The repair-flow wedge needs a removable unit, not just a valuable one

Exploration input for the founders. Refines the earlier analog scan by adding one gate. Per RDI methodology this surfaces patterns and ends in a question; it does not conclude.

The proposed gate

The prior scan matched industries on: B2B, high unit value (so repair beats replace), a dwell-heavy triage flow, and capital trapped in that flow. Dustin’s refinement: the value has to be high enough that you repair instead of replace, but not so high that the servicing model flips to deploy a technician on-site instead of pulling the unit and cycling it through a central repair flow.

The gate is real. But the variable that actually decides on-site-vs-send-out isn’t dollar value — it’s removability: whether the unit is a modular, swappable part (a “line-replaceable unit” / rotable) or a fixed, integrated, installed asset. Value only correlates loosely, and one counterexample breaks the value framing cleanly:

A jet engine is the highest-value item on the list ($12–45M [Public: IBA / SafeFly, 2024–2026]) and they still pull it and ship it to a shop — it’s a rotable. A gas turbine of similar value is bolted into a power plant and gets an on-site field-service outage. Same value, opposite servicing model.

So the predictive variable is form factor / removability, not a price ceiling. The gate is right; restate it as: the unit must be modular enough to swap in the field and cycle a spare through a central pool. That is aviation’s own LRU/rotable concept, and it is exactly what the financing wedge attaches to.

Candidates re-run through the removability gate

CandidateRotable / removable?Verdict under the gate
Datacenter / networking (GPU / board / server)Yes — swap the board, RMA the dead oneStrengthened — purest LRU flow (our anchor)
Aviation — engines, APUs, avionicsYes — explicitly removed and shoppedPasses — rotable side is the target
Heavy-equipment reman (Caterpillar)Yes — core exchange on engines / transmissionsPasses — component cores cycle
Rail (traction motors, engines)Yes — UTEX / UX unit-exchangePasses — literal exchange pool
Oil & gas downhole rental toolsYes — rental-return-inspect-repair loopPasses — holds up
Telecom / RAN (RRU / baseband)Yes — removed and depot-repairedPasses form; still weak on unit value
Medical imagingSystem = no (magnet serviced on-site); coils / tubes / boards = yesRefined — target the component-exchange layer, not whole-system service
Semi capital equipmentTool = no ($150M EUV is field-serviced in the fab); modules / parts = yesRefined — same split; target the parts / refurb flow
Wind turbineMostly no — up-tower / crane-down field workDowngraded — the gate exposes this as mostly on-site
Power-gen turbinesNo — on-site outage modelFails — confirms the weak read
ElevatorsNo — installed, serviced in placeFails — gate confirms it

How the gate changes the recommendation

1. It confirms the top tier and tightens the reason. The best analogs aren’t just “high value” — they’re rotable-pool / unit-exchange businesses: datacenter / networking (anchor), aviation rotables and avionics, rail unit-exchange, Cat core-exchange, downhole rental. Those are where trapped capital literally rotates through a pool — the strongest attach point for financing.

2. It reframes medical imaging and semi-cap-eq from “target the system” to “target the parts layer.” The whole MRI or the whole EUV tool fails the gate (both get on-site field engineers). But the coil / tube / board exchange in imaging, and the module / refurb-part flow in semi-cap-eq, both pass. That’s a narrower but sharper target — and it de-risks the “too few buyers” problem, because component-exchange pools serve a wider base than the handful of fab or hospital-system buyers.

3. It drops wind, power-gen, and elevators. All three lean on on-site field service, so there’s no rotating trapped-capital pool to lend against. Wind is the one genuine loss — it had strong parametric-downtime characteristics (weather variability) even though it fails the trapped-capital gate. Keep it in a separate “parametric insurance, not financing” bucket rather than discarding it.

Net: the gate doesn’t shrink the funnel so much as re-sort it — it promotes the pure rotable-pool plays, splits imaging / semi-cap-eq into “system (out) vs component (in),” and cleanly explains why the installed-asset industries never fit.

The open question

Do we treat parametric downtime insurance as a separate wedge with its own funnel? That wedge doesn’t need the removability / trapped-capital gate — wind, power-gen, even elevators qualify for it. This gate is really a filter for the financing / trapped-capital wedge specifically; applying it to the whole thesis would throw out industries where the insurance wedge still works.

Refines the oil-analog scan and the analog work built on the NVIDIA reverse-logistics case (reverse-logistics-warranty-tam-2026-05-29, rma-reverse-logistics-warranty-software-landscape-2026-07-08).