This group sells driving without a human driver. The product is a car or truck that moves itself — perceiving the road with cameras, radar and laser scanners (LiDAR, short for Light Detection and Ranging, which builds a 3D map of the surroundings from laser pulses) and deciding what to do with onboard AI. The makers split into three camps: robotaxi operators who run their own driverless ride-hail fleets (Alphabet's Waymo, and Tesla), autonomous-trucking developers hauling freight on highways (Aurora), and component/software suppliers who sell the "self-driving brain and eyes" to traditional carmakers (Mobileye). What the world needs it for is simple: per the scan, the US spends roughly $2 trillion a year on transportation, and the single biggest cost in trucking and ride-hail is the human driver. Remove the driver and that cost largely disappears, which is the core economic prize.
Per the scan, potential demand for driverless miles is very large versus today's near-zero supply, so the product is structurally "short" (demand exceeds available supply) for years. What limits supply here is NOT factories or parts — vehicles are straightforward to build — it is software intelligence: solving the rare, unusual "long-tail" driving situations safely, plus regulatory permission. In money terms, the leaders are valued largely on the promise of that future, not on today's small driverless revenue, so an owner buying today is paying a large amount of market value per dollar of current cash from this product. The reader should judge whether that trade is worth it.
The product is autonomous driving capability, and it is sold in a few distinct ways, which matters because each turns into cash differently:
The key plain-English point for an owner: today almost none of these earn meaningful profit from the autonomous product itself. Robotaxi fleets are early and cash-hungry (heavy capex — capital expenditure, the cash spent on long-lived assets like vehicles and depots — and operating losses); trucking is pre-large-scale; the supplier (Mobileye) is the one that already books real revenue, but mostly from older driver-assist chips, not full autonomy. The cash from true autonomy is still ahead, not behind.
Source: 500-stocks scan, Robotics & Automation module, "Autonomous Vehicles & Self-Driving" section (/Users/ravf/projects/work/.claude/worktrees/sector-hub/research/investments/500-stocks/08-robotics-automation.html).
Today (real, exists now): demand for actually-driverless miles is small but growing and is, crucially, already proven to exist — people pay for Waymo robotaxi rides in several US cities, and freight customers are piloting driverless trucks. The scan frames the addressable spend as the ~$2 trillion/year the US puts into transportation, where the driver is the largest single cost in ride-hail and trucking. real demand exists
Forward (forecast): the demand case is that almost every paid mile a human drives today is a mile that could be served driverless once the software is good enough and legal. That is a very large pool — US ride-hail and trucking labor alone is on the order of hundreds of billions of dollars a year. est. Commonly-cited industry projections put the global robotaxi/autonomous-driving market somewhere in the high tens to low hundreds of billions of dollars by the early 2030s, growing at compound annual rates often quoted in a roughly 20-40% range. est. Treat these as directional, not precise. forecast
Why demand could rise under an AGI lens: the thing that has blocked this for a decade is the "long tail" — the rare, novel situations a fixed rule-set can't handle. A genuinely general, human-level (or better) driving intelligence is what would collapse that long tail: it can reason about unseen situations the way a person does, but without fatigue, distraction, or wage cost. Given AGI is arriving, the realistic question shifts from "will full autonomy ever work?" to "how fast does it roll out and who is allowed to operate it." That tends to pull demand forward: once the capability is trusted, switching from a paid human driver to a near-zero-marginal-cost machine becomes a straightforward cost decision for fleet owners. The same intelligence that powers the broader AI build-out is the input that unlocks this physical-AI market. This is a forecast, not a contracted outcome. forecast
Who the buyers are: (1) riders and freight shippers who ultimately pay per mile; (2) fleet/ride-hail operators and logistics companies who buy or license the capability to cut labor cost; (3) automakers who buy the chips/software to put autonomy in cars they sell. The demand is not one market — it is consumer mobility, freight, and OEM (original equipment manufacturer, i.e. the carmaker that builds the vehicle) supply stacked together.
The headline fact (per scan): this product is NOT supply-constrained in the usual sense. Unlike data-center power or advanced chips, you do not run out of raw vehicles, sensors, or factory capacity — those can be built. The scan states it directly: "Not supply constrained. The bottleneck has always been software intelligence, not vehicle hardware." So "supply" here means supply of trustworthy, approved autonomous capability, and that is genuinely scarce today.
The real limiters are:
Market-share structure (who controls supply): the scan flags "winner-take-most" dynamics — whoever solves the AI problem first and at scale captures an outsized share, because a better-driving model compounds (more miles → more data → safer model → more permission → more miles). Today supply of real driverless service is concentrated: Waymo leads mature L4 robotaxi; Tesla leads on fleet/data scale and is rolling out unsupervised self-driving in select markets; Aurora leads US autonomous trucking; Mobileye supplies driver-assist chips/software broadly across OEMs; GM holds Cruise as its AV (autonomous vehicle) unit. per scan
Source: 500-stocks scan, "Autonomous Vehicles & Self-Driving" — Supply/Demand Dynamics and Key Companies (/Users/ravf/projects/work/.claude/worktrees/sector-hub/research/investments/500-stocks/08-robotics-automation.html).
Put the two sides together and the product is, on the scan's framing, short, not long (potential demand far exceeds available supply) for the foreseeable future: potential demand is "almost every paid mile," while supply of trusted, approved driverless miles is today a tiny fraction of that. The shortage is unusual because it is a capability shortage (software + permission), not a physical-capacity shortage — so the usual signs of a tight market (sold-out factories, rising part prices) don't apply. Instead, the evidence that demand exceeds supply is qualitative: limited robotaxi service areas and waitlists, city-by-city rollout, and the fact that operators are racing to expand rather than discount.
| Dimension | Demand side | Supply side |
|---|---|---|
| Size of pool | ~$2T/yr US transport; near-universal paid miles | Driverless miles tiny today, scaling |
| Binding constraint | Trust + price vs human driver | Software intelligence, then regulation |
| Direction (AGI lens) | Pulled forward — long tail collapses | Unlocks fast once software is solved |
| Current state | Proven, far exceeds supply | Scarce, concentrated in a few players |
When could it flip to oversupply? Plausibly only after the software problem is broadly solved (so multiple players can each offer "good enough" autonomy) AND regulators open many markets AND fleets are built out. At that point the scarce thing stops being capability and becomes ordinary fleet capacity, and per-mile pricing could fall toward the cost of running vehicles — a normal, competitive transport-services market. The winner-take-most framing in the scan implies that even then, share may stay concentrated rather than fragmenting into a price war. This is a forecast, not a contracted outcome. forecast
Source: 500-stocks scan section (same file as above) plus neutral arithmetic on its figures.
Each name has very different exposure to this specific product. The descriptions below are rough framing of how much of the company's value/revenue is tied to autonomous driving versus everything else; market caps are approximate and not live-verified. est.
| Company | What it makes here | Exposure to this product | Rough size | Position / edge |
|---|---|---|---|---|
| Tesla (TSLA) | Self-driving software (FSD — "Full Self-Driving," its driver-assist/autonomy package) on its own cars; building robotaxi | Diversified — cars/energy today; autonomy is the large future slice of the story, small share of today's revenue | Roughly several-hundred-$B to ~$1T-class cap est. | Largest real-world data fleet (6M+ vehicles, per scan); vertically integrated |
| Alphabet (GOOGL) / Waymo | L4 robotaxi service (own fleet, paid rides) | Tiny slice of a giant — Waymo is a very small share of Alphabet's revenue today | Roughly ~$2T-class cap (mostly Search/Cloud/Ads) est. | Most mature driverless L4; deepest robotaxi operating experience |
| Aurora (AUR) | Autonomous trucking (driverless highway freight) | Pure-play — essentially all of the business is this product, early-stage revenue | Small/mid-cap (roughly single-digit $B) est. | Leading US autonomous-trucking developer (per scan); freight focus |
| Mobileye (MBLY) | ADAS chips + self-driving software sold to automakers | Pure-play on driving tech, but mostly today's driver-assist (ADAS — Advanced Driver-Assistance Systems, features like lane-keeping and automatic braking), not full autonomy yet | Mid-cap (roughly low-to-mid tens of $B) est. | Supplies many major OEMs; real, shipping revenue today |
| GM (Cruise) | Robotaxi unit inside a legacy automaker | Small slice — Cruise is a minor, loss-making piece of GM's car business | Large-cap automaker (roughly tens of $B) est. | Owns an AV unit but trails Waymo/Tesla in deployment |
Plain framing for an owner: only Aurora and Mobileye are anything close to pure plays on this product (a "pure play" = a company whose business is almost entirely this one thing). Tesla is a partial play where autonomy is the largest piece of the future narrative but a small piece of today's cash. Alphabet and GM give exposure to the product as a tiny fraction of much larger businesses, so buying them is mostly buying those other businesses rather than autonomous driving. The reader can weigh which exposure they want.
Source: 500-stocks scan Key Companies section (same file); sizes are general-knowledge approximations, not live-verified.
Here is the money-in/money-out shape, in plain terms, because it differs sharply across the group:
Capex shape of the group: robotaxi operation is capital-heavy (you must own/maintain fleets, depots, charging, mapping, remote support). The chip-supplier model (Mobileye) is comparatively capital-lighter — it sells a part rather than running a fleet. Today, the only members reliably generating owner cash (free cash the business throws off after its own spending) are the diversified giants (Alphabet from Search/Cloud, GM from cars) — and that cash comes from their other businesses, not from autonomy. The pure-play autonomy names are net cash consumers for now. So an owner is, in effect, funding the build-out of supply and being paid back (if at all) only later, once the demand-supply gap is converted into paid miles. Neutral arithmetic, no verdict: high promise, low current cash, with the price of the pure plays set by belief about how fast and how completely the gap closes. est.
Source: scan-derived facts on business models + general-knowledge framing of the valuation shape (not live-verified). Re-pull live market caps, revenue, and cash burn before any decision.
Where a company-level deep-dive would be most informative, by type:
Practical reading: to study the product itself, dig into AUR and Waymo (inside GOOGL) for operating reality and MBLY for real financials; to study who captures it at scale, compare TSLA's data edge against Waymo's operating maturity. This is a pointer to where the information is, not a recommendation.
What was used:
Hard vs approximate:
Source: file above + general knowledge (cutoff ~early 2026, not live-verified).