides more than the technology choice does.
6. Did each candidate pass a site-specific permit and fuel screen?Air-district treatment, gas availability, land, noise, and water are site facts. A study that never names your air district never studied your site.
7. Is the analysis built backward from a dated deadline?Time-to-power is the scarce commodity. An analysis without a date optimizes cost per kilowatt-hour while the actual constraint goes unmanaged.

A recommendation that passes all seven can still be wrong; engineering risk never goes to zero. But it will be wrong the honest way, on the merits, rather than wrong by construction.

Section 06What a defensible study contains

The work product that survives the test has a consistent anatomy, whoever prepares it:

• A measured load basis. Interval data where it exists; a stated, defensible proxy where it does not, labeled as such.
• The paths matrix. Every credible path from Section 03, priced with sourced and dated inputs, including the three that pay the analyst nothing.
• Sensitivity on the assumptions that move the answer. Usually fuel price, the energization date, and load growth. If the recommendation flips inside a plausible range, the study says so plainly.
• A permitting and fuel screen by district and by site. Named district, named rules, current as of a stated date.
• An incentive screen at statutory values. Under current federal law the investment tax credit for qualifying clean-energy property is 30 percent; bonus adders exist in statute but must be individually qualified, not assumed.⁵ A study that books adders it has not qualified is padding someone's margin. Program-level incentives change; every incentive line carries its as-of date.
• A decision memo. One recommended path, the conditions that would change the answer, and the next three actions with owners and dates. Analysis without a decision is research; owners are buying a decision.

One structural feature matters more than any section: the party writing it should have nothing to gain from which technology wins. Independence is not a virtue we claim; it is a fee structure. Fixed price, paid for the decision, with no equipment margin behind it.

Section 07Neutrality is the authority

The next decade of California energy will be decided at sites like yours, one bounded decision at a time, under deadlines the grid did not used to impose. In that environment, authority does not belong to whoever has the largest catalog. It belongs to whoever tells the truth about alternatives, including the alternatives that pay them nothing.

Our founding conviction is that green energy is only sustainable if it is profitable, and profitability begins with an honest comparison. That is the entire method: all the paths, sourced numbers, dated facts, one decision. Everything else in our field is marketing.