The Wrong War

Earlier this month, in a summit room in Beijing, Xi Jinping opened his meeting with Donald Trump by invoking a 2,500-year-old Greek historian. Can China and the United States, Xi asked, overcome the Thucydides Trap? The reference was to Thucydides's observation that the growth of Athenian power, and the fear this inspired in Sparta, made war between them inevitable and has since appeared in every serious publication covering the summit. The idea is simple and brutal: when a rising power threatens an established one, structural fear pulls both toward conflict regardless of whether either wants it. The trap is the frame itself. The only way out is to recognise it before it closes.

The marketing industry press spent this week applying precisely this lens to the holding company competition. Publicis rising. WPP declining. Omnicom distracted. Will the established order fight back? Who wins the agency Peloponnesian War?

The problem is that it is the wrong application of the idea.

Because the Thucydides dynamic in this industry doesn't run between the holding companies at all. That is the wrong war.

Instead, it runs between the agency layer as a whole and the platform layer. Between the established order of marketing services and the genuinely rising powers of Meta, Google, and Amazon. That is the structural shift. That is where the fear should be directed. Most holding companies are so absorbed in their local rivalry that they haven't looked up long enough to see it.

But there is a historical parallel that is more instructive than Athens and Sparta, and it sits four centuries closer.

It is also, not coincidentally, French.

Early in the 17th Century, the Holy Roman Empire fractured along religious lines and Europe descended into what would become the Thirty Years War. Everyone understood the conflict in religious terms: Catholics versus Protestants, the Habsburgs defending the old faith, the Protestant princes of Germany fighting for reform. That was the frame. That was the war that everyone joined, funded, or fled.

Cardinal Richelieu, chief minister of Catholic France, looked at the same conflict and saw something completely different. Not a religious war. A power contest. The Habsburgs, controlling Spain, Austria, and significant territories on France's eastern and southern borders, were becoming too dominant.

The religious frame was the obvious frame and Richelieu recognised it as the wrong one.

So Catholic France agreed to fund the Protestant Swedish army with one million livres annually to maintain forces in Germany against the Catholic Habsburgs. A Cardinal of the Roman Catholic Church, directing French gold to Protestant soldiers, to fight a Catholic emperor. In the accepted frame of the conflict, it was incomprehensible. In Richelieu's frame of pure European power it was the only logical move. Germany consumed itself over three decades. France played the contest nobody else recognised was being fought, and emerged as the dominant power in Europe.

Richelieu didn't win the Thirty Years War. But he recognised it was the wrong war, invested in the right one alongside it, and won that instead.

The French thread is not coincidental. And in 2019, Sadoun made a move that reads almost precisely from Richelieu's playbook. While every other major holding company was fighting the agency war of market share, creative awards, media billings and merger scale Publicis paid $4.4 billion for Epsilon. Not an agency. A data company. The logic, stated explicitly on record, was to enable clients "to take back control on their data from the walled garden, by shifting from cookies to identity." Not from WPP. Not from Omnicom. From the walled garden. The named opponent was the platform layer. The obvious fight was not the fight Sadoun was fighting.

A week ago, in writing "The Neutrality Gambit", I covered what the LiveRamp acquisition meant for the infrastructure gap between Publicis and its holdco peers with the cash-gap calculation, the WPP runway problem, the concrete divergence in capital allocation. That analysis holds.

But the week of press coverage that followed did exactly what Richelieu would have recognised: it covered the religious war while missing the power contest running alongside it.

Google and Meta combined earn approximately twenty-eight times the net earnings of the top five holding companies, every year. Meta grew revenue 24% last quarter while WPP declined 6.7% like-for-like. No holding company, including Publicis, is in the same financial universe as the platforms.

Sadoun said it plainly on the recent investor call: his capital allocation strategy is "the polar opposite strategy of our peers." He is not measuring against peers. He identified a different opponent, a long time ago, and has been building against that opponent ever since.

The only way that an agency layer player can survive the decade is to justify their existence in a world where platforms work to disintermediate them entirely. There are three ways to do this. These are the three new structural moats that Publicis is showing us how to build, while they are not even looking at WPP and Omincom.

1. Neutrality. Meta's first-party data exists to sell Meta inventory. Google's intent data exists to sell Google search and YouTube. Their data advantage and their media ownership are inseparable, but one exists to optimise the other. An infrastructure layer with no owned inventory, connecting 25,000 publisher domains and 500+ technology partners across 14 markets, serves a brand's interests differently. It can run data collaboration across Meta AND Google AND Amazon AND retail media simultaneously. The platforms cannot do this for you, because they are the platforms. Their agnosticism ends at the garden wall.

2. Creativity and the upper funnel. Platforms are genuinely outstanding at extracting value from existing demand. Meta Advantage+ and Google Performance Max find the person already looking for your product and close the sale efficiently. What they cannot do well is build the brand equity that makes someone start looking in the first place. The platform AI optimises the last mile. It has no mechanism for the journey that makes the last mile worth anything. Brands spending exclusively in the lower funnel are harvesting equity they are not replenishing. The combination of creative network and deterministic identity is the only answer that addresses both problems at once.

3. Data ownership, and this is where the AI dimension becomes concrete, and where the real stakes of the contest become clear. Every company in every sector is currently building AI agents. Agents that identify customers, personalise journeys, coordinate workflows, optimise decisions. The problem is that agents are only as intelligent as the data they are trained on. Most companies are building agents on internal data with historical CRM records, past transaction logs, website behaviour from last quarter. These are data of the past. They report on what happened. They cannot anticipate what is about to.

The platform answer to this problem is straightforward: send us your data, build your agents inside our environment, and we will enrich them with our signals at population scale. Meta has 3 billion users' behavioural data. Google has 8.5 billion daily search interactions. Agents built inside their infrastructure are smarter in the short term because they are drawing on the platform's intelligence, not their own. But that intelligence stays with the platform. Your competitive advantage is borrowed. The moment a competitor makes the same arrangement, the advantage evaporates. You have not built proprietary intelligence. You have rented someone else's.

Publicis's Chief Strategy Officer Carla Serrano made the claim that 93% of companies currently lack the data foundation required to build effective AI agents. An estimated $1 trillion in AI investment is returning only 5% in meaningful value. The problem is not the AI. It is the data.

LiveRamp's clean room architecture is the structural alternative to the platform bargain: a brand connects its CRM data with a retail partner's purchase signals, a publisher's audience data, a supply chain partner's intelligence without any party's raw records being exposed to any other. The resulting combined dataset is unique.

An agent trained on it has proprietary intelligence that a competitor cannot simply replicate, because the competitor does not have access to the same combination of signals. That is competitive advantage that compounds over time.

Platform dependency is its opposite: efficiency that looks attractive today and creates structural vulnerability over years.

The practical question is not which holding company is best positioned against its peers. It is whose interests are structurally aligned with yours when you are trying to build a brand and the AI agents you will need that do not become dependent on infrastructure you do not own.

Platforms are not adversaries. Their tools are frequently the most effective available for specific tasks. But their business model requires them to optimise your data for their inventory. Every customer signal that flows exclusively through a platform enriches the platform's competitive position relative to yours. The efficiency gain is real and immediate. So is the dependency being built alongside it, over years.

The infrastructure that allows you to build AI agents on proprietary co-created data and to activate across all platforms without depending on any single one, is what this acquisition sequence has been constructing for seven years.

One question from the coverage deserves a direct answer: will Publicis actually honour its neutrality commitments regarding LiveRamp? Almost certainly yes, and not because of governance promises in a press release, but because compromising neutrality would destroy the very advantage the strategy was built to create. Publicis cannot function as the platform-agnostic infrastructure layer that competes with Meta and Google if it plays favourites at the data collaboration layer. The neutrality is self-enforcing because the right war requires it. Neutrality concerns are from the wrong war.

Richelieu died in 1642, three years before the Peace of Westphalia confirmed France's dominance. He did not live to see the full vindication of the strategy. But the strategy was correct. The contest he identified was the right one.

The wrong opponents to name are WPP and Omnicom. The right opponents are the companies whose combined earnings dwarf the marketing services layer's by nearly forty times, grow faster every quarter, and are currently building the argument that your AI agents should run on their data, not yours.

Publicis appears to be the only holding company that read the same history Richelieu read, and acted accordingly.

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry's table: senior client, global agency leader, technology founder. First published 25 May 2026 in the Brandflow newsletter on LinkedIn.