Every Saturday morning I read Dan Neil’s “Rumble Seat” column the way other people watch SNL cold opens. Neil is the only auto writer in the English-speaking press who reliably puts a working pull-quote on page D8 of the Saturday-Sunday WSJ. He is also the only one who can drop the words “jocund modernism — call it orbital chic” into a car review and make you reread the sentence twice to confirm it actually ran. That is craft.

The April 25 column is about the 2026 Polestar 4 Dual Motor. It is a $66,900 base, $81,800-as-tested, 544-horsepower, 3.7-second-to-sixty premium electric “SUV coupe” from the Swedish-Chinese collab born out of Volvo’s performance sub-brand. And it has no rear window. Not a small rear window. Not a heavily tinted rear window. No rear window. The hatch lid is fabric-headlined inside and unbroken sheetmetal outside. Where the rearview mirror normally lives there is a video display, fed by a roof-mounted camera, doing its level best to look like a piece of glass.

Neil’s verdict, after a few days behind the wheel: “Drivers won’t miss the mirror.” Mine, after a few days reading the column: the Polestar 4 is the cleanest physical artifact of the auto-AI cockpit thesis we have published all year. And no — we are still not buying Geely.

The premise, in one paragraph

In most of the English-speaking world the front piece of a car is called a windshield. The back piece is called — Neil reminds us — a “backlight,” an archaism from the horse-drawn era. The Polestar 4 does not have one. The hinged trunk lid is concealed in the geometric cutlines of the bodywork. Inside, the glass roof panel ends just over the rear headrests. Beyond that, the black-fabric headliner falls into a shallow, empty space that Neil compares to “a darkened theater stage—that bit should be upholstered, in my opinion, like the cargo hold of a Jaguar E-Type.” Lacking a direct, over-the-shoulder view, the driver depends on side mirrors, short-range ultrasonics, four short-range cameras (for 360-degree views), and one dedicated rear-facing camera that pipes video to a display mounted at the top of the windshield, disguised as a conventional rearview mirror.

This sounds insane. It is not insane. Neil walks through why. Deleting the backlight buys you aerodynamic efficiency, which matters disproportionately on a car whose mortal enemy is range loss. A rear window large enough to be useful would “disturb the laminar flow of air rushing across the rear decklid, increasing shearing turbulence and drag.” And, as he points out, the rearview mirror in most modern full-size SUVs is already functionally useless — the glass in the rear hatch is about ten feet from the driver’s eye, with two headrests in between, the bottom of the backlight five feet off the ground, a typical traffic cone invisible from closer than thirty-five feet. The Ferrari guys figured this out fifty years ago. Marcello Gandini’s 1974 prototype for what became the Lamborghini Countach LP400 specified a rear-facing periscope. The periscope never made production, but the distinctive roof channel did — and earned the early cars the nickname “Periscopio.”

So the engineering case is sound. The aesthetic case is, in Neil’s phrase, “jocund modernism — call it orbital chic.” The driving experience, per the review, is “serene and well-mannered in daily driving” with a Performance Pack that “will be glad to slap you around” — 0-60 in 3.7 seconds, 544 hp combined, 506 lb-ft of torque, super-swole 22-inch forged-aluminum wheels, looks-maxxing tires. The hole-shot acceleration, he writes, “is properly heady and roller-coaster-y.” At highway speeds the car’s “sharp reflexes and deep reserves of squirt create significant moral temptation.” The line that made me stop and laugh: “If you are for some reason campaigning for a life-altering speeding ticket, this car will make it easy.”

Why this is the cleanest auto-AI tell of the year

I wrote on Thursday about the four-piece picture that’s reshaping the auto industry: a million U.S. new-car buyers walking away, Qualcomm signing Stellantis, Pope Leo XIV’s first encyclical on labor displacement, and a Beijing court forcing a Chinese tech firm to pay damages for an AI-driven firing. The Polestar 4 column dropped six weeks before that piece. It is, in retrospect, the artifact those four stories were all pointing at.

Here is what is actually happening when you delete the rear window of a car. You are not making a styling decision. You are admitting that the camera-sensor-compute stack is now more reliable than the glass-and-eyeballs stack it replaced. That is the entire auto-AI thesis in a single design choice. The Polestar 4 ships with four 360-degree short-range cameras, one dedicated rear-facing camera, an ultrasonic array, and enough on-board compute to fuse all of it into a video display the driver trusts more than a physical mirror. The hardware bill of materials of that car has migrated. Forty percent of the cost of a 2026 premium EV is now silicon and sensors. In 2010 it was twenty percent. By 2030 it will be sixty.

The OEMs do not build that stack. Polestar, like every other premium EV maker, buys it. The cockpit silicon comes from Qualcomm (QCOM) or Nvidia (NVDA). The driver-assist and computer-vision stack comes from Mobileye (MBLY) or, increasingly, Nvidia’s Drive Orin/Thor platforms. The base SoCs come from Taiwan Semi (TSM). The OEM is now, structurally, an integrator. They pick a chip partner, pick a sensor partner, design a body, write a UX layer on top, and ship. The economic rent has moved.

The Polestar 4’s missing rear window is the most visible piece of evidence on the market that this shift has happened. Every other piece of evidence — the Qualcomm-Stellantis deal, the Mercedes-Nvidia multi-year, the BMW Snapdragon 5-Series, the Mobileye SuperVision wins at Geely and Polestar — sits in press releases that normal people do not read. Dan Neil’s column sits on D1 of the Saturday WSJ next to a piece about ring watches at Watches & Wonders. That is how a thesis becomes consensus.

You are not making a styling decision when you delete the rear window of a car. You are admitting that the camera-sensor-compute stack is now more reliable than the glass-and-eyeballs stack it replaced.

Why we are still not buying Geely or Polestar

Polestar is, as Neil notes in passing, owned by Geely Holding — the Chinese conglomerate that also owns Volvo Cars. The temptation, looking at a 544-horsepower premium EV that out-styles and arguably out-engineers anything in the BMW or Mercedes lineup at the price point, is to back the manufacturer. Resist it.

Three reasons we are not buying Geely or Polestar in the model.

One: the wrong layer of the stack. Polestar designs and assembles cars. The economic rent — the durable, high-margin, network-effect-protected piece of the bill of materials — sits in the silicon and the perception stack. That is Qualcomm, Nvidia, Mobileye, and TSM. Polestar buys those parts at standard industry margins and adds twenty percent in design and twenty percent in assembly, against gross margins that are perpetually compressed by EV price wars in China. Even if Polestar wins every premium-EV buyer in Europe and North America, it is still a low-teens gross margin business. Mobileye runs in the high forties. Nvidia’s automotive segment runs in the high fifties. Same industry. Different math.

Two: the China discount. Geely is a Chinese state-adjacent conglomerate. Its U.S. ADR is illiquid. The April 25 WSJ also ran a piece on Ford and Geely “discussing” bringing Chinese EV technology to America, with the explicit caveat that “a U.S. deal would be politically fraught.” Translation: a Trump-era administration that has been jawboning the Section 301 tariffs on Chinese EVs is not going to wave through a Geely-Ford joint venture in 2026 or 2027. Polestar’s addressable U.S. market is structurally smaller than its press releases imply.

Three: the battery pack is yesterday’s pack. This is Neil’s sharpest punch in the column, dressed up as a throwaway line. The Polestar 4 ships with a 100 kWh nickel-manganese-cobalt (NMC) pack on 400V charging, delivering 255 miles of official range and a 10-80% recharge in thirty minutes. Geely already has next-generation LFP batteries and 800V charging in the field on other models. Why didn’t this car get them? Neil’s answer is dry: the Polestar 4’s pack “might as well have a sign that says, ‘Watch this space.’” That is a four-word indictment of the product cycle. The car you are looking at is already a half-generation behind its own parent company’s technology. Whatever value Polestar accretes from this model will be repriced when the next pack lands and obsoletes the current fleet.

What we are buying instead

The position book is unchanged. The Polestar 4 column hardens the case for the names we already own and the ones we are adding on weakness.

Nvidia (NVDA) — core overweight. Drive Orin is already in Mercedes. Drive Thor is the 2026-2027 platform. Every premium EV that ships with a rear-window-deletion-class sensor fusion stack is, statistically, a Drive Orin or a Snapdragon Ride. The auto segment is still a low-single-digit percentage of Nvidia’s revenue. It will not be in 2028.

Mobileye (MBLY) — core overweight, added on Thursday. The SuperVision and Chauffeur stacks are the most-deployed eyes-off systems in the premium-EV market outside China. Mobileye’s relationship with the Geely Group dates back to 2022. The Polestar 4’s 360-degree camera array is not Mobileye, but the driver-assist and autonomous-driving stack on the larger Polestar lineup runs Mobileye SuperVision. The stock has been treated as a sentiment proxy for the slow part of the AV story. It is, structurally, a winner on the fast part — the in-cabin perception story that the Polestar 4’s deleted rear window crystallizes.

Qualcomm (QCOM) — added 2.5% on May 26. Snapdragon Cockpit and Snapdragon Ride are the in-dash and ADAS platforms underneath Stellantis, GM Ultifi, BMW’s new 5-Series, and a chunk of the second-tier Chinese OEM lineup. The Stellantis deal that ripped 12% on May 26 is the structural confirmation. We added.

Taiwan Semi (TSM) — core overweight. Every chip mentioned above is fabricated at TSMC. The fab is the chokepoint. We are long the chokepoint.

Geely (GELYY ADR) — watch, not buy. See three reasons above. We revisit if Section 301 posture changes materially or if the next-generation 800V LFP pack lands across the Polestar lineup at a competitive price.

Polestar (PSNY) — watch, not buy. Same logic. Beautiful car. Wrong layer of the stack. Wrong battery generation. Wrong jurisdiction for the U.S. addressable market we need.

The Dan Neil bonus paragraph

Sean’s side-note. The reason I read Dan Neil every Saturday is the same reason I read Michael Lewis books: he is one of about six writers alive who can sneak a real opinion into a product review and have it land. The Polestar 4 column has at least four sentences that would survive a careful edit at The New Yorker. The Italian-driving line lifted from “The Gumball Rally” (1976) — Raul Julia snapping the rearview mirror off a Ferrari Daytona and tossing it out the window: “What’s behind me is not important” — is one of the best closing rhetorical moves in any car review of the last decade. Read the column twice. The first time for the engineering. The second time for the writing.

And if you do buy a Polestar 4, do not text me from it at highway speed. I cannot afford the moral temptation.

Companion Reading In Tonight’s Edition

End-Of-Week Recap — Three Records To Close May. — Friday’s end-of-week wrap. Dow, S&P and Nasdaq all at all-time highs. The book added Tesla (TSLA) and Dell (DELL).

The Auto Industry’s AI Collision. — One million lost new-car buyers, Qualcomm-Stellantis, Pope Leo XIV’s encyclical, and a Beijing wrongful-termination ruling. The four-part picture of the auto industry being rewritten in real time.

Schedule A 15-Minute Portfolio Review →