The Auto Industry’s AI Collision

The Wall Street Journal’s Thursday front page led with a story we have been waiting eighteen months to see in print: New-Car Sales Slump Deepens. Sharon Terlep’s reporting put the number on it: around one million prospective buyers have defected from the U.S. new-car market since the start of the decade, and they are not coming back any time soon. Annual U.S. unit sales are running at roughly 16 million versus the 17 million pre-pandemic pace. Average transaction prices have settled around $50,000. Industry analysts who used to forecast a return to pre-pandemic volumes by 2026 are now pushing those estimates out to 2029 or later. Volvo’s commercial chief told the paper, “This is a real threat to the whole industry.”

It would be tempting to read this as just another rates story — high rates + Iran-driven gas prices = soft new-car demand, simple. It is not that simple, and a portfolio that treats it as a one-line macro print is going to miss the actual trade.

Three other items from the same week tell the rest of the story.

1) Qualcomm + Stellantis: the silicon vendors are taking over the dashboard

On Tuesday Qualcomm (QCOM) closed +12% on news that Stellantis — the parent of Jeep, Ram, Chrysler, Dodge, Fiat, Peugeot, and a dozen other badges — will use Qualcomm’s Snapdragon Cockpit and Snapdragon Ride platforms across its global vehicle lineup. The deal is structurally interesting because of what it confirms: the major automakers, faced with the impossibility of building competitive in-car AI in-house, are outsourcing the cockpit to the chip companies. Stellantis is not the first. GM has been deploying Qualcomm’s Snapdragon SA8540P silicon in its Ultifi platform. Mercedes-Benz has its own multi-year deal with Nvidia (NVDA). BMW has Qualcomm in the new 5-Series. The pattern is the same: the OEMs are abandoning the “we will build it ourselves” strategy and writing checks to the chip designers.

This is the single most important read-through of the new-car sales slump. The headline narrative says cars are too expensive. The structural reality is that the buyers who walked away are also walking away from the dealer-finance complex, the franchise-service network, and the legacy OEM brand premium. The buyers who do come back — or who upgrade their existing cars to AI-enabled vehicles two product cycles from now — will pay premiums for cockpit features, autonomous driving, and integrated AI agents that the OEMs cannot build alone. The economic rent in the new car of 2028 sits with Qualcomm (QCOM), Nvidia (NVDA), Mobileye (MBLY) and Taiwan Semi (TSM) — not with Ford (F) and General Motors (GM).

2) Tesla’s European rebound: the divergence we needed to see

In the same B-section Thursday, Tesla’s European registrations grew “strongly” in April after a year-long slump. We are not breathless about a single month of registrations data. What matters is that the recovery is happening while the broader European auto market is still weak, and while Li Auto — the most-watched Chinese EV peer — warned of a 16-to-20% revenue decline for Q2 and saw its gross margin shrink from 20.5% to 7.9%. That divergence is the buy signal. Tesla is decoupling from the broader auto slump.

We added TSLA to the tactical book Friday at a 2% position. The thesis is not that Tesla is undervalued on traditional auto multiples — it is not. The thesis is that Tesla is the only listed company that captures cockpit silicon (FSD compute), driver-assist (FSD V12 / robotaxi optionality), connectivity (OTA), battery / drivetrain, AND brand premium — all four layers of the stack in a single ticker. When auto buyers come back, they are coming back to a vehicle that looks like a Tesla, not a 2019 Ford F-150 with a touchscreen retrofit.

3) Pope Leo XIV’s Magnifica Humanitas: the cultural backlash is shifting from rhetoric to policy

Pope Leo XIV released his first encyclical last week. Forty-three thousand words on artificial intelligence. The line that has been quoted in every editorial board response is this: “Income from capital risks replacing income from labor.” The pope is not the only one saying it. Greg Ip’s Capital Account column Friday made the same observation in market language: U.S. labor’s share of gross domestic income fell to 51% in Q1, the lowest since records began in 1947. Profits’ share rose to 12.1%, the highest since 1950.

“Income from capital risks replacing income from labor.” — Pope Leo XIV, Magnifica Humanitas, May 2026

This is not just papal rhetoric. Greg Ip notes in his column that California voters will likely be asked to vote on a billionaire tax in November. Proposed taxes on AI are catching on in “some corners.” The House Transportation bill quietly passed last week tucks a new federal “fee” on electric vehicles and hybrids into a $600 billion reauthorization. The bill is being sold to MAGA crowds as “an overdue levy on freeloading Tesla owners who pay no gas tax.” Read the fine print and it is a Trojan horse for a national car tax administered through state DMVs — converted, as Florida Governor DeSantis described it over the weekend, into “a branch of the Internal Revenue Service.” Kimberley Strassel called the proposal “one of the worst Republican ideas to come along in a while” in her Friday Potomac Watch column.

For the auto investor, this matters in two ways. First, EV pure-plays just lost a piece of their tax-arbitrage moat. If the federal “EV fee” becomes law, the gasoline-tax holiday that Tesla owners have enjoyed since 2018 is gone, and the spread between the cost of operating a gasoline car and the cost of operating an EV narrows. Second, auto-AI itself becomes an explicit political target. If Pope Leo’s diagnosis is right, and if Greg Ip’s data is right, the political economy of the next presidential cycle is going to look a lot like 1908–1916 — with AI playing the role of the railroads, and progressive taxation playing the role of the antitrust hammer.

4) Beijing’s wrongful-termination verdict: the labor-displacement clock is starting to run

Buried in the WSJ’s Thursday World News page was the most important AI ruling of 2026 to date. A Hangzhou court ruled in favor of a 35-year-old tech-company employee in China after his employer replaced him with AI and tried to either cut his salary 40% or fire him. The employer was ordered to pay $38,000 in wrongful-termination damages. In a separate case, a Beijing court ruled the same way for a 15-year veteran of a tech company doing map-data work who was displaced by AI. The Beijing municipal government’s human-resources bureau published the case as a warning: “Companies cannot use AI upgrading as an excuse for backdoor layoffs and unfair pay cuts.”

Two things matter here. First, the precedent. If China — not a jurisdiction historically friendly to labor — is willing to fine companies for AI-driven layoffs, the United States, the European Union, and every state-level labor agency will be willing to do the same. We expect to see the first U.S. case along these lines by Q4 2026. The companies most exposed are the ones who have most aggressively flagged AI-driven productivity gains in their 10-Qs — legacy software vendors, large-cap call-center BPO operators, and (yes) the parts of the auto industry where AI is being deployed in the back office.

Second, the auto industry’s workforce vulnerability is asymmetric. Detroit is being squeezed from both ends. From the top, the AI cockpit silicon is being captured by Qualcomm and Nvidia — companies the Big Three cannot acquire and cannot out-engineer. From the bottom, the front-line dealer-service workforce, the call-center workforce that handles warranty and financing, and the parts-distribution workforce are all directly in the line of AI displacement. Ford has 174,000 employees. GM has 162,000. Stellantis has 252,000. Even a 10% productivity gain through AI translates to roughly 60,000 jobs that have to be either reabsorbed or let go — in three of the most politically sensitive industrial workforces in the developed world.

We are not short these names. We are simply not long them. The risk-reward is not there yet.

How Capital Wealth is positioned

We laid the new auto-AI book out earlier this week. The summary is straightforward:

Long the cockpit and the silicon. Qualcomm (QCOM, added Tuesday 5/26 at 2.5%), Nvidia (NVDA, core overweight), Taiwan Semi (TSM, core overweight). These are the names that get paid every time a new car ships, whether the badge says Ford or Stellantis or Hyundai. Long the only OEM that owns the full stack. Tesla (TSLA, added Friday 5/29 at 2% tactical). The European-sales divergence is the catalyst, the Stellantis-Qualcomm deal is the read-through, and the EV-fee political risk is offset by the FSD optionality. Long Dell (DELL) as of Friday, at 1.8% tactical, after the +28.75% after-hours rip on $4.2B of AI-server orders — the cleanest direct line from Anthropic’s $965B private mark to a public-equity beneficiary we have seen this cycle.

Long the AI-power complex that the auto-AI buildout requires. Constellation Energy (CEG), GE Vernova (GEV), Vertiv (VRT), NextEra (NEE), Bloom Energy (BE). Every cockpit-AI deployment needs cloud inference, every inference run needs grid power, every grid power megawatt needs a transformer, a switchgear, and a thermal-management vendor. This is the highest-conviction layer of the book.

Long the cybersecurity layer that the auto-AI buildout creates. CrowdStrike (CRWD), Palo Alto Networks (PANW), Palantir (PLTR). Every connected vehicle is an attack surface. Tesla’s May fleet-wide OTA push touched 5.2 million vehicles in 24 hours. The cybersecurity vendors that protect connected-vehicle fleets are the “shatter-proof windshields” of this cycle.

Watchlist only on legacy OEMs. Ford (F), General Motors (GM), Stellantis (STLA), Toyota (TM). We are watching, not buying. There may be a deep-value trade in F or STLA below current levels, but the entry has to come with either (a) a rate cut, (b) a successful AI-cockpit partnership with read-through similar to QCOM-STLA, or (c) a labor settlement that resolves the AI-displacement risk. None of those are in front of us yet.

Watchlist only on EV pure-plays ex-Tesla. Li Auto (LI), Rivian (RIVN), Lucid (LCID). Li Auto’s Q1 gross-margin collapse to 7.9% from 20.5% is what the rest of the Chinese EV space will look like through year-end. We are not buyers.

The 1908 parallel

One thing we keep coming back to: the auto industry’s great consolidation happened twenty years after the Model T launched. Ford was incorporated in 1903. The Model T was announced in 1908. The shake-out years that left only three U.S. automakers standing — the so-called Big Three — happened between 1925 and 1935. The companies that died in that consolidation were not bad companies. They were good companies competing in the wrong cost structure. The companies that survived — GM, Ford, Chrysler — survived because they captured scale before scale captured them.

The 2026 auto industry is in the equivalent of 1925. The AI build-out is the equivalent of mass-production reaching critical density. The next ten years will look like the late 1920s did: a small number of vertically integrated winners (Tesla in the West, BYD in the East), a larger number of legacy OEMs that survive only by stripping costs and licensing the AI cockpit from the silicon vendors, and a long graveyard of names that simply do not make it through the cycle. The investible thesis is the same as it was in 1925: own the silicon (the “tire factories” of this cycle), the AI-server hardware (the “assembly lines”), the power infrastructure (the “electricians”), and the one vertical integrator with brand premium (the “Ford Motor Company” of 2025 is Tesla).

And own none of the legacy OEMs until the consolidation has run.