Insurance prices from history. It knows what a transformer costs because thousands have run, failed, been repaired, and because that experience accumulated somewhere. A technology with no history therefore has no price, and that is a substantive difficulty, not a question of appetite: nobody knows what they are insuring. Recent large energy projects raise this question daily, and the market's answers are more interesting than a simple decline.
The first answer is an exclusion, often headed unproven technology or prototype. It rules out equipment that has not run in commercial operation for a stated period, or has not been deployed in a stated number of units. The logic holds: while no experience exists, the premium can only be a guess, and a systematically optimistic guess is corrected at the first serial loss. But the clause immediately raises a question it does not answer: from when is a technology proven?
The question is not rhetorical, because industrial innovation almost always proceeds by change of scale rather than by rupture. A fifteen megawatt turbine is not a new machine: it is the same architecture as an eight megawatt one, with longer blades, a taller tower and forces that do not grow linearly. The manufacturer will present it as proven, since the principle is; the underwriter will object that what breaks in a hundred-meter blade did not break in a sixty-meter one. The unproven technology clause does not settle that debate, it starts it, and the definition retained in the contract decides everything else.
The second answer bears on what makes a series frightening, and it is not the failure of one machine. A design defect does not appear once: it is present in every unit from the factory gate, and it waits. A serial loss is therefore an event of a particular kind, where the question of how many losses drives the amount far more than the severity of each. Is the same weld cracking on sixty foundations one loss or sixty? The answer depends on an aggregation clause, and the gap between the two readings is measured in deductibles multiplied by sixty on one side, and a single limit exhausted on the other.
The third difficulty is a calendar problem insurance cannot correct. An industrial series deploys faster than experience accumulates. By the time the defect shows on the twelfth unit, the next forty-eight are already manufactured, often already installed, and the whole fleet carries the same defect. The exposure is not the failed machine, it is the fleet, and it was built before anyone could know. No clause catches that up: it is why underwriters look at the deployment rate as much as at the technology itself, and why a schedule installing everything in eighteen months does not read like one spread over four years.
A fourth question then arises, and it returns to a principle already met. Once the defect is identified on twelve units, what of the other forty-eight, which have suffered nothing but carry the same flaw? Replacing them preventively is not repairing a loss: it is correcting a defective delivery, therefore a matter of manufacturer warranty and recourse, not of damage insurance. The policy pays only for what was damaged, whatever the technical certainty that the rest will follow. That boundary is hard to hold in practice, because waiting for sixty failures in order to repair sixty machines is absurd industrially and perfectly logical contractually.
That leaves what the market makes of all this, which reads in the structure rather than in the price. Capacity is given in stages, following the deployment rather than covering the final fleet from the first machine. Deductibles split in two, one per unit and an aggregate per period, which bounds the serial effect without denying it. And cover often rests on a type certification issued by an independent body, not because that document proves the absence of defect, but because it supplies the only objective marker in a file where everything else is declaratory. A project owner presenting a series project without such a marker is asking an insurer to take its word on the one question it cannot verify.
A battery storage farm of 60 identical containers is installed over eighteen months. In the fourteenth month, thermal runaway destroys container number 12 and damages its two neighbors, for 2.7 million euros. Expert examination establishes a design defect in the cooling system, present on all 60 containers. At that date, 51 are installed and 9 in manufacture. The policy carries a 250,000 euro deductible per loss and a 15 million limit per period. The manufacturer offers to rework the cooling of the 57 undamaged containers free of charge, provided the operator bears removal, reinstallation and unavailability, that is 4.1 million. The operator asks its insurer to pay those 4.1 million on the ground that the loss is certain. What happens?
On the 2.7 million the matter is simple: three damaged containers, one triggering event, a 250,000 euro deductible, the insurer pays 2.45 million. On the 4.1 million the answer is no, and it admits no nuance: 57 working containers have suffered no damage. Their rework corrects a defective delivery, it is a matter of manufacturer warranty, and the manufacturer's own offer implicitly concedes it since it proposes to supply the part. That the future loss is technically certain changes nothing: a damage policy indemnifies what is damaged, never what is going to be. The operator's real negotiation is therefore not with its insurer but with its manufacturer, over removal and unavailability, and it is conducted on the supply contract. That leaves a question nobody has yet asked and which is worth more than the previous two: if the operator does not rework the 57, will the next runaways be that many separate losses with that many deductibles, or one serial loss exhausting the 15 million limit? The aggregation clause decides, and depending on its wording the operator has an interest in accepting the manufacturer's offer or in refusing it. That is the clause that should have been read before ordering sixty identical containers.
- 01A technology with no history has no price: the unproven technology exclusion is not a lack of appetite, it is the admission that nobody knows what is being insured.
- 02Industrial innovation proceeds by change of scale, and the clause does not settle whether an enlarged machine stays proven: it starts the debate, the contractual definition ends it.
- 03A serial defect is present in every unit from the factory gate: what drives the amount is the aggregation clause, one loss or sixty.
- 04The series deploys faster than experience accumulates: by the time the defect shows on the twelfth unit, the whole fleet already carries it.
- 05Preventively replacing undamaged units is not a loss, whatever the technical certainty: it is the manufacturer's warranty, and the negotiation runs on the supply contract.