On-site power, the argument concludes, is a depreciating monument to a temporary problem.
The objection is right about the timeline and wrong about the asset class, and the error hides in one word: depreciating. Walk through what actually happens to the value of an existing machine over the next five years, and the objection inverts into the thesis.
Accounting depreciation, strategic appreciation
Every generation asset on a balance sheet follows a depreciation schedule written for the world that existed when the schedule was drafted: stable equipment markets, normal lead times, replaceable machines. That world is gone for the rest of this decade, and the current market has put numbers on its absence. The figures that follow are Oculus market data, on the same basis as The Turbine Gap. A machine acquired at the end of 2025 at roughly $1.2 million per MW replaces today, for buyers who can even secure new allocation, at more than $2 million per MW, while equivalent existing equipment clears at $1.4 million. That is sixty-plus percent replacement-cost appreciation over the same months in which the accounting schedule marked the asset down. The strategic value of an existing, mobile, serviceable unit has been rising through the entire period its book value has been falling, and as The Turbine Gap documents, the shortage driving that divergence is manufacturer policy, not a passing squeeze.
Source: Oculus market data.
The year-six stranded asset is an appreciated position wearing a depreciated label.
The buyers who understand this are not asking how to avoid owning machines in year six. They are asking how to be the ones holding them. What follows is the option frame: a correctly specified bridge fleet is not a structure, it is an option with three exercises, and the exercises compound.
Exercise one: redeploy
This is where the fleet view replaces the site view, the single largest mental-model upgrade available in this space. A generation fleet across multiple sites is not a collection of stranded per-site assets; it is a system with an internal market. The aeroderivative units that bridged Site A through its energization gap become the time-to-power answer for Site B, in a market where machines that exist command a structural premium for the rest of the decade. Run the portfolio math: a fleet that bridges three sites sequentially over nine years has tripled its deployment value against its acquisition case before any residual sale. The full redeployment cost stack, transport, recommissioning, service-agreement transfer, is knowable to the week and the dollar for anyone who has actually moved fleets; it is simply not published, which tells you something about how much it is worth to the parties who hold it. And the redeployment market is not limited to your own portfolio. The phantom-load dynamics laid out in The Queue Is the Product guarantee a steady supply of failed projects and desperate schedules on the other side of the trade; a mobile fleet is effectively a market maker in time to power, positioned to price other people's emergencies for the remainder of the shortage. That is not a salvage strategy. It is the business model the operator layer described below is being built on.
Exercise two: backstop
Grid arrival does not make on-site generation worthless; it changes the machine's job description, and the new job now comes with a published salary. Capacity in PJM has cleared at the price cap for three consecutive auctions, translating to roughly $120,000 per MW-year at current levels, which means a 200 MW behind-the-meter fleet held as backstop earns on the order of $24 million annually in capacity value before selling a single megawatt-hour of energy. Firm, dispatchable capacity behind the meter is resilience, peak management, and above all negotiating position: a facility that can serve its own load buys grid power on its own terms in every tariff discussion, forever. The regulatory architecture emerging from the orders detailed in Behind the Meter Goes Mainstream makes this exercise more valuable over time, not less. The new service categories for flexible large loads are, by design, structures that reward loads capable of managing their own profile, and in the affordability politics now engulfing every growth market, the load that can flex off the grid at peak is the load regulators hold up as the model citizen. The backstop exercise is also what protects your regulatory vintage: the facility that never fully surrenders its self-supply capability negotiates every future obligation from a position the pure grid-taker never has.
Exercise three: convert
Generation that started life as your bridge can finish life as capacity you meter rather than own: serving your load, or a neighbor's, under per-kilowatt-hour arrangements, with ownership and operation migrating to parties whose entire business is running machines. The financialization sequence traced in From Capex to Opex runs in both directions and at any point in an asset's life, not only at purchase; owned assets convert to consumed services whenever the structure serves the balance sheet better, and the operator layer forming in this market exists to take the other side of exactly that trade. The forward call: as dedicated generation becomes an institutional asset class, the natural buyers of seasoned bridge fleets will be the same infrastructure and private credit capital that absorbed data center shells, which means the conversion exit gets deeper and better-priced every year the asset class matures. The buyer who models that exit on day one preserves it. The buyer who thinks in ownership terms until year six discovers that exits designed under deadline pressure price accordingly.
Mobility is a specification, not a hope
The exception is unforgiving, and it is where lifecycle theses actually die: assets bought wrong do strand, and they strand on details that never make the investment committee deck. A configuration married to one site's water supply is the canonical failure; continuous demineralized-water demand can bind a machine to a location as firmly as its foundations, and water math has quietly killed more redeployment cases than any market variable. The same applies to single-fuel configurations, site-specific interconnection topology, and service agreements that do not transfer. Every one of these is a procurement decision, made years before it matters, cheap to get right at signing and ruinous to fix at repowering. Mobility, fuel flexibility, and LTSA transferability belong in the specification with the same weight as output and heat rate, because they are the raw material of all three exercises above. A machine without them has one exercise, and it is scrap.
The year-six discipline
The underwriting rule compresses to one sentence: model year six before you sign year one. Every bridge deployment should carry, on day one, a written answer to what the fleet does when the grid shows up, with the redeploy, backstop, and convert cases each priced against the appreciation-versus-depreciation divergence this market has already handed you. If the answer is credible, the stranded-asset objection becomes the strongest argument for the structure: you are buying time now, which The Queue Is the Product prices at roughly the cost of the machine itself for every month saved, plus an option whose strike the market keeps moving in your favor. The objection assumed the asset dies when the grid arrives. Specified correctly, that is the day it starts its second career, in a market that will still be paying a premium for the fact that it exists.