Capgemini's $4.8m capitulation

On February 1st Capgemini, one of France's largest listed companies, announced it would sell an American subsidiary rather than defend it. The contract at the centre of the row was worth roughly $4.8m. The group's 2025 revenue was €22.5bn. The gap between those two numbers is the real story.

The subsidiary, Capgemini Government Solutions (CGS), had landed a deal with US Immigration and Customs Enforcement (ICE) on December 18th to provide "skip tracing services for enforcement and removal operations" — locating foreigners whose whereabouts were unknown so they could be picked up and deported. It was one of 13 contracts the unit held with ICE, and was due to run until March 15th. Skip tracing is unglamorous, data-heavy work: checking addresses, reconciling records, handing an agency the last-known location of a person who does not want to be found. There is nothing about it, in theory, that a technology firm could not market with a clear conscience.

The conscience was the problem. In late January two US citizens — Renee Good, shot by an ICE agent, and Alex Pretti, shot by Border Patrol officers — died in Minneapolis. The killings set off protests, and attention turned to the private firms that feed the enforcement machine. The Multinationals Observatory, a French campaign group, revealed the CGS-ICE contract and noted that CGS's final remuneration depended on how many people were helped to be detained and deported — "much like a bounty hunter", it said. It also claimed the unit had supplied skip-tracing services to the administration before the December signing, undercutting Capgemini's suggestion that it had only just learned what its own subsidiary was doing.

The political temperature in France rose fast. The economy minister, Roland Lescure, called for transparency; a left-wing MP demanded sanctions on French firms that work with ICE; questions were asked in parliament. Capgemini's workforce was already raw: on January 20th the group had announced up to 2,400 job cuts in France. An extraordinary board meeting over the weekend produced the decision to sell.

Capgemini's stated reason deserves a close reading. It said that "customary legal restrictions imposed for contracting with federal government entities carrying out classified activities" had stopped the group from exercising "appropriate control" over parts of the subsidiary's operations. The chief executive, Aiman Ezzat, elaborated on LinkedIn: decision-making at CGS was separate, its networks were firewalled, and the parent group could not access classified materials.

Translate that corporate language and it is a confession. US federal contracting law makes classified work structurally opaque to the parent company; the firewall is a condition of winning the business, not a preference. Capgemini is admitting that it ran a subsidiary it could neither see into nor control, and that this blindness — not any moral objection — is why the deal had to go. The same mechanism that let the company disclaim knowledge ("we learned from public sources") is the mechanism that stopped it governing the business in the first place.

That matters because the American market is where Capgemini's future lives. North America generated 29% of group revenue in 2025, growing 7.3% in constant currency and posting the group's healthiest operating margin, near 17%. In the first half of 2026 the region grew almost 20% in constant currency, a pace that has more to do with commercial IT and artificial-intelligence contracts than with immigration work. Capgemini cannot walk away from the United States any more than it could keep this one subsidiary. What it can do — what it has just done — is ensure that no single contract is big enough to hold its home-market reputation hostage.

Hence the investment reading. On the numbers, the sale changes nothing: CGS is about 0.4% of group revenue and less than 2% of US revenue, and the disputed contract is a rounding error in a book near €22.5bn. No shareholder should reprice the stock over it. The meaning is in what it reveals. It caps the upside on one fast-growing, politically sensitive corner of the US market — the digital machinery of deportation — that, under an administration promising more removals, was a niche Capgemini could no longer credibly chase. Few European contractors will envy that concession, and fewer still will be able to enforce it.

There is also a governance lesson tucked into a footnote of a footnote. Every diversified services firm carries subsidiaries that outsiders — and sometimes the parent itself — cannot fully see. The classified firewall is simply the extreme case. What Capgemini priced as near-zero was not just a small revenue line but its own inability to govern one of its own businesses. It sold cheap because the business was tiny and the politics were radioactive; the arithmetic made the surrender rational. For a company earning nearly a third of its revenue in Washington's market and its reputation in Paris, the scarcest input is neither code nor capital but permission. And permission, unlike revenue, does not scale.