A thirty-year concession should buy its technology in ten-year pieces

Every infrastructure owner has a version of the same drawer. In it are binders from the original systems integrator: point lists, printouts of control logic, a commissioning report signed by people who left the industry long ago. The controllers they describe are still running. The manufacturer stopped supporting them years ago, and the engineers who could change them safely can be counted on one hand. Nobody decided to arrive at this position. It is what happens when technology is bought as though it had the lifetime of the concrete around it.

My argument is simple to state and harder to apply. A concession that runs for thirty years should not buy its technology in one thirty-year piece, nor even in one fifteen-year piece. It should buy it in pieces that match how long each layer really lasts, and it should own the joints between those pieces so that each can be replaced without reopening the others. I have watched this question from both sides: for thirteen years as a supplier selling integrated systems, and for nearly twelve as the owner who inherits them.

Lifetimes that do not match

On a motorway the layers of technology age at very different speeds. Cable ducts, gantry foundations and tunnel cable trays may last as long as the road. Field devices such as cameras, detectors, fans and signs typically need replacing after fifteen to twenty years. The controllers that coordinate them may manage ten to fifteen. Operating systems, databases and application software run on a cycle of five to eight years, and cloud and telecommunications contracts are renegotiated every three to five. Those figures vary with environment and maintenance regime, but the ratio between them changes surprisingly little.

An integrated procurement smooths these differences into one price and one completion date. That looks attractive to everyone at signing. The grantor sees a single point of responsibility. Lenders see a fixed-price contract with a creditworthy counterparty. The operator sees one supplier to call. The difficulty comes later, when the software layer needs replacing and it turns out that it can only talk to the controllers through a protocol the original supplier owns, or that the controllers accept updates only from tooling licensed by the same supplier. At that point the owner is no longer buying a replacement. It is negotiating with a supplier who knows the owner cannot leave.

When we replaced the tunnel control system on our northern Portuguese concession between 2021 and 2023, a large part of the effort went into reconstructing what the existing system actually did. The documentation described the design intent of 2007; years of later modifications lived in the code and in the memories of two engineers. We spent several months turning that knowledge into control narratives and interface specifications before we could ask anyone to price a replacement. That work should have been a deliverable of the first contract, and I say so as someone who helped write it.

Own the seams, not every component

The answer is not for owners to become integrators. Elvaran does not want to write controller logic, and most concessionaires should not try. What we can own is the definition of how the pieces connect: the data model, the interface specifications, the control narratives that describe in plain language what each system must do in each scenario, and the test scripts that prove it. If those documents belong to us and are kept current, any competent supplier can bid for the next piece.

In practice that has changed our tenders in four ways. We require interface documentation to be delivered in a form we can maintain, and we pay for it as a separate line so that it is not quietly reduced in scope. We ask every bidder for an exit plan alongside the delivery plan: what we would need, and in what form, to replace their system in ten years without their help. We avoid support contracts whose price depends on our inability to leave. And we write the control narratives ourselves, with operators and tunnel safety officers, before procurement begins.

This costs money and time. A supplier asked to deliver open documentation will price that risk, and our own staff must do work an integrated contract would have hidden. On the tunnel replacement I estimate the owner-side effort at around €1.6 million over three years. Set against the cost of being unable to compete the next replacement, I regard it as one of the better investments we have made.

There are fair objections. A single point of responsibility does reduce arguments when something fails at the boundary between systems, and grantors are right to worry about an owner who splits accountability so finely that nobody answers for a fire scenario. My response is that splitting procurement need not mean splitting assurance. For each tunnel we still name one integrator responsible for end-to-end scenario testing at every cutover, and the scenarios themselves are ours.

The last replacement is a handback decision

The question becomes sharper as a concession approaches its end. Our coastal concession returns to the grantor in December 2031. Any system we install now will be handed over in its sixth or seventh year, and the grantor, or the next operator, will inherit our choices exactly as we once inherited someone else's.

Two temptations appear in those final years. One is to defer renewal and hand back systems that still function but are close to obsolescence, leaving the problem to the successor. The other is to buy something expensive and proprietary late in the term, because the cost falls on a shrinking number of revenue years and the consequences fall on someone else. Both are poor stewardship, and grantors increasingly write handback clauses on remaining useful life to prevent them.

I would rather plan the final replacement cycle backwards from the handback date. That means knowing which systems will cross the line with how much life left, agreeing early with the grantor what documentation and data will transfer, and choosing technology the next operator could maintain without us. It is in our interest as well as theirs: an orderly handback is one of the things a grantor remembers when the next concession is tendered.

None of this is glamorous. Interface specifications do not appear in investor presentations, and an exit plan sounds pessimistic at a launch meeting. But the infrastructure owners I respect are the ones whose second replacement went more smoothly than their first. That is the measure I apply to our own work. A concession is a long contract with the public, and its technology should be bought by people who expect to be around, or to be succeeded, when each piece has to be renewed.