Volkswagen is not the story. Germany's cost structure is.

On 4 September 2026 Volkswagen's supervisory board unanimously approved a plan to remove around 100,000 jobs from the group by the end of the decade, roughly 15% of its worldwide workforce.

It is being read as a Volkswagen story. It is not. It is the clearest reading yet of a cost structure the German car industry built while one market was paying for it, and that market has stopped paying. Volkswagen is the flagship carrying the wreckage, not the wreck.

Two things have to be separated before any of that can be argued. The job numbers in circulation do not measure the same thing, and the cause most often named is not the cause.

13.5 against 29.0: productivity, not China, explains the headcount

Start with the numerator and the denominator, and make them match. Most of the productivity figures quoted in this debate fail on that alone.

The Volkswagen Group delivered 8,983,900 vehicles worldwide in 2025 and employed 662,942 people at 31 December 2025, both figures including the Chinese joint ventures. That is 13.5 vehicles per employee.

Toyota sold 11,322,575 vehicles in calendar 2025, including Daihatsu and Hino, and employed 390,927 people on a consolidated basis at 31 March 2026. That is 29.0 vehicles per employee. Toyota sold 2.3 million more cars than Volkswagen with roughly 272,000 fewer people on the payroll.

These are audited full-year figures, not quarterly ones, which is the difference between a structure and a bad period.

One warning about the denominator, and it is not pedantry. Volkswagen publishes a second headcount, 602,659, which excludes the Chinese joint ventures. Divide the same deliveries by that and you get 14.9, because you have credited Volkswagen with cars built by people you just removed from the payroll. Anyone who hands you a productivity number for a carmaker should be able to say which of the two they used. The same discipline decides the rest of this article.

A competitive shock does not create a cost structure. It makes it visible.

The China explanation is seductive because it is partly true and entirely external. Group deliveries in China fell 8.0% in 2025, to 2.69 million. The group operating result halved, from €19.1bn to €8.9bn, and the operating margin fell from 5.9% to 2.8%, hit by US tariffs, the reset of Porsche's product strategy, currency and price effects. AlixPartners puts the Chinese cost advantage at around 30% per vehicle, with development cycles less than half as long and 40% to 50% less investment per model.

All of that is real. None of it explains why the organisation was that size in the first place.

For most of its modern history Volkswagen was not run purely as a business. Lower Saxony holds a blocking stake. Labour holds half the supervisory board under co-determination. The VW Law was written to keep it that way, and the arrangement did exactly what it was designed to do: it turned Volkswagen into a machine for producing stable, well-paid German industrial jobs. That is a defensible social objective and a catastrophic capital allocation discipline. When headcount is treated as an output rather than an input, the organisation slowly fills with process, redundancy and complexity that nobody is paid to remove.

This is not a peculiarity of German industrial policy. It is the same mechanism that emptied the Italian dealer network of its owners one layer down: the fat margin was the slack, and the slack is what made a badly run business and a well run one look alike in the accounts. In Volkswagen the Chinese market was the slack. Remove it and the difference is the entire result.

The December 2024 agreement removes 35,000 jobs and no plant. The 100,000 is a different number.

Three figures are in circulation. They have three different perimeters and three different degrees of certainty, and stacking them as though they were chapters of one number is how a board ends up briefed wrongly.

More than 35,000. The VW brand, German sites, by 2030. Signed in December 2024 with IG Metall and the works council as the "Zukunft Volkswagen" agreement, alongside 734,000 units of German production capacity removed, collectively agreed labour costs down €1.5bn a year, savings of over €4bn a year from that round and over €15bn a year in the medium term. This one is a contract.

Around 50,000. What Volkswagen counted as already agreed across the group before September 2026. The 35,000 is its largest block, with Audi's and Cariad's programmes inside it.

100,000. The group, worldwide, by the end of the decade. First reported by Manager Magazin on 26 June 2026, when Volkswagen would not comment on "internal, confidential documents". Still unconfirmed on 10 July, when management announced a model range cut without attaching a headcount figure. Approved by the supervisory board on 4 September 2026, as roughly 50,000 on top of the 50,000 already agreed.

Note what happened over those ten weeks. The same figure travelled from a press report the company refused to discuss to an approved plan. Anyone who wrote it down as a fact in June was right by accident, and would have been wrong on the perimeter anyway: the 35,000 is a German brand number, the 100,000 is a worldwide group number, and neither is a version of the other.

What the December agreement did not contain was a single plant closure, and nor does September's. Management's opening position in 2024 was to shut at least three German sites. In September, union and supervisory board representatives stated that no closure had been approved, while conceding that four German plants, Hannover, Emden, Zwickau and Neckarsulm, have no secured production plan for the next decade.

"No closure" is carrying a great deal of weight in that sentence. On 16 December 2025 the last vehicle left Dresden, a red ID.3 GTX, after 6,200 cars that year. It was the first time Volkswagen had ended vehicle production at a German plant. The site was not closed: from 2026 it becomes an innovation campus with TU Dresden, working on artificial intelligence, robotics and chip design.

That is the German playbook running exactly as designed. Production stops, the address stays, and the write-down arrives in instalments. It is also not only German: an Italian plant running at a fraction of its capacity is being handled with the same instinct, one floor up.

If this were a Volkswagen problem, the rest of Germany would be fine

It is not fine.

Audi is cutting up to 7,500 German jobs by 2029, 6,000 of them by 2027, concentrated in administration, sales, planning and development rather than on the line. It stopped building cars in Brussels on 28 February 2025, a plant of around 3,000 people, on weak demand for the Q8 e-tron. Porsche, the group's profit engine for two decades, reported a 2025 group operating result of €413m against €5.64bn the year before, an operating return on sales of 1.1% against 14.1%, after nearly €3.9bn of one-off charges for restructuring, a product strategy reset, battery activities and US tariffs. Cariad, the software unit, shed up to 2,000 staff, and Volkswagen's answer to the software it could not build was a joint venture with Rivian worth up to $5.8bn.

Move outside the group and the pattern holds. ZF Friedrichshafen is cutting 11,000 to 14,000 German jobs by the end of 2028, out of around 54,000, concentrated in electrified powertrain. Continental did not restructure its automotive division at all: it spun it off as Aumovio, listed in Frankfurt on 18 September 2025 with around 82,000 employees and about €18.5bn of 2025 sales. That is a group deciding the business is worth more outside its own balance sheet than inside it.

This is not a company in trouble. It is an ecosystem repricing itself, and the decisions being taken inside it are ownership decisions, not cost decisions.

Which is where this stops being German news. A supplier reading that ZF is removing a quarter of its German workforce, or that Continental has put its automotive business into a separate listed company, does not have a strategy question. It has one question about its own exposure, and the answer is a piece of work rather than an opinion: which of our programmes sit on platforms that survive, and what is our position worth to the people consolidating around us. That is the shape of a commissioned piece of market research, which can be asked for on its own, on a single question, without a transaction behind it.

Zwickau says the capacity was sized to a political deadline, not to demand

If you want the single most damning detail in the whole file, it is not the size of the cut. It is the address.

Zwickau was converted at enormous cost into Volkswagen's all-electric flagship, six EV models across VW, Audi and Cupra, the showpiece of the group's electric transition. It built around 204,000 vehicles in 2024. It has since run through repeated production pauses and shift reductions on the ID.3 and the Cupra Born, and it now sits on the list of four German plants with no secured production plan for the next decade. From 2026 it takes on a new role as the group's competence centre for the circular economy.

Zwickau was not beaten by BYD. It was sized for a demand curve that policy promised and consumers declined to deliver, which makes it a forecasting failure rather than a competitive one. You can mandate supply. You cannot mandate demand. The difference between the two is now being paid in Saxony, in jobs.

That gap between a mandated volume and a real one is the same gap that shows up three years later in the used market, where the residual value assumption written into a leasing contract decides who absorbs it. Getting a demand curve wrong is not a modelling embarrassment. It is the single most expensive decision a manufacturer makes, because every plant, supplier contract and residual value guarantee downstream is sized off it, which is why we spend most of an automotive strategy mandate on the assumption rather than on the plan built over it.

A tariff wall is a diagnosis, not a cure

Brussels' answer to Chinese competition was duty. Since 30 October 2024 the EU has applied definitive countervailing duties on battery electric vehicles built in China, for five years, ranging from 7.8% for Tesla Shanghai to 17.0% for BYD, 18.8% for Geely, 20.7% for other cooperating producers and 35.3% for SAIC and for non-cooperating producers.

That last figure is worth stating precisely, because it is usually not. The 35.3% sits on top of the standard 10% import duty that every imported car pays. The two are different instruments and they are worth keeping apart: one is a targeted anti-subsidy measure aimed at named manufacturers, the other is the ordinary tariff every imported car carries.

Either way it is a tell, not a solution. You do not need a duty wall to protect an industry that is winning, and the wall is already being negotiated into something softer: in January 2026 the Commission published guidance on how Chinese producers can offer price undertakings in place of the duty. The first undertaking accepted, in February, went to Volkswagen (Anhui) for the CUPRA Tavascan, a group car built in China, and the carmakers that announced a European plant to get over the duty had already redone the arithmetic before that. Meanwhile the 2035 combustion ban has acquired an e-fuels carve-out and CO2 fleet targets have been smoothed across multiple years to avoid fines nobody had budgeted for. The policy apparatus committed to a timeline it could not execute and is now managing its retreat at the same negotiated pace as Volkswagen manages its job cuts.

The test has an answer already, and it is not the comfortable one

In September we put a test on the table so the thesis could be judged rather than believed. If the VW brand margin recovered toward 6.5% by 2026 and China stabilised, call it a cyclical shock and move on. If instead there were further cuts, a slipped margin target, or guidance reversals at Porsche and Audi, the structural reading holds.

Scored on what is now published, it is not close.

The margin target slipped first, and earlier than the test assumed. The 6.5%-by-2026 goal for the VW brand was already being described to investors as 6% "in the medium term" in January 2025. The margin itself has not moved: the brand turned €2.6bn of operating result on €86.6bn of revenue in 2025, about 3%, and 0.4% in the first quarter of 2026, or 3.5% before restructuring and ID.4 costs. The further cuts arrived on 4 September. The guidance reversals arrived at Porsche, whose return on sales went from 14.1% to 1.1% in a single year with 2026 guided at 5.5% to 7.5%, and at Audi, which cut its profitability forecast again. China did not stabilise.

Four of the conditions were set. Four have been met. On the author's own test, this is structural.

For anyone supplying this industry, the operative number is not the 100,000. It is the model range, which Volkswagen said in July 2026 it intends to cut by up to half. Halving a range does not halve volume. It concentrates it, and it does so by decision rather than by cycle, which means the part you supply either survives the cut or it does not, and the answer exists inside your customer's planning today rather than in next year's registrations.

That changes the question in front of a supplier. It is not whether to wait out a downturn, because this is not one. It is whether the business is worth more on its own or inside the balance sheet of whoever is consolidating around it, and that is a question with a closing date attached. Continental answered it in September 2025. ZF is answering it now. The suppliers who answer it last will answer it at somebody else's price.