The Restricted Group

On 25 August two credit funds filed in a Manhattan federal court for discovery against a lender, a brand-licensing group and two investment banks, in aid of litigation they intend to bring in London against a 113-year-old British car maker. Six weeks earlier that car maker had raised £550 million secured on assets placed in a newly incorporated subsidiary. In the same half-year its revenue rose 38 percent, its gross margin widened by nearly six points and it turned an EBITDA loss into an EBITDA profit. Every one of those facts is true simultaneously. Hold them together and you have the clearest available specimen of what is actually failing in European industry, which is not the engineering.

Architecting Alpha, Part X

Three Borrowed Pillars took a Chinese state journal's autopsy of the German model and ran the audit: security, energy, market, three inputs rented from three landlords and repriced inside four years. That essay worked at the level of a country. This one runs the same audit at the level of a single balance sheet, because a country is an abstraction and a capital structure is a contract, and contracts are where you can see the loss being allocated in real time.

Aston Martin Lagonda is the specimen. Not because it is representative in size, which it is not, but because it is small enough to read end to end and slow enough to have been failing in public for eight years. Everything that is happening to European industrial credit at the aggregate level is happening to this one company in legible form, with dates and filings attached.

And it rented a fourth pillar that Germany also rented and nobody counted. Time.

The claim, and how it fails

European industrial decline is not primarily a manufacturing problem. It is a duration mismatch. Capital structures underwritten for a forty-to-fifty-month product cycle are now financing competitors running twenty-four, and the resulting adjustment does not arrive as a price in the bond market. It arrives as a legal event. Assets move across the covenant perimeter, the newest money ranks ahead of the money already there, and the loss gets allocated by documentation rather than by the market. This is why the aggregate spread indices show nothing wrong. Nothing is wrong with the average. Everything is wrong with the dispersion, and an index cannot price the seniority of your particular claim.

Four ways this is wrong. If Aston Martin is simply a badly capitalised company, which on its record it obviously also is, then the specimen is diseased for idiosyncratic reasons and proves nothing about a continent. If the operating recovery is real and the financing bought exactly the runway it was meant to buy, then this is competent balance-sheet management being misread as decline, and the bondholders are arguing about the division of a going concern rather than the wreckage of one. If the development-cycle gap is softer than the consultancies selling transformation programmes claim, and I will show below that it is, then the duration argument is built partly on marketing. And if European high-yield spreads sitting inside their 2021 tights are telling the plain truth, then I am reading a structural crisis into an asset class that is pricing none, which is precisely the error I would attack in someone else. I take all four at the end.

Part I. The number is wrong, and the correction is the story

The figure in circulation is 550 million dollars. The figure is £550 million, and the currency is the least interesting thing about the error.

On 22 July Aston Martin announced a new debt financing: a £450 million senior secured term loan plus a £100 million delayed draw facility, provided by funds managed by HPS Investment Partners, priced at SONIA plus 6.75 percent and maturing in July 2031. Proceeds repaid and cancelled the fully drawn £170 million revolving credit facility and the £20 million drawn under Lawrence Stroll's consortium facility. Pro forma liquidity went from £145.2 million at 30 June to roughly £340 million. On any ordinary reading this is a company terming out short debt and buying two more years. It is also, on its face, expensive rather than exotic.

The exotic part is the security. The announcement says the loan is secured on "certain of the Group's assets situated in a newly incorporated subsidiary, together with certain other assets of the Group." Two days later the company issued a response to investor questions confirming that a further new subsidiary had been designated an unrestricted subsidiary, and reassuring holders of its 2029 notes that those notes remain secured by a pledge over the shares in Aston Martin Lagonda Limited, "an indirect parent entity" of the new structure.

Read that sentence the way a credit analyst reads it. Your security is no longer a claim on the assets. It is a claim on the shares of a company that is a parent of a company that holds the assets, which is a different instrument with a different recovery. Fitch said so in its own language on 31 July, cutting the senior secured debt of the issuing entity from CCC+ to CCC and the Recovery Rating from RR4 to RR5, because the new facility ranks above the debt in the restricted group, carries a different security package, and leaves limited visibility on the underlying collateral. That is a rating agency stating that expected recovery fell, not because the business deteriorated, but because the assets moved.

On 25 August Arini Capital Management and Tresidor Investment Management filed a Section 1782 application in the Southern District of New York seeking documents and testimony from HPS, Authentic Brands Group's UK entity, Moelis and Lazard, in aid of anticipated English proceedings. Aston Martin, through Latham & Watkins, says the transaction is lawful and consistent with the bond documentation. It may well be right; that is what the litigation is for. Note also what has not been disclosed: reporting that a majority of the non-automotive brand rights went into a new vehicle alongside a licensing partner is press coverage, not a company filing, and the 22 July announcement names neither the counterparty nor the assets. I am not going to treat unfiled reporting as fact in an essay that is about the difference between what documents say and what people assume they say.

One more correction while I am in the weeds, because it will otherwise propagate. The issuing entity, Aston Martin Capital Holdings Limited, is incorporated in Jersey, not the Cayman Islands. The Cayman reference floating around the diagrams belongs to the new 2026 structure, which is a different entity doing a different job. If you are going to argue from a corporate structure chart, get the jurisdictions right, because jurisdiction is the entire point of the chart.

What this is, stripped of the specifics, is a liability management exercise. The technique is American. It was refined on J.Crew, Serta Simmons, Envision and a dozen others, and it works by finding the gap between what a covenant package literally permits and what its signatories assumed it permitted. Europe imported that technology considerably faster than it imported any manufacturing method invented in the same decade. That is worth sitting with for a moment.

Part II. The company is not failing at building cars

Here is the part that makes this interesting rather than merely sad.

The first half of 2026 was the best operating period Aston Martin has had in years. Revenue of £628.6 million, up 38 percent. Wholesale volumes of 2,331 against 1,922. Gross margin of 33.8 percent against 27.9. Adjusted EBITDA of £62.7 million against a £3.0 million loss. Free cash outflow of £197.6 million against £321.0 million. Valhalla, the mid-engined hybrid that the whole product plan leans on, delivering around 225 units in the half. Full-year guidance reaffirmed at both the first quarter and the half, which is not something a company in operational freefall does twice in a row.

Set that against the trough it is climbing out of. FY2025, reported on 25 February 2026: revenue £1,257.7 million, 5,448 wholesales, a £493.0 million loss for the period, £409.9 million of free cash outflow, gross margin down 750 basis points. Up to 20 percent of the workforce removed for roughly £40 million of annual savings. Five-year capital expenditure cut from about £2.0 billion to about £1.7 billion by rephasing battery-electric investment.

Now the two numbers that matter, and they are both the company's own guidance.

Full-year 2026 wholesales are guided to be similar to FY2025's 5,448. Net cash interest for 2026 is guided at approximately £160 million, raised from £150 million at the half. Divide one by the other. Aston Martin is carrying roughly £29,000 of interest per car sold, before a single pound is spent designing the next one. That is the whole essay in one ratio. The gross margin can go to 40 percent, the factory can run perfectly, Valhalla can sell out, and the company still has to produce twenty-nine thousand pounds per vehicle for its lenders before it funds the future.

The coupons explain it. In March 2024 the group refinanced into $960 million of 10.000 percent senior secured notes and £400 million of 10.375 percent notes, both due 2029. It then tapped the same instruments twice, in August and December 2024, to repay a drawn revolver. Outstanding at 30 June 2026: $1,050 million and £565 million. Net debt £1,544.7 million. Market capitalisation, at 34.30 pence on 15 September, roughly £348 million. The notes have been quoted around fifty-seven cents.

Put those two paragraphs side by side. The equity is worth about a fifth of one year's revenue. The debt is worth about two and a half years of revenue. The operating business is improving on every line. This company is solvent on product and insolvent on time, and that distinction is the thing the aggregate data cannot show you.

Part III. What actually broke was the cycle

Why does a luxury car maker end up paying ten percent? The easy answer is that it has raised equity seven times since listing at £19.00 in October 2018 and has a market capitalisation down roughly 92 percent since. That answer is true and insufficient, because it describes the symptom in the tense of the cause.

The structural answer is duration. A capital structure is a bet on a product cycle. If you amortise a platform over seven years and your competitor replaces theirs every two, your depreciation schedule is a liability and theirs is a weapon.

Look at what the numbers confess. Aston Martin's 2026 guidance is for adjusted depreciation and amortisation of £375 to £400 million against capital expenditure of about £300 million. The company is amortising materially more than it is spending. That is the arithmetic signature of a business living off a platform investment made in a longer-cycle world, and it is the same signature you will find across the European supply base if you look.

The cycle-length claim needs handling honestly, because it is the weakest link in the chain and the version circulating is inflated. The widely repeated framing is eighteen to twenty-four months in China against forty-eight to sixty in Europe. The sourced range is narrower. AlixPartners puts Chinese new-energy development cycles at as little as twenty months with Western counterparts roughly double. McKinsey puts Chinese new entrants at about twenty-four months concept to launch against forty to fifty for legacy makers. Nobody publishes a methodology, the clocks being compared start at different points, and the Chinese figures often describe derivative programmes on existing platforms measured against Western clean-sheet ones. Both firms sell transformation work to Western manufacturers, which means the gap is the product. Treat the headline months as directional and the mechanism as the real finding.

The mechanism is where it becomes uncomfortable. McKinsey's decomposition puts the largest single lever at virtual validation: Chinese manufacturers simulate roughly 65 percent of testing in software against 40 to 50 percent elsewhere, worth nine to eleven months on its own. That is not a labour-cost story or a subsidy story. It is the same substitution this series has been tracking since The Tolerance War: the displacement of physical iteration by simulated iteration, which converts calendar time into compute time, and compute time is the one input whose price has been falling.

And the strongest evidence that the compression is real is adversarial. Chinese regulators have opened a year-long safety campaign with unannounced inspections and proposed doubling mandatory road testing for new-energy vehicles to 30,000 kilometres. Chery's vice president put it plainly: "There are simply some development timelines we cannot afford to shortcut." A state telling its champions to slow down is a better confirmation that they sped up than any consultancy chart.

Part IV. The index says nothing happened

Here is the anomaly that should stop anyone reading European credit from the top down.

The ICE BofA Euro High Yield option-adjusted spread was 268 basis points on 11 September. Roughly 285 a year earlier. The long-run average is around 566. Against the June 2021 tights of about 300, current spreads are not merely unstressed, they are through the last cycle's low, and sit in roughly the tightest five percent of readings in twenty-five years. A caveat on my own source: the Federal Reserve's series now retains only three years of history, so the 2021 comparison comes from secondary commentary rather than a primary print, and I would not build a trade on that specific number.

Now look underneath it. The distress ratio, the share of the index trading beyond a thousand basis points, was 5.7 percent at the end of July, its highest since March. The index is roughly two thirds BB-rated, at record levels of quality composition, which is what an index looks like when its weakest members have been downgraded out of it rather than repaired. And the structural fact that matters for this essay: automotive is roughly nine percent of the European high-yield index against roughly two percent of the American one. Europe's speculative-grade market is levered to cars in a way America's is not.

The migrations are the evidence. ZF Friedrichshafen was investment grade a cycle ago; S&P cut it to BB- in April 2025 on leverage, and the company then reported a €2.1 billion net loss for 2025 with €10.2 billion of net debt and an equity ratio of 13.3 percent, while cutting 7,600 jobs in its powertrain division. Stellantis was downgraded by both Moody's and S&P on 10 February 2026 to the bottom rung of investment grade after a €22.3 billion net loss and €25.4 billion of unusual charges. Volkswagen's outlook went negative in December 2025 with published downgrade triggers. Porsche's automotive division operating margin went from 14.5 percent to 0.3 percent. Bosch is now at roughly 22,000 announced Mobility job cuts. Continental's spun-off automotive arm announced 4,000 more and a plan to push research and development below ten percent of sales.

And the aggregate, which is unambiguous. German large insolvencies, meaning filings by companies above €10 million of revenue, ran to 471 in 2025, up 25 percent, against 163 in the 2021 low. Fifty-nine of them were automotive suppliers. German automotive employment stood at 691,500 at the end of the first half of 2026, down 5.8 percent year on year and the lowest since 2005. The industry association's own projection in May put the loss at about 100,000 jobs since 2019 and 225,000 by 2035, worse than its own study two years earlier by 35,000.

So we have an asset class at multi-decade tight spreads sitting on top of a sector shedding its second-largest employer base and filing insolvencies at three times the 2021 rate. Both readings are correct. The resolution is that the average has not moved because the average is not where the adjustment is happening. Loss is being allocated issuer by issuer, document by document, and in this case courtroom by courtroom. An index prices the average claim. It cannot price the seniority of yours, and seniority is now the variable doing the work.

Part V. The second European case, read against the first

Run the Part IX audit on this one balance sheet.

Security, meaning who guarantees the perimeter. Aston Martin builds in Warwickshire and sells heavily into the United States, where UK-built cars face 10 percent within an annual quota of 100,000 vehicles and, on the company's own stated all-in reckoning, 27.5 percent outside it. The quota is administered quarterly at 25,000 units. In its half-year results the company describes having carefully managed its US imports towards the end of each quarter and says it will try to optimise production schedules to reduce quota risk. Read that literally: a manufacturer is sequencing its factory against a foreign customs calendar it has no vote in. That is not a production schedule. It is a tariff schedule wearing a production schedule's clothes. Note also what the company has never published, despite the figure being widely attributed to it: any quantified tariff cost. The £65 million that gets cited is warranty, dealer support and quality investment, stated as such in its own results.

Market, meaning who is permitted to buy from you and on what terms. Britain's zero-emission vehicle mandate requires 33 percent of car sales to be zero-emission this year, rising to 80 percent in 2030. It is also, right now, under a consultation that opened on 14 August and closes on 23 October 2026, covering trajectory adjustments, flexibilities, alternative approaches and the definition of the 2030 phase-out itself. No decision has been announced. Aston Martin has rephased its battery-electric investment and taken a £42.7 million impairment on discontinued vehicle programmes in the first half. It is making seven-year product commitments against a regulatory trajectory that is itself an open question. If your consultation period plus decision lag exceeds a competitor's entire development cycle, you have lost that argument before you decide it, and the content of the decision is close to irrelevant.

Time, the pillar nobody counts. Germany rented security, energy and market. Aston Martin rented all three and then rented time as well, at ten percent, from a bond market that was pricing a world in which a platform lasted seven years. When the rent came due it did not find more time. It found a law firm, a new subsidiary, and a perimeter it could move assets across.

That is the general case. A restricted group, in bond documentation, is the ring of entities the covenants actually bind. What is inside it is promised to you. What is outside it is not. The entire craft of liability management is the art of moving value across that line while the paperwork still says yes.

Europe's industrial base is a restricted group in exactly that sense, and its assets are already moving. BYD is in full production at Szeged in Hungary at 150,000 units a year. Chery is building Omodas in Barcelona, in a plant Nissan left, at another 150,000. Leapmotor is in Zaragoza, Polestar is going to Košice, MG is building in Galicia for 2028. Nissan has taken Sunderland to a single line at roughly half capacity and is marketing the freed capacity to Chinese partners, with Chery the leading candidate. Britain's stated path back to volume runs partly through Chinese manufacturers using British plant.

Chinese brands took somewhere between nine and eleven percent of European registrations in the first half of 2026, roughly double a year earlier, depending on whether you count by brand or by group and whether you include the UK. The sharpest detail is not the share but the route: roughly a quarter of EU hybrid and plug-in hybrid sales now go to Chinese brands, which is a deliberate walk around a tariff regime written in October 2024 specifically for battery-electric vehicles, at rates from 17.0 percent for BYD to 35.3 percent for SAIC on top of the standard 10 percent duty. Build the wall around one drivetrain and the traffic reroutes through another. In February the first exemption under the minimum-price mechanism went not to a Chinese manufacturer but to a European one's China-built car, the CUPRA Tavascan, in exchange for a minimum price, a volume cap and committed European investment with milestones.

And the timing tell. AlixPartners forecast in July 2025 that Chinese brands would reach ten percent of the European market by 2030. They passed it in the middle of 2026. The consultancy baseline was four years slow, which is the same forecasting error, in a different domain, as underwriting a seven-year amortisation schedule against a two-year competitor.

The three positions

Operator: read your own documentation before your creditors read it for you. You are inside somebody's restricted group and you probably do not know its boundary. The specific test is short. If your last financing required new collateral, where did that collateral come from, and who previously had a claim on it? If your answer involves a subsidiary incorporated within the last eighteen months, you have already run this playbook or had it run on you. The second test is the ratio from Part II: divide your annual cash interest by your annual unit output, and if that number is large relative to your gross profit per unit, your product plan is decorative. Fix the duration before you fix the product.

Investor: price the seniority, not the spread. The distress in European industrial credit is entirely in the dispersion, and the index will keep telling you otherwise. Where covenant packages written between 2019 and 2021 are financing assets whose competitive cycle has halved, the assumption that is wrong in the price is recovery, not default probability. The instruments worth reading are the ones where a structurally subordinated new facility has been layered above an existing secured claim, because that is where the loss is being allocated first and where the documentation, rather than the business, determines who takes it. This is the recognition mechanism from Regime Change running in reverse: instead of an unbooked asset being moved onto the ledger, a booked claim is being quietly moved off it.

State: an industrial policy that does not price time is not an industrial policy. Britain has committed £4 billion through DRIVE35, against £15 billion from Jaguar Land Rover, £4 billion for a gigafactory and £2 billion of Nissan programmes, targeting 1.3 million vehicles a year by 2035 from 2026 output of roughly 740,000, itself down from a 2016 peak above 1.7 million. The capital is not the binding constraint. Decision latency is. A permitting regime, a grid connection queue or a mandate consultation that runs longer than a competitor's product cycle is a subsidy paid to that competitor, funded by your own manufacturers, and it does not appear in any budget line.

How this could be wrong

Aston Martin may simply be a bad company and nothing more. Seven equity raises since 2018, a 92 percent decline in market capitalisation, chronic undercapitalisation that predates every macro claim in this essay. If the specimen is diseased for reasons specific to itself, then it is a case study in balance-sheet mismanagement and the continental argument has to be carried entirely by ZF, Stellantis and the insolvency count. That is a defensible position and I would not fight it hard.

The financing may work exactly as designed. Revenue up 38 percent, margin up nearly six points, EBITDA positive, guidance reaffirmed twice, liquidity roughly £340 million pro forma, and a car in Valhalla that people actually want. If free cash flow turns in 2027, the £550 million bought precisely the runway it was meant to buy, the notes recover from the fifties, and the litigation resolves into an argument about the division of a healthy enterprise rather than the remains of a sick one. The company's lawyers say the transaction is lawful and consistent with the documentation. If the English courts agree, then the covenants did what they were written to do and the story is about diligence by creditors, not decline by a continent.

The duration argument leans on numbers I have already discounted. If the true gap is forty-five months against twenty-four rather than sixty against eighteen, it is half the size the popular framing implies, and half of a soft number is not a thesis. The mechanism, virtual validation displacing physical iteration, is the part I would defend; the month counts are the part I would surrender first.

And the market may simply be right. European high-yield spreads are at multi-decade tights, the trailing default rate is under two percent including distressed exchanges, and S&P expects the European speculative-grade default rate to fall towards 3.25 percent by December. Thousands of people whose job is to price this have looked at the same insolvency data and concluded it is contained. Reading a structural crisis into an asset class that prices none is a real failure mode, and if I am wrong that is the shape the wrongness will take: the dispersion normalises, the migrations stop, and the whole argument turns out to have been a description of one sector's bad decade rather than a continent's changed condition.

If one of those four breaks first, write to info@selfbuiltsystems.com and tell me which. Short and specific gets an answer.

The lesson is not about cars. A restricted group is any perimeter inside which your claim is honoured, and outside which it is merely mentioned. Every operator sits inside one, every investor holds a claim defined by one, and every state is one. The question that decides the next decade is not whether you can build. Germany could build. Aston Martin can build. The question is whether the documentation governing you was written for the speed you are now running at, and what happens to the things you thought you owned on the day somebody discovers that it was not.

Architecting Alpha is published in the spirit of bold conjecture and ruthless criticism. Every claim is linked to its source. Where a figure is widely repeated but untraceable to a filing, it is marked as such in the text rather than borrowed.