Waymo, Zoox, Tesla: The Three Business Models Fighting for Driverless Cities

01 One Race, Three Mutually Exclusive Bets

The robotaxi market is often framed as a horse race, but the three leading entrants are not running the same race. Waymo, the Alphabet spinout, is scaling a lidar-and-radar-based system inside carefully mapped geofenced territories. Zoox, owned by Amazon, designed a purpose-built two-directional vehicle with no steering wheel at all, betting that the fleet of the future deserves a chassis designed for it from the ground up. Tesla is wagering that cameras alone, trained on hundreds of thousands of customer cars, can achieve full autonomy and that its existing fleet is a distribution weapon no startup can match. These bets are structurally incompatible: each implies a different cost curve, a different regulatory posture, and a different definition of victory. Understanding the architecture of each bet matters more than tracking weekly headlines.

02 Waymo: The Perfectionist Expansion

Waymo's approach is the most conservative and, by measured results, the most proven. The company operates paid, driverless rides across Phoenix, San Francisco, Los Angeles, and Austin, with additional cities announced or in testing, and has publicly reported scaling past 250,000 paid rides per week as of mid-2026. Its vehicles carry a suite of lidar, radar, and cameras, which costs far more per vehicle than a camera-only setup but produces a sensor-fused model of the road with direct depth measurement. Expansion happens city by city, because each territory requires high-definition mapping and a safety case reviewed by state regulators. The measured track record, drawn from Waymo's own published safety data, shows crash rates per mile meaningfully below human benchmarks in its operating domains. The constraint is capital intensity and the slow, deliberate pace of geographic growth.

03 Zoox: The Ground-Up Gamble

Zoox is the least conventional of the three. Rather than retrofitting an existing car, it built a purpose-built robotaxi: a compact, symmetrical vehicle with no steering wheel, no pedals, and identical front and rear ends, so the vehicle never needs to turn around on a narrow street. The design maximizes cabin utility and lets Zoox fully control the passenger experience, but it also means the vehicle cannot be sold to consumers, cannot share a service network with conventional cars, and requires an entirely dedicated supply chain, factory footprint, and maintenance operation. The economics of that bet are unproven in public data: Zoox has not released per-mile cost figures, and its service, which began carrying members of the public in Las Vegas and testing in San Francisco and the Bay Area, remains small in scale compared with Waymo. Our read is that Zoox is making a long-horizon bet that purpose-built fleets will eventually undercut retrofitted ones on cost per mile, but as of today that remains an estimate, not a demonstrated fact.

04 Tesla: The Camera-Only Mass Fleet

Tesla's wager rests on a single engineering premise: vision is sufficient. Its vehicles carry no lidar and rely on cameras and neural networks to infer depth and scene structure, which keeps per-vehicle hardware dramatically cheaper than Waymo's sensor suite. The strategic payoff is scale. Tesla has built the general capability into hundreds of thousands of cars already on the road, which in principle allows it to flip autonomy on for existing fleets rather than manufacture new ones. Tesla began operating a limited supervised robotaxi pilot in Austin in 2025, subject to a state permit with operational restrictions on geography and conditions. The open question is empirical: whether camera-only perception, in rain, low sun, and unusual geometry, matches the safety profile of systems with direct depth measurement. Regulators, not marketing, will adjudicate that question. Until independent crash data accumulates, the claim remains unproven.

05 The Cost Equation per Mile

The economics reduce to a brutal comparison that current data cannot fully resolve. Waymo's fleet carries expensive sensors, but its utilization is high, its rides are fully paid, and Alphabet has absorbed the capital cost. Tesla's per-vehicle hardware cost is far lower, but camera-only autonomy may require greater remote supervision and software maturity before it can run unsupervised at commercial density, which eats into the saving. Zoox's purpose-built vehicle promises the lowest theoretical long-run cost per mile at maximum fleet utilization, but it has not yet published the operating data to prove the curve, and its capital burden is entirely front-loaded. As an illustrative order of magnitude, retrofit lidar suites historically added tens of thousands of dollars per vehicle, though costs have been falling; the precise per-mile cost of any operator in 2026 is a closely held number. This is an area where confident public figures deserve suspicion.

06 Regulators Are the Real Referee

Regulation will decide this race as much as engineering will. Waymo's city-by-city expansion is effectively co-designed with state regulators, which is slow but builds a durable license to operate. Tesla operates its Austin pilot under a Texas permit with defined operating boundaries, and its national rollout depends on federal and state authorities accepting camera-only autonomy without lidar redundancy, something no US regulator has yet endorsed at commercial scale. Zoox faces a different hurdle entirely: a vehicle with no manual controls requires an exemption pathway from federal motor vehicle safety standards, since those standards presuppose a human driver able to take over. Each company has chosen a regulatory mountain to climb, and the timelines attached to those mountains, not technology demos, will determine which fleets actually reach the streets first.

07 Verdict: Three Ways to Win, One Way to Lose

On measured evidence, Waymo leads today: real paid rides at scale, a published safety record, and steady, if slow, geographic growth. Tesla has the most valuable endgame if its camera thesis holds, because a fleet already in customer hands erases the manufacturing ramp that constrains everyone else. Zoox is the highest-variance bet, with a plausible long-run cost advantage that remains almost entirely unproven in public data. The one strategy that looks wrong is the middle path: partial autonomy that still needs a human behind the wheel, which competitors and regulators have pushed to the margins. What would falsify each bet is clear. If camera-only perception plateaus below commercial safety thresholds, Tesla pivots to a smaller supervised fleet and its advantage evaporates. If lidar costs keep falling and Waymo's expansion stays disciplined, the perfectionists win by default. And if purpose-built economics never materialize for Zoox, Amazon will have funded the industry's most beautiful science project.