China is not flooding Europe. It is filtering it.

Chinese manufacturers have announced ten new production sites in the European Union since September 2023, on Transport and Environment's count. For anyone who has to supply one of them, underwrite the cars it will build, or compete with it from an existing plant, the useful question is narrower than the map: which of those announcements turns into an order for steel, and when.

The answer does not follow the map. BYD has shelved a billion dollar plant in Turkey and is installing equipment in Hungary. Its decision on a second European site has been described as imminent since late 2025 and is still open. Chery took two years of slipped dates to get a former Nissan plant near Barcelona building cars. Same company in two of those cases, the same strategic pressure in all of them, and different outcomes inside the same six months.

What separates them is not China. It is the jurisdiction on the other side of the table. And the reason that has become the deciding variable is a change in Brussels in January 2026 that was reported, almost everywhere, as the opposite of what it was.

The wall never had a door. It had an application form.

At the end of October 2024 the European Union imposed definitive countervailing duties on battery electric vehicles built in China, following a formal anti-subsidy investigation, topping out at 35.3%. The full scale, the named manufacturers behind each rate, and the ordinary import duty all of it sits on top of, we set out separately. What matters here is what happened to that wall fourteen months later.

On 12 January 2026 the European Commission published a guidance document on price undertakings. In its own words it "provides Chinese exporters of BEVs to the EU with general guidance on the submission of price undertaking offers", and it "covers various aspects to be addressed in a possible undertaking offer, including the minimum import price, sales channels, cross-compensation, and future investments in the EU".

Read that for what it is not. It is not a repeal. It is not a negotiated price floor switched on across the market in place of the duty. It is a procedure for asking. The duties imposed at the end of October 2024 were in force on the day the guidance was published, they are in force now, and they apply to every exporter that has not been through the procedure and been accepted.

The gap between those two readings is not a matter of wording. It is the entire planning assumption underneath a European pricing plan. A repeal changes the landed cost of every China-built electric car in Europe at once, and nobody has to commit to anything to collect the benefit. A procedure changes the landed cost of one exporter's cars, for the models named in one offer, after the Commission has examined that offer and accepted it, and only for as long as the exporter keeps its side of the bargain. The first is a market event. The second is a queue, and queues sort people.

The first through was not a Chinese brand. It was a Volkswagen built in China.

On 10 February 2026 the Commission accepted the first price undertaking under that procedure. The exporter was Volkswagen (Anhui) Automotive. The product was the CUPRA Tavascan, a Spanish-badged electric SUV built in China. Volkswagen (Anhui) can export the Tavascan into the Union without the duty. Every other Chinese exporter still pays it.

That one fact carries more information than most of the forecasting written around it. The first company through a mechanism designed for Chinese exporters was a German group's Chinese operation, exporting a Spanish brand.

It is not an accident of timing. The mechanism does not reward being Chinese, and it does not punish it either. It rewards being able to file. The guidance asks for a minimum import price the exporter can hold and police through its own sales channel, a group structure that can be examined for cross-compensation, and a credible answer in the box marked future investments in the EU. A Chinese brand three years into Europe, selling through an importer network it does not own, has none of those ready. A European group that has built cars in China for three decades has all of them, plus the legal and compliance apparatus to assemble the file and live with the audit afterwards.

So the shape of the opening is now known, and it is not the shape the January headlines drew. It is granted per exporter and per model, not per country and not per brand. It is granted to whoever can document, not to whoever can price. And on current evidence the applicants best equipped to use it are European groups with Chinese production, which is a very different competitive picture from the one where the duty simply disappears.

The price of the exemption is written down, and it is called investing here

The Commission also published what Volkswagen (Anhui) gave up for it. Three commitments: to sell above a minimum import price, to limit its import volumes, and to invest "in significant BEV-related projects in the EU with clearly defined milestones". Failure to comply, the Commission says, "may lead to withdrawal of the undertaking by the Commission and a retroactive reinstatement of duties".

Put those three side by side and the incentive runs the opposite way to the one the commentary assumed. The reading that spread in January was that a price floor makes a European plant pointless: if you can price above the line and skip the duty, why pour the concrete? That reading only survives if the exemption is a pricing decision. It is not. It is a pricing decision, plus a cap on units, plus a capital commitment with dates on it and a clawback behind it.

The volume cap alone kills the arbitrage. An exporter that prices above a floor and then sells as many cars as it can has traded a duty for a margin, which is a good trade. An exporter that prices above the floor and is capped on units has traded a duty for a ceiling on its European business, and the only way through a volume ceiling is production inside the Union. The investment milestones then write that conclusion into the contract rather than leaving it to be inferred, and the retroactive reinstatement means the promise is not free to break later.

This is what a filter looks like when it is drafted rather than described. You do not step over the wall with a factory. You promise the factory, in writing, with dates, and you lose the exemption backwards if you do not build it. Anyone modelling a China-built model's European margin off a price floor alone is modelling a third of the instrument.

Turkey stopped, Spain still open, Szeged installing: one company, three outcomes in six months

BYD is the cleanest test available, because the same capital allocation committee produced three different answers in three jurisdictions inside the same year.

Turkey stopped. In 2024 BYD signed with Turkey's industry ministry for a plant reported at around one billion dollars, 150,000 units of annual capacity and 5,000 jobs. Construction never started. In June 2026 BYD executive vice president Stella Li told Reuters that Hungary is "the number one priority right now" and that there is no timeline for production in Turkey. Ankara then suspended BYD's import tax exemptions and warned that incentives already taken may have to be repaid if the investment does not happen. In parliament, an opposition deputy asked the industry minister how much the company had gained from those exemptions while importing 26,610 vehicles over the last six months of 2025 and the first four months of 2026, and whether any guarantees had ever been secured. That figure comes from a parliamentary question rather than from a published customs series, which is worth saying plainly, but the shape of it is not in dispute: the cars arrived, the plant did not.

Hungary is being installed. Szeged started trial production in the first quarter of 2026. Mass production, originally targeted for late 2025, is now expected in the fourth quarter of 2026. Full capacity is put at up to 300,000 units a year, and the investment has been reported as high as four billion euros, a figure that has always travelled with that qualifier and should keep it. The site is a year behind its own plan. It is also being built.

Spain is still open. In July 2026 Alfredo Altavilla, BYD's special adviser for Europe, said the decision on a second European site needs to be made "very soon", with Spain and France the candidates and a brownfield acquisition the favoured route. Spain had already been reported as the frontrunner in late 2025. Nine months on, the call has not been made.

Three outcomes, one company, one continent. If Chinese onshoring were a wave, the three would look alike. They do not look alike at all.

The filter is not China. It is the jurisdiction that turns an announcement into a building site

Compare what Turkey and Hungary were actually offering when the money had to move.

Hungary offered membership of the Union and functioning customs status, so a car built there is an EU car with no duty question attached to it, and a government that had courted the investment rather than using it as leverage in a domestic argument. Turkey offered a customs union with the Union, lower costs than western Europe, and an incentive package that turned out to be revisable. When the plant slipped, the incentives were pulled. That is a rational response by Ankara and a decisive one for the investor, because it reprices every future commitment made in that jurisdiction, not just this one. Turkey has more leverage over trade terms than Hungary does. It has less leverage over where the plant gets built, and the two are not the same lever.

The second Spanish data point says the same thing from the other side. Chery's joint venture with Ebro at the former Nissan site near Barcelona was announced in 2024, slipped repeatedly, and is only now building Omoda 5 cars towards a target of around 50,000 units a year by the end of 2026. Spain kept the project alive through the delays. The delays were real and the project is still real, which is precisely the distinction a supplier needs and a map cannot show.

Then there is the detail that says most about how these decisions are now being timed. Altavilla put the urgency of BYD's second site not on the duty but on the proposed European content rules, which Europe's largest carmakers have themselves asked Brussels to shape, a request we take apart alongside an idle Italian plant and the partners circling it. The capital is being timed to a rule that has not been written yet. That tells you the duty was never the binding constraint. Regulatory certainty was, and still is.

The wall does not cover plug-in hybrids, and that is where the Chinese brands walked through

The measure Brussels adopted applies to battery electric vehicles built in China. Plug-in hybrids are not in it. Chinese manufacturers noticed immediately, and the registrations show what they did about it.

Between January and May 2026 the five largest Chinese groups registered 619,353 cars across the European Union, EFTA and the United Kingdom, around 10.6% of the market, on Dataforce figures reported by Bloomberg in June 2026. The perimeter matters and is usually left out: that is a count by ownership rather than by badge, and it includes the United Kingdom, which never adopted the extra duty and is now the strongest market in Europe for Chinese electric cars. Counted by badge inside the Union alone the share is materially smaller, which is why two honest analyses of the same market can be six points apart.

On the brand level the crossing already happened. BYD registered 174,144 vehicles in Europe in the first half of 2026 against Tesla's 170,351, a lead built from a deficit of around 39,000 units a year earlier.

And the duty itself did something measurable, just not the thing it was sold on. Transport and Environment's July 2026 analysis found that China-built battery electric cars fell from 22% of the EU market in 2024 to 17% in the first quarter of 2026, and that the group facing the top rate almost halved its China-built exports between 2023 and 2025 while a group facing a rate less than half as high more than doubled them. The wall sorted Chinese exporters by rate. It did not stop them. Meanwhile the CEPR work on the same period found that consumer prices for Chinese electric cars in Europe did not rise and in many cases fell, that the effect on Chinese import share was small and not statistically significant, and that the clearest measurable result was fiscal, with tariff revenue running at roughly half a billion euros a quarter.

A measure whose most reliable output is customs revenue is a diagnosis of a competitiveness problem, not a treatment for one.

The people building for real are not doing it for the duty, and that is the part that does not slip

The most quoted number from the T&E work is the fall in China-built electric cars from 22% to 17% of the EU market. The most useful one is who caused it. The drop was driven mainly by Western manufacturers moving production out of China and into Europe. Tesla's share of China-built imports fell from 26% to 19%, and European manufacturers' share of those imports collapsed from 38% to 23%. Imports of Chinese-branded cars kept rising throughout.

So the repatriation that actually happened was European, and the Chinese localisation that actually happened was not driven by the duty either. Geely has built cars in Europe for more than a decade through Volvo, Polestar, Lotus, LEVC and Lynk & Co: European plants, European engineering, European service networks, with a Chinese parent. On 23 July 2026 Geely and Ford announced a joint venture at Valencia, targeting production in the first half of 2027 with the Ford Kuga, then the Bronco and a jointly developed crossover in 2028, and two Geely electrified models from 2028. That is not a company trying to get over a wall. That is a company buying its way into a manufacturing base, a supplier network and a service footprint it will still want in 2035, which is exactly what it did when it bought Volvo. It is also the exception rather than the pattern: the Chinese arrivals in Europe have overwhelmingly built and partnered rather than acquired.

Which is the point a fleet or leasing desk should take from all of this. A plant inside the Union is not tariff arbitrage. It is the only credible route to parts availability, a service network, a warranty somebody will honour in year four, and therefore to a used-car bid that is not a guess. Those are the things a lease price is actually a bet on, and they are the ones that decide what the car is worth when it comes back, just as they decide whether a new brand still has a network three years after the launch. A European compliance plate on its own has never moved that number.

These are entry, footprint and network decisions taken at the same table, and they are usually taken with worse information than the product decisions sitting next to them, which is most of what an automotive strategy mandate is spent on.

For anyone who has to act on this rather than read about it, the undertaking mechanism has handed over a usable test, because Brussels wrote down what it accepts as a real commitment. Three questions separate an announcement from a building site. Has equipment been ordered and installed, or is there only a signed memorandum? Is there a named model with a homologation timetable, or a capacity figure with no product behind it? And is the commitment attached to something that bites if it is missed, a milestone, a clawback, an incentive that can be withdrawn, or is it attached only to a press release?

Szeged passes all three. Turkey failed the third one first, and Ankara then proved the point by pulling the incentives. If the answer to any of the three is missing for a plant your order book, your residual values or your market share depends on, that is not a strategy question. It is one question with a researched answer, and it is the shape of a commissioned piece of research: one question, the base rebuilt from registrations, filings and capacity, and a written answer you can act on or ignore. The map will not tell you. It was never built to.