McDonald's is selling third-party advertising on 450 of its drive-thru boards. It operates those 450; franchisees paid for the other thirteen thousand.

The screen you order from at a McDonald's drive-thru in the United States may now show you an advertisement for something McDonald's does not sell. It has done so since August, at 450 restaurants. The company did not say so publicly until 23 September.

That is the date of its investor day in Chicago. The release McDonald's filed with the Securities and Exchange Commission that morning runs to three pages of targets to 2030 — a percentage point and a half of chicken share, the same again in beverages, operating margin in the low-to-mid 50% range, about 250 basis points of restaurant-level efficiency, about $8.5 billion of partnering support to franchisees through 2036. It does not contain the word advertising.

The McDonald's Media Network was described from the stage instead. Morgan Flatley, the global chief marketing officer, said the pilot had begun the previous month across 450 US company-owned restaurants, and that the aspiration is to build it into a billion-dollar business. She called it revenue with, in her words, "no disruption to our customer experience". We read a third-party transcript of the day rather than the company's own, because there is no company transcript we could find, and the filing carries none of it.

CNBC, which reported the announcement that afternoon, says the placement is the digital drive-thru order boards and that the pilot has not reached the franchisees who operate the rest of roughly 14,000 US restaurants. Trade coverage says the advertisement appears once the order is complete. We could not establish that from the company and will not print it as fact. It matters less than it sounds. Either way a customer who came for a burger is queueing in front of an advertisement for something else, in a lane with a painted kerb on both sides and no gap to pull out through — and that last fact is the only property of the inventory nobody disputes.

Take the company's version first, because it is the likely one

A retail media network is not a quarter's work. Amazon and Walmart built theirs from strength rather than weakness, and Amazon reported $68.6 billion of advertising service sales in 2025 on CNBC's figures. Flatley put the US commerce media market above $100 billion by 2028. On that framing this is a company with 45,356 restaurants noticing late that it owns a great many screens, which is a question any board would ask.

The framing also survives the obvious objection. None of it is a response to one soft quarter, because none of it could be. A pilot running in August was commissioned long before the August results.

So the honest version of the sceptical reading is not that McDonald's is reacting. It is that a business builds this when it has concluded that each visit needs to be worth more, and that the levers it has been using to make each visit worth more have stopped working.

A segment went backwards for a year, and it was nothing to do with the food

Comparable sales exclude currency and hyperinflationary markets, which makes them the right series for a question about demand and the wrong one for a question about revenue. Read the three segments together across twelve quarters and the shape is not subtle.

The same figures, as read.

Quarter	US	International operated	Developmental licensed
Q3 23	+8.1%	+8.3%	+10.5%
Q4 23	+4.3%	+4.4%	+0.7%
Q1 24	+2.5%	+2.7%	-0.2%
Q2 24	-0.7%	-1.1%	-1.3%
Q3 24	+0.3%	-2.1%	-3.5%
Q4 24	-1.4%	+0.1%	+4.1%
Q1 25	-3.6%	-1%	+3.5%
Q2 25	+2.5%	+4%	+5.6%
Q3 25	+2.4%	+4.3%	+4.7%
Q4 25	+6.8%	+5.2%	+4.5%
Q1 26	+3.9%	+3.9%	+3.4%
Q2 26	+0.8%	+1.5%	+1.9%

The developmental licensed segment is the one that carries the Middle East, among more than seventy-five countries. In the fourth quarter of 2023 it grew 0.7% while the US grew 4.3% and the international operated markets 4.4%. Three quarters later it was down 3.5%. The 2024 annual report puts the full-year decline of 0.3% down to the continued impact of the war in the Middle East and to negative comparable sales in China. The word boycott appears once in that document, in a generic risk factor about economic action, and is attached to no result. The company's own word for what happened is war.Worth being precise about what the filings do and do not say. They name a cause for a segment. They do not quantify it, separate it from China, or name a country. A reader who wants the Middle East on its own will not find it in an SEC filing.

That is the demonstration. For four consecutive quarters, a segment covering a quarter of the world went sideways and then backwards for reasons that had nothing to do with the price, the product or the service, and the company said so in writing. No menu decision would have fixed it. Neither would 250 basis points of restaurant-level efficiency.

The case against reading it that way

State it at full strength, because it is strong.

The effect faded. The developmental licensed segment turned positive in the fourth quarter of 2024, up 4.1%, and has grown in every quarter since; for the full year 2025 it grew 4.6%, the best of the three segments. If exposure to a shock the company cannot control were the motive, the motive weakened two years ago.

The recent weakness is also somewhere else entirely, and management has explained it operationally. In the second quarter of 2026 the US grew 0.8% with guest counts falling. On the August call, executives said restaurant teams were overwhelmed by too many deployments, that service times rose and satisfaction scores fell, that value execution accounted for about two-thirds of the traffic underperformance, that only 60 to 65% of the system was running the recommended pricing architecture, and that April lapped the previous year's Minecraft campaign. A new president of McDonald's USA was announced in the same release. None of that is a company saying its demand is structurally exposed. It is a company saying it ran the quarter badly.

And a media network is simply a normal thing for a large retailer to build in 2026. Reaching for a geopolitical explanation when a commercial one is sitting on the table is the sort of move this publication is meant to be suspicious of.

The chart does not tell you what the board concluded. It tells you what there was to conclude something from.

The third rent on the same square metre

Here is the reading, and it is ours. McDonald's has not said it and will not.

The company already makes money from a restaurant three ways over. It takes a royalty on a percentage of the franchisee's sales. It generally owns or holds a long lease on the land and the building, and takes rent. And it sells food in the restaurants it operates itself. Two of those three are a claim on somebody else's turnover, which is to say they move when demand moves.

Advertising is the first line that does not. What an advertiser pays for a slot on an order board is a function of how many cars pass it, not of what those cars buy, what beef costs, or whether the pricing architecture has been deployed correctly in that market. It is the same square metre, let a third time, to a tenant who does not care what is on the tray.

That is not independence from demand, and we should not pretend it is. Empty the lane and the slot is worth nothing. But it is independence from every lever that has actually gone wrong in the last three years: the menu, the price, the ticket time, and a war. A board looking at the chart above would be negligent not to want a revenue line with that property, whether or not anyone in the room used the word exposure.

The scale is also worth holding lightly. A billion dollars was described as an aspiration across the McDonald's system, which is not the same number as a billion of the company's own revenue, and nobody said which. For comparison, the only line in the accounts that holds money from neither restaurant sales nor rent nor royalties — Other revenues, being technology fees from franchisees and brand licensing — came to $647 million in 2025.

The screen has an owner, and it is usually not McDonald's

This is why the pilot is company-operated restaurants only, and it is not a technology problem.

The 10-K is explicit about the conventional franchise arrangement: the company generally owns or secures a long-term lease on the land and the building, and the franchisee pays for "equipment, signs, seating and décor". Approximately 95% of the 45,356 restaurants are franchised. In most of the estate, the operator bought the screen.

So every step beyond 450 restaurants is a negotiation with somebody who owns the inventory, has paid to maintain it, is being asked to accept about $8.5 billion of partnering support for a modernisation programme at the same time, and will reasonably ask which side of the table banks the advertising revenue. Flatley's "no operational complexity" is true of the 450. It is a claim about the other thirteen thousand that nobody has tested, because nobody has asked them yet in public.

What would have to be true before any of this reached the EU

Nothing has been announced for Europe. What has been announced is the plumbing: a single global mobile application platform, piloting in France later this year and scaling across the ten largest owned markets by the end of 2028. Targeting across kiosks, boards and an app implies loyalty accounts, order history and device identifiers. No one at the investor day connected the data foundation to the advertising business, and we are not connecting it for them. But the questions a European deployment would have to answer are answerable now, and three of them bite.

The basis for the profiling. Under the GDPR, selecting an advertisement from a person's order history is profiling within Article 4(4) — automated processing used to evaluate personal aspects including preferences, interests, behaviour and location. It needs a lawful basis. Consent under Article 6(1)(a) means nothing happens until somebody says yes. Legitimate interests under Article 6(1)(f) means it starts and the person may object. Which one is chosen will not be visible in the product, and Article 6(1)(f) ends with a clause that will matter in a restaurant: the interests of the controller give way where the data subject is a child.

The device, which is a separate question. Article 5(3) of the ePrivacy Directive requires consent to store information on, or read information already stored in, a user's terminal equipment, after clear and comprehensive information. It does not care whether the information is personal, and it is not satisfied by whatever basis was chosen for the GDPR question. An identifier read from a phone to pick an advertisement is a consent question on its own terms, in twenty-seven differently transposed versions.

The right to object, and where it is kept. Article 21(2) makes objection to direct marketing absolute — no balancing, no compelling grounds, and it covers the related profiling. Article 21(3) says the processing then stops. Article 21(4) is the one to read twice: the right must be brought to the person's attention at the latest at the time of the first communication, clearly and separately from any other information. An opt-out at the foot of a privacy notice is not that.

Then the two instruments everyone will reach for, which do not reach back. The Audiovisual Media Services Directive protects minors from commercial communications in Article 9(1)(g), but Article 9 binds media service providers, and an audiovisual commercial communication is defined as images accompanying or included in a programme or a user-generated video. A drive-thru order board is not a programme. Article 9(4) on food high in fat, salt and sugar does no more than have Member States encourage codes of conduct, and only for children's programming. The Digital Services Act forbids advertising based on profiling to a person the provider reasonably believes is a minor, in Article 28(2) — but it binds providers of online platforms, and an online platform under Article 3(i) is a hosting service that stores and disseminates information to the public. A restaurant ordering app does not. We found no guidance or case law on in-restaurant screens either way.

Which leaves the children question where nobody has organised it: in national advertising rules that differ by Member State, in a room that reliably contains children, in front of a screen that is not broadcast, not an online platform, and not going anywhere. We have not opened those national rules and make no claim about them.

The company's own risk factors say that as it expands digital engagement, data collection and personalisation through AI it faces new and heightened risks, and name the GDPR, US state law and the EU AI Act. They do not name ePrivacy. They do not name the Digital Services Act. Nor, as of the 2025 annual report filed in February, do they name a media network at all.

The third-quarter release is due late October. The document that would actually settle this is the first agreement with a franchisee to sell time on a screen the franchisee paid for, and that one is not filed anywhere.