Physical AI on wheels: electric cars, robotaxis and self-driving trucks, 2026 to 2030

00Executive summary

Self-driving transport is the first big, money-making use of "physical AI": the same chips, models and data loops behind chatbots, pointed at moving people and goods. It is a need, not a want, because the human driver is the most expensive and most dangerous part of every mile. But the money through 2030 is smaller and more concentrated than the headlines suggest.

The five things to know

1. The best self-driving businesses sit inside giant companies where they are under 3% of the value. Waymo (owned by Alphabet, Google's parent) and Zoox (owned by Amazon) are the two most advanced robotaxi operations on earth, and neither moves its parent's stock price. Tesla, where self-driving is nearly all of the value, trades at about 219 times next year's expected profit.
2. Electric cars are the volume; self-driving is the growth rate. U.S. electric-car sales fell to a 5.7% share after the federal tax credit ended and have been flat for three quarters. The rest of the world is having a record year. Robotaxis, self-driving trucks and delivery robots are all electric, so the two stories rise and fall together.
3. Chip content per car rises whether or not robotaxis scale. ON Semiconductor, Allegro and Ambarella get paid every time a carmaker adds driver-assist features or moves to a higher-voltage electrical system. That is the lower-risk way to own the theme.
4. Pure plays have a graveyard. Luminar, Nikola, Fisker, Canoo, Arrival and Lordstown all went to zero. Kodiak, Serve and indie Semiconductor have fallen below the $1 billion size floor. Aurora is the one surviving U.S. self-driving pure play with real cash in the bank, and it still needs to raise money in 2027.
5. Rules, not technology, now set the pace. Zoox's federal permission slip, Nevada lifting its fleet cap, and the federal probe into Tesla's Cybercab matter more to 2027–2028 outcomes than any software release.

01What it is: first principles

A self-driving vehicle is a robot that learned to drive by watching hundreds of millions of miles of human driving, then practicing billions more in a video-game-like simulator. Every part of that sentence maps to a layer of the AI stack, and every layer has public companies attached.

How one ride actually works

The stack, bottom to top

Why electric cars and self-driving are the same trade

Four reasons, and none is ideology. First, a robot needs electronic control of steering and brakes, which electric platforms have by default. Second, the self-driving computer draws 500 to 1,000 watts nonstop, which a gas car's 12-volt electrical system cannot supply. Third, a robotaxi runs 18 hours a day, so energy and maintenance cost per mile dominate its economics, and electric wins both. Fourth, fleet vehicles charge at their own depot overnight, sidestepping the public-charging problem that slows ordinary buyers.

Why this belongs to the AI stack

The training clusters, the inference chips, the model design and the data loop are the same components, from the same suppliers, as the chatbot boom. NVIDIA now reports automotive inside a bucket it calls "Edge Computing." Alphabet trains Waymo on the same TPUs that serve Gemini. The physical world is simply the next place the compute gets pointed, and it is the first place where the output is a paid service (a ride, a truckload, a delivery) rather than a chat window.

02Want or need?

A need. Human driving is the single largest cost in moving people and goods, and the leading cause of accidental death for Americans under 45. Self-driving attacks both.

The economics of the human in the seat

In ride-hailing, the driver keeps most of every fare. In trucking, the driver is 44% of the cost per mile, and the truck sits idle whenever the driver sleeps. In food delivery, Serve Robotics puts the human courier's cost at $8 to $10 per order. Each of these is a labor bill that self-driving converts into a hardware-and-software bill, which is exactly the shift that turned cloud computing into a multi-trillion-dollar industry.

Aurora's own 2030 plan makes the math explicit: $5 billion of revenue on 30,000 trucks is about $167,000 per truck per year, or roughly $0.74 per mile at 225,000 miles. That prices the robot at the driver's wage and leaves the extra operating hours to the trucking company. It is a credible offer to customers earning sub-1% margins.

The safety case

Waymo publishes its record against human drivers on the same roads: 95% fewer crashes with serious injury or worse, 82% fewer crashes with any injury, and 93% fewer injury crashes involving pedestrians, across 271 million driverless miles through June 2026. Tesla does not publish comparable numbers for driverless operation, and the federal safety agency (NHTSA) has escalated an investigation into Tesla's driver-assist software in poor visibility to cover 3.2 million vehicles. The safety case is real but so far belongs to one company.

The people case

The trucking industry's often-quoted driver shortage of about 160,000 by 2030 is contested; a stronger reading is that it is a retention problem, with each replacement costing near $12,800. Either way, wages rise, and rising wages are the demand curve for automation. On the passenger side, an aging population that cannot or should not drive is a market ride-hailing already serves and robotaxis serve more cheaply.

03Market size through 2030

Every market-size number in this space mixes different things (fares vs. company revenue, U.S. vs. global, 2030 vs. 2035). We separate them. Electric cars are a few hundred billion dollars of vehicle sales. Robotaxis are tens of billions of fares. Self-driving freight and delivery are single-digit billions by 2030. The growth rates run in the opposite order.

Checking the robotaxi number from the bottom up

Take Waymo's real productivity: about 500,000 paid rides a week on about 4,000 cars, or 125 rides per car per week. At a $22 average fare that is about $143,000 of fares per car per year. Goldman's $19B therefore implies a U.S. robotaxi fleet of roughly 130,000 cars in 2030, up from about 4,600 today (Waymo ~4,000, Tesla ~500 registered, Zoox a few hundred). That is a 28-fold fleet increase in four years, which requires Waymo's Hyundai and Magna assembly lines, Tesla's Cybercab factory and Zoox's California plant all to work. The bear case of $4B is about 28,000 cars, roughly what Waymo alone could reach on its current path. The bull case of $40B is about 280,000 cars and needs Tesla to solve driverless operation on customer-owned cars.

For context, U.S. ride-hailing fares today total on the order of $65–75B a year (Uber's U.S. rides plus Lyft). The base case is therefore a 20–25% robot share of ride-hailing by 2030 plus some new demand from lower prices. Aggressive, but not absurd: Waymo already carries a meaningful share of San Francisco's rides.

04State of play, September 2026

Three companies charge money for driverless rides in the United States. They are taking three different roads, and one of them is two orders of magnitude ahead.

Three strategies

Where Waymo runs today (15) and where it has announced (not yet live)

The electric-car backdrop

The $7,500 federal tax credit ended September 30, 2025. U.S. electric-car share hit a record 10.6% in the rush before the deadline, then fell to about 5.8% and has stayed there for three quarters. August 2026 sales were down 47% from a year earlier. Ford's first-half electric sales fell 60%; the F-150 Lightning is discontinued. Tesla's U.S. share rose to about 52% only because rivals retreated faster than Tesla shrank.

Underneath, the market got healthier. The price premium for an electric car over a gas car is down to 9.7% ($4,847), dealer inventory is at parity, and used-electric sales are up 15%. Meanwhile the rest of the world is booming: global sales jumped 35% quarter over quarter in the spring on an oil-price shock, Europe's battery-electric share hit 30.5% in August, and China's share hit 65%. The U.S. is the outlier, and the reason is policy.

05Policy and rules

The 2026 story: Washington opened the front door to robotaxis while closing the door on electric-car subsidies. States and cities hold the keys to everything in between.

Self-driving: opening up

• The federal exemption path works. NHTSA granted Zoox a Part 555 exemption from eight safety standards (July 30–31, 2026): up to 2,500 vehicles a year for two years under extra oversight. It is the template for any vehicle without a steering wheel.
• Self-certification is being tested. Tesla chose to certify the Cybercab itself rather than seek an exemption. NHTSA opened investigation AQ26002 on September 3. The outcome decides whether Tesla can sell Cybercabs to the public in 2027 as planned.
• No federal self-driving law yet. Operations rest on NHTSA's exemption program, a standing crash-reporting order, and state law.
• State patchwork. Texas, Florida, Arizona and Nevada are permissive (Nevada now allows up to 1,000 vehicles in the Las Vegas area). California requires a safety driver for Tesla and runs a slower two-agency permit ladder, though it issued heavy-truck testing permits to Aurora and Kodiak in August. Massachusetts still bans driverless operation. New York allows eight Waymo cars with a hand on the wheel.
• Enforcement is active. NHTSA has an Engineering Analysis (the step before a forced recall) open on 3.2 million Tesla FSD vehicles, a school-bus probe and a voluntary recall on Waymo, and the NTSB (the crash-investigation board) opened its first Waymo case in January after a child was struck. A $243M jury verdict against Tesla's Autopilot was upheld in Miami. California's DMV ruled Tesla's "Autopilot" and "FSD" names are misleading; Tesla sued.
• Drones wait on one document. The FAA's rule for flying drones beyond the pilot's line of sight (Part 108) has sat at the White House since July 10. Every drone-delivery forecast depends on it.

Electric cars: closing down

• Credit gone. The $7,500 consumer credit ended September 30, 2025. Harvard's Salata Institute estimates the repeal alone cut 2030 U.S. electric share by 6.2 points.
• Fuel-economy rollback signed. On September 26, 2026 the administration approved new CAFE standards that end the prior path that effectively required electric cars. Regulatory-credit income, already down 67% at Tesla, is structurally gone.
• California's rules in court. A federal judge blocked the EPA from stripping California's authority to set its own tailpipe rules (Sept 2); the EPA appealed Sept 22. California separately repealed its own electric-truck mandate.
• Charging money frozen. NEVI's $5B is frozen; states sued; money trickles out one state at a time. There are 76,236 fast-charging ports; Tesla's share of new ones fell from over 40% to 26% as the automaker-backed Ionna network took 74%.
• China walled off. A 102.5% tariff on Chinese electric cars, which is why Waymo's Chinese-built Zeekr robotaxis cost about $100,000 each landed. A separate rule bans Chinese car software from model year 2027 and hardware from 2030, which is why China's robotaxi companies expand in Europe and the Gulf, not the U.S.

06Large companies best positioned

Ranked on a 1–10 positioning score that weighs asset quality, how much of the company the theme represents, and what the share price already assumes. Latest reported quarter is April–June 2026 for every name.

What the ranking says

The two best robotaxi operations are inside companies where they are rounding errors. That is the opportunity, not a flaw. You can own Waymo through Alphabet at 17 times profit with a cloud business growing 82% attached, or own the same theme through Tesla at 219 times profit with no ride data. The scoring rewards the first and penalizes the second, while keeping Tesla in the basket because it is the only company with a credible path to a sub-$30,000 robotaxi at scale.

Uber is the customer layer, and the market is pricing the risk that Waymo cuts it out. We think that risk is real in three or four dense cities and irrelevant everywhere else: mixed human-plus-robot fleets handle rush hours, airports and weather that robot-only fleets cannot, and Uber has now committed roughly $10B across 30 partners so that no single technology, including Waymo's, is essential to it.

07Small and mid-size suppliers

The screen: U.S.-headquartered, U.S.-listed, worth at least $1 billion, and making something an electric car, robotaxi, self-driving truck or delivery robot cannot ship without. Fifty-five names checked. The uncomfortable finding: most of the best "self-driving suppliers" were re-rated in 2026 for AI data-center demand, not for cars.

Fell below the $1 billion floor

Six names investors still associate with this theme no longer qualify, and the reasons matter. Kodiak AI ($488M) earned more first-half revenue than Aurora but has a $189M shareholder deficit and about four quarters of cash. Serve Robotics ($386M) grew revenue 404% and then cut full-year guidance from $26M to $9–10M after Uber sold its entire stake; two-thirds of orders in its own markets cannot use a robot. indie Semiconductor ($700M) is improving but cash fell to $139M against $431M of debt. EVgo ($434M) and ChargePoint ($240M) are decent charging businesses on broken balance sheets. Hertz ($620M) has no verifiable robotaxi deal.

Excluded because they are not U.S. companies

Mobileye (Israel, $6.8B) is the most painful exclusion: the best driver-assist pure play on any exchange, guiding to about $2B of revenue. NXP (Netherlands, 14 times profit), Aptiv (Switzerland), Sensata (UK), Hesai (China, five straight profitable quarters and the reason Western lidar prices keep falling), Pony.ai and WeRide (China, barred from U.S. roads), and Einride (Sweden, more revenue than Aurora or Kodiak) are all noted and left out.

Dead

Luminar filed for bankruptcy on December 15, 2025 after Volvo dropped it as a supplier; its laser unit went to Quantum Computing Inc. for $110M and its lidar assets to MicroVision for $33M. Shareholders were wiped out on April 6, 2026. It is the sixth of eight companies in the 2020–21 startup cohort to reach zero (chapter 10).

08The ten picks

Ordered by conviction, with a suggested weight inside a 100% thematic sleeve. Five profitable anchors carry 59%; three growth suppliers carry 29%; two option-value bets carry 12%. Every pick comes with what to watch, what would break the idea, and the single best argument against it.

Bench: first alternates

Ambarella replaces NVIDIA if you want independent self-driving chips at small-company volatility. Qualcomm replaces ON if you want the $65B car-chip pipeline at 19 times profit and can stomach the phone-chip decline. Gentex at 10.8 times profit is the value substitute for Allegro. MP Materials is the magnets chokepoint if you want one. Samsara and Symbotic are the fleet-operations and warehouse layers. DoorDash replaces Amazon for delivery purity at a much higher price.

09How to hold them

The basket is a barbell: heavy, cash-generating anchors on one end, two small speculative bets on the other, and nothing in the middle that can quietly sink you. Nothing losing money gets more than 10%, and nothing without revenue gets more than 4%.

Rules

• Correlation warning. NVIDIA, Alphabet, Amazon, ON and Allegro all trade with the AI-spending cycle. If that cycle breaks, this basket falls with it regardless of what robotaxis do. Treat the sleeve as an AI bet with a self-driving tilt, not as diversification.
• Entry. Buy the anchors in stages on weak days; do not chase Ouster or Aurora after a 20% up-day. Tesla's next decision points are the October 2 delivery report and the October 22 earnings call.
• Rebalance triggers. Trim any position that doubles its target weight. Add to Alphabet if Waymo reports 750k rides a week. Add to Uber if it shows it keeps as much of a robot fare as a human fare. Cut Aurora to zero if the December truck count is under 100.
• Quarterly thesis check. Three questions: Did Waymo's rides grow? Did U.S. electric-car share move off 5.8%? Did any regulator suspend a driverless operation? Two of three going the wrong way means cut the sleeve by a third.
• Distress screen. Any holding with under six quarters of cash at its current burn and no committed financing gets sold before the share sale, not after. That rule alone would have avoided every zero in the 2020–21 cohort.

Scenario map for 2030

10What could go wrong

If this report is wrong, it is wrong for one of the reasons below. Each is a live risk, not a hypothetical.

The theme as a whole

• Unit economics are unproven at scale. Waymo's Chinese-built cars land near $100,000 each after tariffs; Alphabet's Other Bets lose about $7B a year against roughly $600M of Waymo fares. Nobody has shown a profitable robotaxi city. The sub-$20,000 sensor set and the Hyundai assembly line are the fix, and both are 2027 events.
• Growth has stalled at the leader. Waymo has been flat at 500,000 rides a week for six months against a 1M year-end goal, even as the fleet grew. If the limit is permits and car supply, it is fixable. If it is demand at current prices, every number in chapter 03 is too high.
• Tesla's timeline history. Driverless FSD has been "next year" since 2016. Cars with the older HW3 computer, sold as robotaxi-capable, have been conceded not to be. That is a retrofit liability and a lawsuit magnet.
• Regulatory backlash. One child fatality could do to Waymo what the 2023 pedestrian-dragging incident did to GM's Cruise. Massachusetts still bans driverless cars; London's regulator has publicly warned it will refuse licenses; New York allows eight cars with hands on the wheel. Labor politics have not been tested in a city with strong taxi or Teamsters presence.
• Capital intensity. Alphabet, Amazon and NVIDIA's customers are spending hundreds of billions on AI with negative or shrinking free cash flow. A spending pullback re-rates this basket regardless of self-driving progress.
• The U.S. electric-car slump could be permanent. The credit is gone, the fuel-economy mandate is gone, charging money is frozen, and hybrids are winning the mainstream buyer. If the Harvard 32% share estimate is wrong and the bear case (10%) is right, U.S. electric-car revenue is a quarter of the base case and Rivian and Tesla's U.S. volumes do not recover.
• China. BYD, Hesai, Pony and WeRide have lower costs and more profitable lidar and robotaxi businesses than any Western peer. Tariffs protect U.S. revenue but not U.S. valuations; Hesai's pricing sets Ouster's ceiling.

The graveyard: the 2020–21 startup cohort

Eight electric and self-driving companies went public through SPAC shells in 2020–21 at a combined peak value above $65B. Six went to zero. The two positive outcomes both sold before the cash ran out.

What separated the survivors: Velodyne merged while it still had cash and a product; Embark wound down before it ran dry. What killed the rest: no revenue (Embark, Arrival, Canoo), one customer (Luminar and Volvo, Lordstown and Foxconn), convertible debt that came due for a business too small to pay it (Luminar's $625M, Fisker's $667M), and fraud (Nikola). Kodiak and Serve today carry three of those four markers. Aurora carries one (no revenue) and, for now, the cash to survive it.

The best argument against each pick

11Glossary

Every abbreviation and piece of jargon in this report, in one place.

ADAS

Advanced driver-assistance systems: lane keeping, adaptive cruise control, automatic braking. The driver stays responsible (Level 2).

AI

Artificial intelligence: software that learns patterns from data instead of following hand-written rules.

ATM program

At-the-market offering: standing permission to sell new shares into the market over time, diluting existing holders.

ATRI

American Transportation Research Institute: the trucking industry's research arm, source of the cost-per-mile data.

Buyback

A company repurchasing its own shares, which raises each remaining holder's slice.

CAFE

Corporate Average Fuel Economy: the federal miles-per-gallon rule for automakers' fleets.

CAGR

Compound annual growth rate: the steady yearly growth that turns a starting number into an ending number.

Capex

Capital expenditure: spending on long-lived assets such as factories, data centers and fleets.

Cash runway

Cash on hand ÷ quarterly cash burn. Under six quarters means a share sale is coming.

Class 8

The heaviest truck category: 18-wheelers hauling freight between cities.

Convertible notes

Loans that can turn into shares at a set price. Cheap in booms, dangerous in busts.

Design win

A carmaker choosing your chip or part for a model that ships in two to four years.

Drive-by-wire

Steering and brakes controlled electronically rather than mechanically. Required for a computer to drive.

End-to-end model

One neural network from camera pixels to steering, instead of separate see/predict/plan modules.

EV

Electric vehicle.

EV/Sales

Enterprise value (market cap + debt − cash) ÷ revenue. Used when there is no profit to measure.

FCF

Free cash flow: cash from operations minus capital spending. What is actually left over.

FMVSS

Federal Motor Vehicle Safety Standards: the rulebook every U.S. car must meet.

Forward P/E

Share price ÷ next year's expected profit per share. Higher means more growth already priced in.

FSD

Tesla's "Full Self-Driving" software. Supervised (Level 2) on every customer-owned car as of this report.

Geofence

The mapped area a robotaxi may operate in.

GPU / TPU

Graphics processing unit (NVIDIA) and tensor processing unit (Google): the chips that train and run AI models.

Gross bookings

Total fares riders pay. A platform's revenue is only the slice it keeps.

Gross margin

Revenue minus direct production cost, as a share of revenue. Negative means each sale loses money before overhead.

HW3 / HW4

Generations of Tesla's in-car computer. HW3 cars (2019–2023) cannot run the driverless software.

Level 2 / 3 / 4

SAE autonomy levels: human must watch / eyes off on approved roads / no human needed within a defined area.

Lidar

Laser range-finding sensor that builds a 3D map of surroundings. Waymo uses it; Tesla does not.

Market cap

Market capitalization: share price × shares outstanding, i.e. what the market says the company is worth.

NEVI

National Electric Vehicle Infrastructure program: $5B of federal money for highway chargers, frozen in 2025.

NHTSA

National Highway Traffic Safety Administration: the regulator that sets vehicle safety rules and can force recalls.

NTSB

National Transportation Safety Board: investigates crashes and issues recommendations.

OEM

Original equipment manufacturer: in this report, a carmaker.

Part 555 exemption

NHTSA's process for letting a maker sell a limited number of vehicles that skip certain FMVSS rules.

Radar

Radio-wave sensor that measures distance and speed and works in rain, fog and dark.

Robotaxi

A driverless car carrying paying passengers, booked by app.

SAAR

Seasonally adjusted annual rate: the pace of U.S. car sales, about 16 million a year today.

Silicon carbide (SiC)

A chip material that handles high voltage efficiently; used in electric-car inverters.

SPAC

Special-purpose acquisition company: a listed shell that merges with a startup to take it public.

TaaS / DaaS

Aurora's two prices: Transportation-as-a-Service (Aurora owns the truck, ~$2+ per mile) and Driver-as-a-Service (customer owns the truck, ~$0.85 per mile).

Take rate

The share of a fare a platform keeps after paying the driver.

TAM

Total addressable market: yearly revenue available if a product reached everyone who could buy it.

Tier-1

A supplier that sells parts directly to carmakers.

Utilization

How many hours a day a vehicle earns money.

12Sources and method

Research was run September 26–27, 2026 by a five-lane team (sector and market sizing, large-company operators, small- and mid-cap suppliers, freight and delivery plus the electric-car market, and a red team), with a separate SEC-filings pass on the 2020–21 SPAC cohort. Live prices, market caps and analyst estimates come from TradingView via the tvremix connector at the September 25–26 close. Company financials are from SEC 10-Q and 8-K filings and XBRL company facts wherever available; press and analyst figures are used only where filings do not exist. Items the team could not verify from a primary source are marked unverified in the working memos and have been excluded or caveated above. The sizing and red-team lanes were interrupted by an API limit; their chapters were completed from the other lanes' primary data and the SPAC post-mortem.

Primary sources

1. Alphabet Q2 2026 earnings release (SEC EX-99.1); Waymo updates hub and Safety Impact Hub; Waymo $16B Series D (Feb 2, 2026); TechCrunch on the Denver, San Diego and Tampa launches (Sept 1, 2026) and Waymo fleet data (Sept 24, 2026); CNBC on the end of Uber exclusivity (July 24, 2026); NHTSA recall 26E026; Japan Times on the Tokyo launch (Sept 15, 2026).
2. Tesla Q2 2026 10-Q and shareholder update (July 22, 2026); Q2 production and deliveries release; TechCrunch on the Cybercab investigation (Sept 4, 2026); California DMV finding; CNBC on Tesla's suit against the DMV (Feb 23, 2026); The Chargeport robotaxi tracker; CNBC and Electrek on the Semi factory (Sept 24–25, 2026); Seeking Alpha on the CAFE rollback (Sept 26, 2026).
3. Uber Q2 2026 8-K and 10-Q; TechCrunch's Uber AV deal tracker (Aug 1, 2026); Uber release on NVIDIA robotaxis across 28 cities; Nuro–Lucid–Uber program page; Reuters and Dow Jones on the Delivery Hero bid (Sept 24, 2026); Lyft Q2 2026 release; Lyft–Waymo Nashville announcement.
4. NVIDIA FQ2 and FQ1 FY2027 8-Ks (segment change); NVIDIA newsroom on DRIVE Hyperion 10, Halos, Alpamayo, the Uber fleet and the Mercedes CLA; Qualcomm FQ3 FY2026 results and June 24, 2026 investor day.
5. Amazon Q2 2026 release; NHTSA press release and Federal Register notice on the Zoox exemption (July 30–31, 2026); CNBC on Zoox paid rides (Aug 5, 2026); Reuters and Fox Business on Prime Air expansion (Aug 19, 2026); UAV Coach on the FAA Part 108 rule.
6. Aurora Q2 2026 10-Q, shareholder letter and XBRL facts; Aurora 2030 Vision release (Sept 23, 2026); FreightWaves on per-mile pricing; Kodiak Q2 2026 8-K; Serve Robotics Q2 2026 10-Q; TechCrunch on Uber's Serve stake sale (Aug 11, 2026); Reuters on Gatik (Aug 25, 2026); ATRI 2026 Operational Costs of Trucking via Transport Topics.
7. GM Q2 2026 results and eyes-off announcement; Ford Q2 2026 release and 8-K; Rivian Q2 2026 results and deliveries; Lucid Q2 2026 reset release and 8-K; DoorDash Q2 2026 results.
8. Cox Automotive Q2 and August 2026 electric-vehicle data; Kelley Blue Book price reports; Harvard Salata Institute 2030 study (via CleanTechnica, Sept 25, 2026); IEA and Argus on global Q2 sales; Reuters on Europe's 30.5% share; Gasgoo and The Next Web on China; AP on the NEVI freeze; DTN on the EPA appeal.
9. StockAnalysis company and forecast pages for ON, ALGM, OUST, AMBA, GNTX, MOD, IOT, SYM, MP, LFUS, AEVA, QS, CAR, NVTS, WOLF, MBLY, NXPI, HSAI, PONY, WRD, ENRD; Luminar Chapter 11 8-K, plan-effective 8-K and Form 15; Reuters on Volvo dropping Luminar (Nov 17, 2025).
10. SPAC post-mortem: SEC XBRL facts and 8-Ks for Nikola, Lordstown/Nu Ride, Canoo, Fisker, Arrival, Luminar, Velodyne and Embark; Epiq court documents for Nikola; TechCrunch on the Fisker, Canoo and Arrival sales; the Ouster–Velodyne merger 8-K.
11. Market sizing: Goldman Sachs Research on robotaxis ($19B U.S. by 2030, $400B global by 2035); MarketsandMarkets global robotaxi $45.7B by 2030 (June 2026); Oppenheimer on autonomous trucking (via MarketWatch, Aug 25, 2026); ARK Big Ideas 2026 (referenced, not quoted).
12. Live quotes, market caps, ratings, price targets and consensus estimates: TradingView via tvremix MCP, Sept 25–26, 2026.