Bolloré Owns 93% of the Top Holding Company. Only the 29.9% at the Bottom Went to Court

Vincent Bolloré's family owns 93.05% of Compagnie de l'Odet.¹⁴ Nobody has ever filed a claim about that. Compagnie de l'Odet owns 71.6% of Bolloré SE.¹⁵ Nobody has filed a claim about that either. Bolloré SE owns 29.9% of Vivendi, and that number has occupied the Autorité des marchés financiers, the Paris Court of Appeal twice, and France's supreme civil court, over a question worth an estimated €6 billion to €9 billion.²

Read those four sentences again, because the asymmetry is the whole story. Three links in a chain, three fully disclosed percentages, and the regulatory apparatus of the French Republic has spent twenty-one months and counting on the one closest to the number 30. The other two links, the ones where the family's grip is total rather than marginal, have never been anyone's problem.

That is not an oversight. It is what the rule is built to do, and once you see it you cannot unsee it in your own structure.

Start at the top, because that is where the ownership is least contested and most complete.

Sofibol, a family holding vehicle, crossed two-thirds of the voting rights in Compagnie de l'Odet in August 2024, ending up with 56.07% of the capital and 70.63% of the votes. Add the rest of Vincent Bolloré's family entities and the total reaches 93.05% of the capital and 71.47% of the votes.¹⁴ Compagnie de l'Odet's own bylaws grant double voting rights to shares registered in the same name for four years, which is how a 56% capital stake becomes a 70.63% voting bloc without anyone buying an additional share.

Compagnie de l'Odet, in turn, held 70.5% of Bolloré SE's capital at the end of 2024 and 71.6% at the end of 2025.¹⁵ Bolloré SE holds 29.9% of Vivendi.² Underneath Vivendi sit the operating businesses: Canal+, Havas, Louis Hachette Group, and until 2024, Universal Music Group.

Multiply the disclosed layers and the family's ultimate look-through economic stake in Vivendi lands somewhere under 20% of the company, built on top of a structure it owns outright at every level above it. We are showing that arithmetic to make the shape visible, not publishing it as a precise figure. Each rung comes from a different filing in a different jurisdiction, and the site's own Control Wedge methodology deliberately will not score a chain like this, because a wedge is defined against one company's one beneficial-ownership table, not four companies' worth of them stapled together. That refusal is itself informative. The instrument built to expose the gap between votes and economics cannot be pointed at a cascade at all, because nothing forces the cascade to report itself as a single number. It reports as four numbers, in four documents, and reassembling them is a research project rather than a filing.

Which is the actual function of the structure. It was never built to hide that the family controls everything above Vivendi. Every one of those percentages is sitting in a public disclosure right now. What the cascade does is let the family own 93% of one entity, 71.6% of the next, and calibrate only the last link to two decimal points, because the last link is the only one wired to a tripwire. Ninety-three percent of a private holding company answers to nobody. Twenty-nine point nine percent of a listed one sits a tenth of a point from a mandatory bid that would cost billions.

Bolloré did not start at 29.9%. He built there, deliberately, over a decade.

He took the Vivendi chairmanship in 2014 with a stake that started around 8%.¹¹ By April 2015 he had spent €2.84 billion in six weeks to reach 14.5%.¹³ By 2016 the stake was 20.4%, and he was pushing for the loyalty-share mechanism that French law makes available to any company: double voting rights for shares registered for at least two years, unless two-thirds of shareholders vote it down.¹² Once that mechanism matured in April 2017, a stake in the low twenties converted into voting power in the high twenties, and reporting from the period describes him as having achieved effective control that year.¹²

Notice what did the work. Bolloré did not need to buy his way past 30% of the capital. He needed to buy enough capital that the statutory loyalty bonus, which every French listed company can adopt and which shareholders approved for Vivendi specifically because two years is a long time for an activist fund to stay put, would do the rest. The double-voting default rewards patience, and patience is the one resource a controlling family always has more of than an index fund with a quarterly redemption cycle.

By the time of the 2024 restructuring, the capital stake itself had settled at a number that reads less like a coincidence and more like a specification: 29.9%, a tenth of a point under the French mandatory-bid trigger.² A structure this carefully engineered does not arrive at a number like that by accident, and nobody involved has ever claimed it did.

For most of the previous decade, Bolloré's stake in Vivendi was a large minority position in a conglomerate that also, at various points, held Universal Music Group, Havas, Canal+, and stakes in Telecom Italia and Lagardère. A dispute over "control" of that sprawling asset would have been a dispute over a great deal of everything.

The December 2024 restructuring changed the target. Vivendi's shareholders approved, at 97.57% and 97.58% for the two resolutions, the partial demerger of Canal+ and Louis Hachette Group, alongside the separate listing of Havas.⁵ Shares of Canal+ began trading on the London Stock Exchange, Havas NV on Euronext Amsterdam, and Louis Hachette Group on Euronext Growth Paris on December 16, 2024, with settlement to Vivendi's own shareholders two days later.⁶

What remained was smaller and stranger. The reconstituted Vivendi reported a net asset value of €4.83 billion at the end of 2024, sitting on top of an investment portfolio worth €7.12 billion, wholly owning the video game publisher Gameloft, and holding minority stakes in Universal Music Group, Banijay, MFE, Telecom Italia, Telefónica, Prisa, and Lagardère.⁷ It is, in other words, a holding company for other people's operating businesses. Bolloré SE kept its 29.3% stake in that smaller thing.²

Shrinking the company did not shrink the family's percentage. It could not, arithmetically; a demerger distributes shares pro rata, so a 29.9% holder of the whole stays a 29.9% holder of what is left. What it did was concentrate the argument. Vivendi's own materials described the split as unlocking shareholder value by letting each business be valued on its own terms.⁷ A side effect nobody put in the press release was that it also let one shareholder's stake be argued over on its own terms, undiluted by three other businesses' worth of noise. A €4.8 billion investment holding company, with most of its float already accounted for by the same family's declared stake, is a much easier thing to describe as "controlled" than a diversified media conglomerate with four operating divisions, and CIAM made exactly that argument to the AMF within weeks of the split closing.

France's answer to "who controls this company" is not a vibe check. Article L.233-3 of the Commercial Code sets out a specific, mechanical test: a person controls a company when they determine, in fact, the decisions taken at that company's general meetings through the voting rights they hold.¹⁰ That is a backward-looking, evidentiary question about actual meetings, not a forward-looking judgment about influence.

CIAM's theory, run through the AMF's article 236-6, was that Bolloré SE's post-demerger position satisfied that test and should trigger a mandatory withdrawal offer for the rest of Vivendi's shares.¹ The AMF disagreed on November 13, 2024, ruling that Bolloré SE did not control Vivendi within the meaning of article L.233-3.¹ CIAM appealed, and on April 22, 2025 the Paris Court of Appeal reversed the regulator, finding de facto control. Part of that court's reasoning turned on Vivendi's own treasury shares, the stock the company holds in itself and cannot vote: exclude that stock from the denominator, and Bolloré's effective share of the votes that actually count starts to look like more than 30%.

Then the AMF, following the appellate court's instruction, went further. Its second decision on July 18, 2025 ordered Bolloré SE and Vincent Bolloré personally to file a public withdrawal offer for the whole of Vivendi, on the explicit ground that Vivendi's own 3.7% of self-held shares should be counted toward the 30% threshold calculation, and it gave a filing deadline of January 18, 2026.¹⁷ Analysts put the resulting bill at €6 billion to €9 billion.²

That is where the story would have ended, with a forced buyout, except that Bolloré appealed to the Cour de cassation. On November 28, 2025, France's supreme civil court quashed the April 2025 ruling. Its holding was narrow and specific: de facto control under article L.233-3, I, 3° must be assessed solely on the basis of votes actually expressed at general meetings, and no other criterion, meaning the personal prominence and broader influence factors the Court of Appeal had leaned on were the wrong test entirely.⁸⁹ The case went back to a differently composed panel of the Paris Court of Appeal.

On July 8, 2026, that panel ruled that Vincent Bolloré and Bolloré SE do not control Vivendi.³ Vivendi's stock fell roughly 10% within the day.³ Sit with that for a second, because it is the driest part of this entire file. A ruling that a shareholder does not control the company, which is supposed to be reassuring news for everyone who is not that shareholder, made the stock go down. It went down because the minority shareholders CIAM was nominally protecting had spent the previous year pricing in a mandatory buyout at a premium, and the court just told them the check was not coming. The rule exists to protect minority shareholders from being trapped under an unaccountable controller. Its correct application, in this instance, is what trapped them. CIAM was also ordered to pay Bolloré's side €350,000 in costs, and said it would appeal again.³

Two years, two courts, one regulator, and the entire proceeding turned on a definition that measures a single fact: who showed up to vote at the last two annual meetings, out of a shrinking pool of shareholders who bothered to show up at all.

Here is the detail that should bother a founder more than the litigation does. Bolloré's percentage of Vivendi did not hold still while all this was happening. It moved from 29.3% at the close of the demerger to 29.9% during the dispute, and Bolloré did not need to buy a single additional share to get there.²

Vivendi ran a share buyback program in 2023 and 2024, and in March 2026 cancelled 7,000,000 of the shares it had bought, leaving the company holding 27,887,667 of its own shares, or 2.73% of its capital, against a total share count of 1,022,918,125.¹⁸ Every cancellation shrinks the denominator that every other shareholder's percentage is measured against. A buyback authorized by a board that answers, several rungs up a cascade, to the family whose stake sits a tenth of a point under a legal tripwire is not a neutral capital-return decision. It is a lever that moves the tripwire closer to the family without the family lifting a finger, funded by the company's own cash rather than the family's.

Founders authorize buybacks constantly, usually framed as returning excess cash to shareholders or supporting the stock price. Rarely does anyone in the room ask the other question: whose relative position does this transaction improve, once the shares actually leave the count. If a controlling holder sits close to any bright-line threshold in your jurisdiction, a corporate buyback is not company business happening near their stake. It is company business happening to their stake, priced in company money, decided by a board they appointed.

None of this requires a French conglomerate or a nine-figure legal bill to matter. A two-tier holding structure with a family vehicle on top produces the same shape at any size.

Name every rung and its percentage, out loud, in one document. If you cannot state in one sentence who owns what all the way from the ultimate family vehicle down to the operating company, you have already lost the ability to explain your own structure to a regulator, and they will not do the reconstruction generously.

Find the one rung that touches a bright-line rule, and treat it differently from every other rung. Ownership above 90% in a private holding company is unexamined because nothing forces an examination. Ownership at 29.9% of a listed operating company sits next to a number that changes what regulators are entitled to demand. Those are not the same kind of percentage, even though they are both percentages.

Learn your jurisdiction's actual control test, not the one you assume. A US founder assumes control is about raw ownership or board seats. France's L.233-3 asks a narrower, more mechanical question: whose votes actually decided the last two annual meetings.¹⁰ That test rewards low turnout among everyone else as much as it rewards a large stake, and a controller who wants the safety of a low reported percentage should want other shareholders to show up, not stay home.

Watch what a buyback does to the denominator. Any authorization to repurchase and cancel stock changes every other shareholder's percentage, including the controller's, quietly and by the same amount every time. If a controlling shareholder sits within a few points of a legal threshold, ask who approved the buyback and who benefits from the arithmetic before you approve the next one.

Decide in advance whether "control" for you means "safely under the trigger" or "genuinely uncontested." Bolloré's family is not remotely uncertain about who runs Compagnie de l'Odet or Bolloré SE. The only uncertainty engineered into this structure sits at the exact layer where uncertainty is worth billions of euros. If your own structure has a rung like that, know which one it is before a minority shareholder finds it for you.

Not under the French mandatory-bid rule, which attaches to crossing 30% of the capital or voting rights of a company whose shares are admitted to trading on a regulated market.² A private, unlisted holding company like Compagnie de l'Odet can be 93% family-owned indefinitely, because the trigger is a feature of the listed layer, not the ownership concentration itself. The regulatory attention in a cascade always lands on whichever rung is publicly traded, regardless of how concentrated the private rungs above it are.

Because the two groups wanted opposite outcomes from the same fact-finding. CIAM and the broader minority shareholder base stood to receive a cash exit at a court-supervised price if control was found, an outcome analysts valued at €6 billion to €9 billion in aggregate.² When the July 2026 ruling found no control, that expected payout disappeared, and the market repriced the stock down roughly 10% the same day.³ A minority-protection rule only pays out when it is triggered. A court correctly declining to trigger it is good news for the controller and bad news for the shareholders the rule was written to protect, which is an outcome the statute permits without resolving.

Every percentage in this structure is public. Sofibol's stake in Compagnie de l'Odet is in a regulatory filing. Compagnie de l'Odet's stake in Bolloré SE is on the company's own website. Bolloré SE's stake in Vivendi has been printed in nearly every article written about this dispute for two years. There is no hidden document here, no offshore trust, no jurisdiction nobody can see into. The facts were never the problem.

The problem, if you are a minority shareholder standing underneath a cascade like this, is that the law only picks a fight over the rung it was built to police, and the family that built the structure gets to choose which rung that is by choosing where to stop buying. Bolloré stopped at 29.9%. Whether that was foresight or luck, the effect is the same: everything above that number is uncontested because nobody is entitled to contest it, and the number itself has now survived a regulator, two appellate panels, and a supreme court, largely by being exactly one-tenth of a percentage point smaller than the number that would have ended the argument in CIAM's favor.

If you are building your own structure, ask which single number in your chain plays that role, because it is the only one anyone will ever measure you against.