Autonomous Driving Weekly Analysis — 2026-10-23: Waymo and Tesla announced expansion schedules while permitting and certification processes remain unfinished

Briefings covered: 2026-10-23 news briefing

Frame: Four value-added layers in the robotaxi value chain — AV software and data, and demand platforms (high margin) / vehicle manufacturing (medium) / operations and maintenance (low). Where this week's evidence conflicts with the frame, the evidence takes priority.

This Week's Key Points

The expansion plans the two leading operators announced this week are both schedules set before permits or certification were finalized. Waymo announced entries into Toronto (10-16) and Madrid (10-20), but both cities start with manual-driving tests, and the targets for public rides are 2028 and 2027 respectively. Ontario's pilot regulations expire in October 2027, and in Spain driverless paid service requires a separate passenger transport authorization. At its Q3 earnings call (10-21), Tesla stated targets for Cybercab service in Dallas in November and customer deliveries within the year, but NHTSA's query (AQ26002) on the basis for self-certifying federal safety standards for a vehicle without brake pedals or side mirrors, with an answer deadline of 10-30, and the review of the school-zone defect petition (DP26010) have not been concluded.

Waymo's 20,000-vehicle plan comes from a single Semafor report, and the company declined to comment. According to the report, the fleet would grow from about 4,000 vehicles now to 8,000 by end-2026 and 20,000 by end-2027, using the US$5 billion term loan raised in October. A vehicle count target is not the same as the number of vehicles in driverless operation, and the allocation by city has not been disclosed. In the same week, pickups at Las Vegas Harry Reid Airport began, bringing Waymo's airport service to six locations.

Capital went into the operations layer and low-speed delivery vehicles. Hertz spun off its robotaxi fleet management subsidiary Oro Mobility and raised US$80 million, with the investor group receiving 50% of voting rights and about 15.9% of economic interest. Chinese driverless delivery vehicle maker Neolix raised about US$400 million in a Series E. Meanwhile, no new NHTSA investigation, robotaxi recall, permit suspension or fatal or injury accident was identified during the period.

In Korea, RideFlux was selected as the operator of the Level 4 taxi service in Sangam, Seoul, and is preparing to launch at the end of November. The initial fleet is two vehicles, and while the driver's seat is empty, a safety manager rides in the front passenger seat. Read together with the Seoul Level 4 city bus agreement covered last week, commercialization in Korea is starting from public procurement and small-scale pilots, and MOLIT's final driverless permit is the reference point for the schedule.

1. Industry Implications

(1) Waymo's overseas entries are at the announcement stage, and their schedules are set by each country's paid-service authorization system

Both Toronto and Madrid start with manual-driving tests with safety operators on board. Toronto received approval to join Ontario's AV pilot program, and vehicle operations will be handled by Element Fleet Management. The pilot regulations expire in October 2027 while the public ride target is 2028, so without an extension of the regulations or a new legal basis there is no path to paid service (Electrek). Madrid starts testing with Jaguar I-PACE vehicles in the city, Alcobendas and Getafe, and fares and fleet size have not been disclosed.

In Madrid, WeRide–Uber–AVOMO received Spain's first L4 operating permit earlier and was running 20 test vehicles as of September. Since U.S. and Chinese operators will go through the same authorization process in the same city, how Spanish authorities' driverless authorization conditions work for each side will be the first case for gauging the cost of entering the European market. That the bottleneck for overseas entry is authorization rather than technology is close to established fact, and that authorization schedules set the pace of expansion is a reasonable inference.

Las Vegas airport pickups were initially opened only to some existing riders, and Zoox is already operating at the same airport. Waymo, Zoox and Motional via Uber (paid service with safety operators) operate together in Las Vegas, making it an unusual market where share by operator can be compared within one city. Airport trips carry higher fares and directly affect revenue per vehicle.

(2) Tesla set an external sales schedule for Cybercab before the federal certification query was concluded

Dallas service in November and customer deliveries within the year were announced before the AQ26002 answer deadline (10-30) and the DP26010 decision. Tesla said current production is going into its own fleet. Musk's statements that cumulative unsupervised mileage rose 8 times quarter on quarter and that the Austin Cybercab fleet grew 10 times in about a month since launch did not disclose base fleet counts or absolute figures, and they conflict with last week's company social media figure of "8 times in 30 days." The "official absolute figures in earnings materials" that last week's analysis listed as a watch point did not appear in this earnings release either. Musk said robotaxi revenue will not be reflected meaningfully in the financial statements before Q2 2027.

If external sales materialize, part of the risk of vehicle ownership and of the related regulatory outcome moves from the manufacturer to the fleet operators that buy the vehicles. If the certification query is concluded after sales, that risk stays with the buyers. This is an inference made while the terms of sale are undisclosed, and it needs to be rechecked once the contract structure is published.

(3) In the split robotaxi model, the fleet operations layer raised independent capital

Hertz raised US$80 million for Oro Mobility, with the investor group receiving about 15.9% of economic interest along with 50% of voting rights and three of six board seats. Oro holds Uber's fleet management contract for the Lucid–Nuro robotaxis and targets the start of AV operations by end-2026, beginning in SF (per the company). It currently leases about 45,000 vehicles to Uber and Lyft drivers. In a structure where the platform (Uber), vehicle (Lucid), AV stack (Nuro) and fleet operations (Oro) are separate companies, this is a case of the operations layer attracting separate capital, following Moove and AVOMO into the same layer.

In the four-layer frame, operations and maintenance is the lowest-margin layer. The valuation of this deal was not disclosed and there is no margin data, so it is not grounds to change the frame. However, the fact that the investors secured voting rights larger than their economic interest suggests that investors in this layer may value control over fleets and access to vehicle supply contracts more than returns (inference). It connects to (2) in that demand for the same layer will grow if Tesla's external sales materialize.

In the vehicle manufacturing layer, if Waymo's 20,000-vehicle plan is accurate, vehicle procurement (Magna's Mesa plant, the Ojai based on a Zeekr body, the Hyundai IONIQ 5) and sensor and compute demand would need to grow 2.5 times within a year from end-2026. Because borrowed funds are being used to expand the fleet, a vertically integrated operator also carries the capital intensity of the vehicle and operations layers directly. Zoox's lease of a 350,000-square-foot site in Hayward could be read as an expansion of production capacity, but its use has not been disclosed and it comes from a single source.

(4) Chinese autonomous driving capital went into low-speed driverless delivery vehicles rather than robotaxis

Neolix's Series E of about US$400 million is the largest single round in Chinese autonomous driving this year. It was led by Boyu Capital (reports say a state-level fund co-led; fund name unverified), with new participation from Alibaba's AEF NextGen, Granite Asia, Vertex and others. Cumulative funding is more than US$1.4 billion, and the valuation was not disclosed. The company said it will use the proceeds to build overseas bases. "More than 30,000 driverless vehicles operating in about 300 cities in 20 countries" is a claim by the company with no external verification.

During the same period, Baidu, Pony.ai and WeRide made no announcements of new cities or driverless transitions. Overseas expansion by Chinese companies is also proceeding through low-speed delivery vehicles, separately from robotaxis, and the entry of Chinese driverless delivery vehicles into last-mile markets in Southeast Asia and the Middle East could accelerate (inference). Delivery vehicles are sold with the vehicle and operations bundled by one company, so the four-layer distinction is less clear than in robotaxis.

(5) Korea's first taxi with an empty driver's seat starts with two vehicles and a safety manager in the front passenger seat

The Seoul Metropolitan Government selected RideFlux as operator of the Level 4 taxi in Sangam, which is preparing to launch at the end of November, with detailed contract terms still under negotiation. The initial vehicles are two Genesis GV80s, and a safety manager rides in the front passenger seat. The schedule for the safety review for the final driverless permit is being coordinated with MOLIT (Electronic Times). It is Korea's first case with an empty driver's seat, but it differs from full driverless operation by overseas standards.

With additional permits in October, Autonomous A2Z reached a cumulative 101 MOLIT temporary operation permits (14 of its own Level 4 ROii shuttles and 87 retrofitted vehicles, per the company). The number of permits is not the same as the number of vehicles operating at the same time. Read together with last week's Level 4 city bus agreement (target of 500 buses by 2030), demand in Korea is coming through public procurement and small-scale pilot operations. Valuation of Korean operators depends more on the timing of MOLIT's final driverless permit than on vehicle counts, and RideFlux's KOSDAQ listing review (filed in August) is proceeding at the same time as permit progress.

2. What to Watch

• Tesla's submission of its AQ26002 answer (deadline 10-30) and the form of the November Dallas launch — Confirm through NHTSA's public correspondence and the operating conditions on launch day (whether safety operators are on board). If the answer is not filed by the deadline or additional measures follow, the likelihood of the external sales schedule within the year should be viewed as lower; if launch occurs without safety operators, it will be the first case of commercial operation of a vehicle without a steering wheel beginning ahead of the certification outcome.
• Company confirmation of Waymo's 20,000-vehicle plan — Confirm through an official company announcement or the parent company's quarterly earnings materials, and the output pace of Magna's Mesa plant. If confirmed, 2026–2027 demand estimates for the vehicle, sensor and compute supply chain should be raised; if not, it remains a single-report figure. Utilization per vehicle against the target of 1 million weekly paid rides by year-end (per the company) is the indicator to watch alongside it.
• Whether Ontario's pilot regulations (expiring October 2027) are extended, and the timeline for Spanish authorities' driverless passenger transport authorization — Confirm through provincial regulatory amendment notices and announcements by Spanish transport authorities. If an authorization path is set, the Toronto and Madrid schedules can be viewed as execution plans; if not, they continue to be classified as announcements.
• Launch of Lucid–Nuro vehicles in SF within the year and the start of Oro's AV operations (end-2026 target, per the company) — Confirm through Uber and Oro announcements. This is the first data showing whether the fleet operations layer in the split model starts generating actual revenue.
• Pony.ai Q3 results (11-18) — Confirm through disclosure of the Chinese robotaxi company's fleet size, driverless share and overseas revenue. This allows judging whether this week's shift of Chinese capital toward delivery vehicles is connected to weaker robotaxi results.
• The end-November launch of the Sangam Level 4 taxi and the schedule for MOLIT's final driverless permit — Confirm through Seoul Metropolitan Government announcements and MOLIT notices. If a schedule for a driverless permit without a front-passenger safety manager emerges, the judgment on the timing of commercialization in Korea can be moved forward.
• Whether DP26010 is accepted (≈2027-01-21) — Confirm through NHTSA's decision document. If accepted, it will be the first robotaxi case in which traffic law compliance becomes the subject of a formal investigation, and the burden of proof on the operations layer will grow.

3. Data Reliability Notes

• Citable: Waymo's Toronto and Madrid entry announcements and target years for public rides (per the company, multiple outlets), start of Las Vegas airport pickups (10-21, airport authority and multiple outlets), Tesla's November Dallas service and within-year delivery targets (earnings call statements, marked as company targets), AQ26002 answer deadline of 10-30, Oro Mobility's US$80 million and its equity and voting structure (Hertz press release and multiple outlets), selection of the Sangam Level 4 taxi operator and initial two vehicles (Seoul Metropolitan Government, multiple outlets), A2Z's cumulative 101 temporary operation permits (marked as per the company).
• Use with caution: Waymo's 20,000 vehicles by end-2027 and 8,000 by end-2026 (Semafor single source, unverified by the company), Tesla's 8x mileage and 10x fleet (base figures undisclosed, conflicting with last week's company social media figures), Neolix's funding amount ("US$400 million" and "about US$400 million" both reported) and the name of the co-leading state-level fund (unverified), Neolix's 30,000 vehicles in operation (company claim), use of Zoox's Hayward site (single source, unverified), the undetermined legal path to paid service in Ontario (Electrek report).

*This article is industry analysis based on public information and is not a recommendation to buy or sell any security.*