Autonomous Driving Weekly Analysis — 2026-11-06: Uber set driverless vehicle targets, Waymo decided to end Uber exclusivity in some cities, and Waymo and Kodiak moved to vehicles built at OEM factories

Briefings covered: 2026-11-06 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

Uber set targets of about 700 fully driverless vehicles on its app by year-end and more than 2,000 in 2027. These figures were given by the CEO in the third-quarter results (11-04) and are all company targets. For 2028 and beyond, the CEO said "tens of thousands" and the CFO said "more than 10,000", which is conflicting. The CFO said long-term vehicle procurement will be funded with third-party capital rather than on Uber's balance sheet. In the same quarter gross bookings were US$59.8 billion and trips were 4.1 billion; the AV share of total trips was not disclosed. About 700 vehicles by year-end is small relative to this scale.

Waymo confirmed it will end Uber exclusivity in Austin and Atlanta in January 2028 and switch to its own app. Waymo's co-CEO cited the fact that riders on the Uber app cannot tell whether they have been assigned a driverless vehicle (per the company), and Uber said Waymo had rejected an option for booking robotaxis directly within the Uber app. The number of Waymo vehicles and rides through Uber in the two cities has not been disclosed. It is a reasonable inference that, within the driverless vehicle targets Uber set the same day, the share of partners other than Waymo — Zoox (Las Vegas, launch targeted for later this month, per the company), Motional, Nuro and others — will grow.

Installing AV hardware at OEM factories appeared in both passenger robotaxis and trucks in the same week. Waymo and Toyota announced that a robotaxi whose bZ4X assembly and Waymo hardware installation are done on one line at the Takaoka plant will be tested in Japan in 2027 (10-30). Kodiak said it delivered the first two Western Star 49X driverless trucks built by Daimler Truck North America to Atlas (11-03). Neither case disclosed production volumes or unit prices.

On the regulatory side, the US and Japan showed different approaches to vehicles without steering wheels or pedals. Tesla submitted its AQ26002 answer by the deadline and asked that all documents for Q8 — which asks which FMVSS it judged it met and how — be kept confidential (single-outlet report; content of the original in the NHTSA docket unverified). Japan's MLIT was reported to be planning to apply, from December, a common standard for certifying approved Level 4 and 5 vehicles without steering wheels or pedals, but the official notice is unverified. No new NHTSA investigations, robotaxi recalls, permit suspensions or fatal or injury crashes were identified in the period covered.

Among last week's watch items, AQ26002 was confirmed only as far as the submission of the answer, and NHTSA's assessment has still not been issued. Kodiak's December I-45 driverless launch, vehicle numbers in Waymo's expanded areas, the DC ordinance vote and the launch of the Sangam Level 4 taxi (scheduled for late November) were not resolved this week. Last week's view that vehicle production sets the timing of truck AV expansion was reinforced by Kodiak's delivery of OEM factory-built vehicles, and no view needs correcting.

1. Industry Implications

(1) This week produced evidence that demand platform margins can shrink depending on operator scale

Waymo's switch to its own app was confirmed as a move by an AV operator that has grown in scale to take the rider relationship directly. The four-layer frame places demand platforms alongside AV software in the high-margin layer, but that position rests on the premise that operators cannot generate demand on their own. Waymo cited its inability to tell riders that a driverless vehicle had been assigned, which means a platform structure that dispatches without distinguishing vehicle suppliers does not fit the interest of an operator trying to build its brand. Uber said it had offered an in-app direct booking option, indicating that the platform side has also moved toward conceding terms.

This evidence does not change the frame, but it adds a qualification. The interpretation that fits this week's evidence is that a platform's negotiating power weakens in its relationship with large operators able to generate their own demand, while it keeps its position as the supplier of demand in relationships with operators that have few vehicles and no app of their own (the partners Uber named, such as Zoox, Motional and Nuro). The switch takes place in January 2028 and the existing contract expires in May 2028 (earlier FT report), so no actual revenue effect has appeared yet.

(2) Uber said it will grow its fleet with third-party capital rather than owning vehicles, bringing vehicle financing into view as a separate layer

The CFO said long-term vehicle procurement will be funded with third-party capital. If a demand platform keeps vehicle depreciation off its balance sheet, the asset burden passes to leasing and finance companies, and the platform maintains a structure with low capital requirements relative to revenue. This points the same way as last week's Kodiak–DTL contract (the carrier owns the trucks and the AV company receives a per-mile license fee): in both passenger cars and trucks, the parties avoiding asset ownership are in the software and platform layers, and vehicle capital is carried by other parties. Financing terms and counterparties were not disclosed, so the profitability of this layer cannot be estimated.

In terms of scale, about 700 vehicles by year-end and more than 2,000 in 2027 are all company targets. Even the 2027 target is a small share relative to 4.1 billion trips per quarter. Uber's third-quarter GAAP net income of US$512 million reflected a US$476 million (pre-tax) loss on the revaluation of equity stakes; the briefing did not identify which stakes the loss related to.

(3) The Waymo–Toyota and Kodiak–Daimler Truck cases enlarged the role of the vehicle manufacturing layer

In both cases AV hardware is installed at the OEM factory stage rather than through a retrofit process. Toyota joins Zeekr and Hyundai Motor as a Waymo vehicle supplier, and the bZ4X is for the right-hand-drive Japanese market. The Japan program is at the testing and validation stage in 2027, and the UK schedule and production volumes have not been disclosed. Kodiak plans to fill 65 of Atlas's initial 100-truck order with Western Stars by mid-2027 (per the company); Atlas has 37 driverless trucks and expects about 45 by year-end.

Factory installation reduces retrofit cost and ties production pace to OEM lines. In that case OEMs gain weight in expansion speed and unit-price negotiations, so it is a reasonable inference that, even within a frame that treats vehicle manufacturing as a medium-margin layer, OEMs hold a stronger position than in the retrofit era. Taken together with last week's Aurora production delay, the bottleneck for expansion schedules in both trucks and passenger cars lies in production execution outside the AV companies.

(4) Certification paths for vehicles without steering wheels are proceeding through confidential individual audits in the US and a common standard in Japan

With Tesla requesting confidentiality for all of AQ26002 Q8, the basis on which Cybercab meets FMVSS remains undisclosed. NHTSA has not said whether it will accept the confidentiality request, has not found any violation of standards, and has issued no order to halt service. This is a single-outlet report, and the content of the original in the docket is unverified. If the basis is not disclosed, the standards other OEMs preparing vehicles of the same design could refer to will also arrive later. Tesla's November Cybercab launch in Dallas and external sales within the year are company targets and depend on NHTSA's assessment. The decision deadline for the school-zone petition (DP26010) is around 2027-01-21.

Japan's MLIT plan would certify approved Level 4 and 5 vehicles under a common standard without steering wheels, pedals, rear-view mirrors or driver-monitoring devices; driverless freight vehicles would be exempt from crash tests while taxis and buses would keep them (citing a Nikkei report; official notice unverified). The government target is about 10,000 Level 4 buses, taxis and other vehicles by 2030. Vehicle certification and permission for driverless road service are separate, so even if this plan takes effect, operating-permit schedules need to be watched separately. Because it coincides with the Waymo–Toyota 2027 Japan test, the predictability of certification procedures becomes a factor when companies building purpose-built vehicles choose their first production market.

In Korea, the Citizens' United for Consumer Sovereignty filed a complaint with the KFTC alleging that Tesla Korea's FSD advertising violated the Act on Fair Labeling and Advertising and the E-Commerce Act (11-05; whether it was formally accepted is unverified), and on 10-29 the Seoul Central District Court dismissed a claim by 88 owners who bought the FSD option for a full refund. The share of Tesla vehicles registered in Korea that cannot use FSD is conflicting at 76.2% and 77.6% (different reference dates). Tesla Korea submitted an application for technology special-case approval to MOLIT on 10-22. The advertising wording of driver-assistance features is being treated as a consumer regulation issue, and the scope of domestic availability will be set by the special-case review. No new pilot zone designations or new temporary operating permits from MOLIT were identified in the period covered.

(5) The top two US truck AV companies use the same Chinese lidar, so if sensor regulation widens the costs arrive at once

Kodiak confirmed it uses Hesai lidar on its commercial trucks, and Aurora confirmed it uses Hesai products alongside its own lidar. Hesai is on the US Department of Defense 1260H list, but this designation only bars Defense Department contracts and does not directly prohibit civilian use. Hesai denies military ties and is in litigation, and the course of rulings on its designation status is unverified. The current BIS connected-vehicle rule does not cover lidar, and a bill to ban Chinese lidar is pending in Congress.

If regulation extends to lidar, both companies would face sensor replacement costs and revalidation schedules, which would overlap with the production bottleneck in (3) above. Kodiak explained that it designed its system so that Hesai cannot access sensor data, but if the bill is written on a country-of-origin basis, data-access design may not count as grounds for exemption (speculation). US and Israeli lidar makers could see replacement demand, but at present it has not been determined whether the bill will pass.

In truck AV capital, Humble Robotics raised a US$51 million Series A (US$75 million cumulative) for a cabless electric truck (single source). It targets port drayage and short-haul transport between distribution centers; customers have not been disclosed, and pilot operations are targeted within the next 12 months (per the company). As with last week's Bedrock Robotics, early capital is going into closed and semi-closed routes with a lower permitting burden.

2. What to Watch

• The actual year-end number of fully driverless vehicles on the Uber app and the Zoox Las Vegas launch — Check via Uber's fourth-quarter results and Zoox and Uber announcements. If the number comes close to about 700 by year-end, the platform strategy of widening driverless supply beyond Waymo can be seen as being executed; if it falls well short, the platform's dependence on Waymo for driverless supply should be judged higher.
• The structure and counterparties of the third-party capital for Uber's vehicle procurement — Check via Uber filings and leasing or finance company announcements. If the terms are disclosed, the return earned by the vehicle financing layer can be placed on the four-layer frame, and it can be judged whether the platform's asset burden actually stays off its balance sheet.
• NHTSA's decision on the AQ26002 Q8 confidentiality request and the form of the November Dallas Cybercab launch — Check via the NHTSA docket and the operating conditions on launch day (whether a safety operator is on board). If confidentiality is granted and no further action follows, a practice may emerge of keeping the certification basis for vehicles without steering wheels confidential company by company; if further action follows, the view on the external sales schedule for such vehicles and on the FMVSS exemption path should be pushed back.
• MLIT's notice of a common certification standard for Level 4 and 5 vehicles — Check via the original MLIT notice in December and the vehicle types covered. If it is issued as reported, companies building purpose-built vehicles gain grounds to choose Japan as their first production certification market, and how the crash-test requirements for taxis and buses are set will determine robotaxi vehicle design costs.
• Committee progress on the Chinese lidar ban bill and revisions to the BIS connected-vehicle rule — Check via congressional bill records and BIS Federal Register notices. If it advances, sensor replacement costs and schedules for US truck AV need to be built into expansion plans; if not, the current supply structure will hold for the time being.
• Driverless truck numbers in Kodiak's third-quarter results and the December I-45 driverless launch, and Pony.ai's third-quarter results (scheduled for 11-18) — Check via each company's results materials and announcements. If Kodiak grows its fleet with OEM factory-built vehicles and starts driverless freight on public roads, the bottleneck for truck AV shifts from production to operational expansion. For Pony.ai, overseas vehicle numbers and the driverless share will show whether MOUs are turning into actual vehicles.
• The launch of the Sangam Level 4 taxi and whether the KFTC opens an investigation into FSD advertising — Check via Seoul Metropolitan Government announcements, the outcome of MOLIT's special-case review and KFTC announcements. The Sangam taxi is scheduled to start with a safety manager in the front passenger seat initially, and once launched it will provide the first data on commercial operation in Korea with no one in the driver's seat.

3. Data Reliability Notes

• Citable: Uber third-quarter gross bookings of US$59.8 billion, revenue of US$14.8 billion, 4.1 billion trips, a US$476 million equity revaluation loss and GAAP net income of US$512 million (SEC 8-K); Waymo's switch to its own app in Austin and Atlanta in January 2028 (co-CEO interview, multiple outlets); Waymo–Toyota bZ4X installation at the Takaoka plant and 2027 Japan testing (company blog, multiple outlets); Kodiak's delivery of the first two Western Star 49X trucks and Atlas's 37 trucks (press release, multiple outlets); Kodiak's and Aurora's use of Hesai lidar (confirmed by both companies, multiple outlets); the Seoul Central District Court's 10-29 dismissal of the FSD refund claim (multiple outlets).
• Use with caution: Uber's about 700 driverless vehicles by year-end and more than 2,000 in 2027 (company targets); "tens of thousands" vs "more than 10,000" for 2028 and beyond (conflicting); Zoox Las Vegas launch later this month and Nuro in early 2027 (per the company); expiry of the existing Waymo–Uber contract in May 2028 (earlier FT report); Tesla's request to keep all of AQ26002 Q8 confidential (single source; content of the original in the NHTSA docket unverified); Tesla's November Dallas Cybercab launch and external sales within the year (company targets); MLIT's application of a common certification standard in December (citing a Nikkei report; official notice unverified); Japan's about 10,000 Level 4 vehicles by 2030 (government target); the course of rulings on Hesai's 1260H designation status (unverified); Atlas completing 65 of 100 trucks by mid-2027 and about 45 by year-end (company outlook); the share of Korean Teslas that cannot use FSD, 76.2% vs 77.6% (conflicting, different reference dates); whether the KFTC complaint was accepted (unverified); Humble Robotics' US$51 million Series A (single source) and pilot operations within 12 months (per the company).

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