AVFox
Loading AVFox content.
← All rankings
Autonomous mobility
ADS Providers
Building a self-driving system and running it at scale are two different challenges. This ranking measures the second: how far each company's system has moved from supervised testing to commercial driverless service and, once it gets there, how many trips it delivers.
How to read the scores. Every AVFox ranking uses the same yardstick: 10 means 10 billion autonomous trips or deliveries a year, about the number of rides Uber completes worldwide, and 3 marks the start of driverless commercial service. Each point in between is meant to take about the same effort to climb. The score measures one thing: how many autonomous trips or deliveries a company completes today, not its potential, profitability, safety or quality of service. Few companies publish these numbers, so most scores are estimates built from public information.
0–2.9 progress toward commercial driverless service · 3.0–10 annual commercial driverless trips · How the score works
Research preview Based on public deployment data and updated as new figures come in. All scores are estimates. Last updated November 7, 2026.
- Geography
- Global (excludes China-only operators)
- Status
- Research preview
- Methodology
- ADS Providers methodology v0.2
| Rank | Provider | Deployment stage | Score (out of 10) | Summary |
|---|---|---|---|---|
| 1 |  WaymoDriverless commercial Largest driverless service outside China: paid rides open to the public in 14 US cities, with about 4,000 cars. | Driverless commercial | 5.1 | Largest driverless service outside China: paid rides open to the public in 14 US cities, with about 4,000 cars. |
How we got hereHow the score was calculatedWaymo's last ride figure, 500,000+ paid rides a week, was first reported in March 2026, when it had about 3,000 cars; it still cites that milestone. Its fleet is now about 4,000. Counting the added cars at half the utilization of the existing fleet, since most are in cities still ramping up: about 583,000 rides a week, or 30.3M trips a year (estimated) → 5.09, shown as 5.1. Footprint14 driverless public cities in the US; 4+ testing; London, Tokyo and Madrid announced. About 4,000 vehicles. What would move the scoreA new ride figure from Waymo. 5.2 needs about 34.2M trips a year, roughly 657,000 a week. | ||||
| 2 |  Apollo Go (Baidu)Driverless commercial Largest driverless operator in China, also running driverless service in Abu Dhabi and Dubai; volume fell in Q2 2026. | Driverless commercial | 4.6 | Largest driverless operator in China, also running driverless service in Abu Dhabi and Dubai; volume fell in Q2 2026. |
How we got hereHow the score was calculated10.7M fully driverless rides reported from Q3 2025 to Q2 2026 → 4.6. At the Q2 2026 pace alone (about 1M a quarter) it would score 4.2. FootprintAbout 22 driverless public cities, mostly in China plus Abu Dhabi and Dubai. Fleet size not disclosed. Germany and the UK announced for 2027. What would move the scoreQ3 2026 results (due mid-November) will show whether the Q2 drop was temporary. | ||||
| 3 |  Pony.aiDriverless commercial Driverless paid service in Beijing, Shanghai, Guangzhou and Shenzhen, with the largest disclosed robotaxi fleet after Waymo. | Driverless commercial | 4.6 | Driverless paid service in Beijing, Shanghai, Guangzhou and Shenzhen, with the largest disclosed robotaxi fleet after Waymo. |
How we got hereHow the score was calculatedTrips are not disclosed. 1,975 robotaxis (June 30, 2026) × 5,000 trips a year = 9.9M (estimated) → 4.60, shown as 4.6. Footprint4 driverless public cities in China; 4 more where driver status is unverified. Target of 20+ cities by the end of 2026. What would move the scoreA reported trip count, or the driverless share of its fleet. At its reported 23 orders per vehicle a day it would score 4.8. | ||||
| 4 |  WeRideDriverless commercial Driverless paid service in Abu Dhabi, Dubai, Beijing and Guangzhou, with 1,800+ robotaxis. | Driverless commercial | 4.5 | Driverless paid service in Abu Dhabi, Dubai, Beijing and Guangzhou, with 1,800+ robotaxis. |
How we got hereHow the score was calculatedTrips are not disclosed. 1,800 robotaxis × 5,000 trips a year = 9.0M (estimated) → 4.56, which rounds to 4.6. Apollo Go and Pony.ai also round to 4.6, and at most two companies share a score, so WeRide, third on full precision, shows 4.5. Footprint4 driverless public cities; 2 more where driver status is unverified. Madrid, Denmark and Hong Kong announced. What would move the scoreA reported trip count. At its reported 21 rides per vehicle a day in China it would score 4.7. Sources | ||||
| 5 |  TeslaDriverless commercial Driverless rides open to the public in 6 US cities, with 300+ Cybercabs in Austin; rides per car are still low. | Driverless commercial | 3.8 | Driverless rides open to the public in 6 US cities, with 300+ Cybercabs in Austin; rides per car are still low. |
How we got hereHow the score was calculatedTesla reports miles, not trips. About 1.65M driverless miles in Q3 2026 × 4 ÷ 6.7 miles per trip ≈ 1.0M trips a year → 3.76, shown as 3.8, about 9 trips per car a day. Tesla uses occasional remote control; with no one aboard, rides still count as driverless. Footprint6 driverless public cities; supervised service in the Bay Area; 4 cities announced. 300+ driverless vehicles. What would move the scoreReported trip counts. If October's pace holds, the score would be about 4.1. | ||||
| 6 |  ZooxDriverless commercial Purpose-built robotaxis carrying the public in Las Vegas (paid since August 2026) and San Francisco (free). | Driverless commercial | 3.6 | Purpose-built robotaxis carrying the public in Las Vegas (paid since August 2026) and San Francisco (free). |
How we got hereHow the score was calculatedZoox reports riders, not trips. About 100 vehicles × 5,000 trips a year = 0.5M (estimated) → 3.59, shown as 3.6. Footprint2 driverless public cities; about 10 testing cities; Los Angeles announced for 2027. About 100 vehicles (press reports). What would move the scoreA reported fleet or trip count. At its March–August rider pace (about 1.2M trips a year) it would score 3.8. | ||||
| 7 |  May MobilityDriverless · Restricted ODD Driverless public shuttle in Peachtree Corners, GA, limited to set stops and weekday hours; not a full driverless operation. Its other services use safety operators. Restricted ODD: Fixed route with set stops, weekdays only. Scored below 3.0 until the service runs all week, all day or across an area. | Driverless · Restricted ODD | 2.4 | Driverless public shuttle in Peachtree Corners, GA, limited to set stops and weekday hours; not a full driverless operation. Its other services use safety operators. |
How we got hereHow the score was calculated2.0 for driverless operation + 0.3 for public riders + 0.1 for approval (Georgia requires none) = 2.4. No points for all-day service, an open service area, a 25+ driverless fleet or a dated commercial launch. Footprint1 driverless site (restricted); 3 supervised cities; Arlington, TX on Uber announced. About 100 vehicles across all sites. What would move the scoreRunning 7 days a week for 12+ hours, serving an area instead of a fixed route, or a dated driverless launch. | ||||
| 8 |  MotionalSupervised commercial Rides on Uber in Las Vegas with a safety operator behind the wheel; driverless service targeted for the end of 2026. | Supervised commercial | 1.9 | Rides on Uber in Las Vegas with a safety operator behind the wheel; driverless service targeted for the end of 2026. |
How we got hereHow the score was calculated1.5 for supervised public rides + 0.2 for testing with the driver's seat empty (Motional's February 2026 post describes driverless testing in Las Vegas) + 0.1 for Nevada driverless approval + 0.1 for a dated driverless target = 1.9. The post does not establish whether anyone else could take control aboard or in a chase vehicle, so Motional stays below 2.0. Footprint1 supervised city (Las Vegas). 130,000+ supervised rides. Fleet size not disclosed. What would move the scoreEvidence of rides with no one aboard moves it to 2.0 or higher; opening those rides to the public moves it to 3.0 or higher. | ||||
| 9 |  MobileyeSupervised commercial Its system runs MOIA's free ride-pooling pilot in Hamburg and a passenger service in Orlando, both with safety operators. | Supervised commercial | 1.7 | Its system runs MOIA's free ride-pooling pilot in Hamburg and a passenger service in Orlando, both with safety operators. |
How we got hereHow the score was calculated1.5 for supervised public rides + 0.1 for approval (Florida requires none for driverless operation) + 0.1 for a dated driverless target (2027) = 1.7. Footprint2 supervised cities (Hamburg, Orlando); 5 testing; Los Angeles on Uber announced. 100+ ID. Buzz test vehicles. What would move the scoreTests with the driver's seat empty (+0.2), or driverless operation. | ||||
| 10 |  AvrideSupervised commercial Rides on Uber in Dallas with a safety operator behind the wheel; 60,000+ rides by June 2026. | Supervised commercial | 1.6 | Rides on Uber in Dallas with a safety operator behind the wheel; 60,000+ rides by June 2026. |
How we got hereHow the score was calculated1.5 for supervised public rides + 0.1 for Texas driverless authorization = 1.6. No dated driverless launch yet. Footprint1 supervised city (Dallas). 200+ vehicles across all uses. What would move the scoreA dated driverless launch (+0.1) or tests with the driver's seat empty (+0.2). | ||||
| 11 |  WayveSupervised commercial Rides on Uber in London since September 2026 with a safety driver; UK approval to remove the driver isn't in place yet. | Supervised commercial | 1.5 | Rides on Uber in London since September 2026 with a safety driver; UK approval to remove the driver isn't in place yet. |
How we got hereHow the score was calculated1.5 for supervised public rides. No readiness points yet: no UK driverless approval, no empty-seat testing reported, and its spring 2026 driverless date passed. Footprint1 supervised city (London); 15 cars at launch. Tokyo and 11 more Uber markets announced. What would move the scoreUK driverless approval (+0.1), a new dated driverless start (+0.1) or empty-seat testing (+0.2). | ||||
| 12 |  NuroSupervised testing Testing Lucid Gravity robotaxis for Uber in the Bay Area, with Uber employees riding alongside a safety driver; public launch planned for late 2026. | Supervised testing | 1.4 | Testing Lucid Gravity robotaxis for Uber in the Bay Area, with Uber employees riding alongside a safety driver; public launch planned for late 2026. |
How we got hereHow the score was calculated1.0 for supervised testing + 0.1 for employee riders + 0.1 for a CPUC permit to carry passengers + 0.1 for a late-2026 launch date + 0.1 for a California driverless testing permit = 1.4, the top of its band. FootprintTesting in the Bay Area and other US cities. About 100 test vehicles. What would move the scoreOpening rides to the public moves it to 1.5 or higher. | ||||
Based in China · Based in the United Kingdom
Methodology
How the score works
Every company gets one score from 0 to 10, shown to one decimal, rounded.
Commercial scale: 3.0 to 10
Once a company runs driverless rides open to the public, its score comes from annual trips, and each point is meant to take about the same effort to climb. We assume a mature robotaxi completes about 5,000 trips a year, about 14 a day.
| Score | Equivalent fleet | Trips a year |
|---|---|---|
| 3.0 | Launch (≤10 cars) | ≤50,000 |
| 4.0 | 500 | 2.5M |
| 5.0 | 5,000 | 25M |
| 6.0 | 40,000 | 200M |
| 7.0 | 200,000 | 1B |
| 8.0 | 500,000 | 2.5B |
| 9.0 | 1M | 5B |
| 10.0 | 2M | 10B, about the number of rides Uber completes worldwide each year |
Before commercial service: 0 to 2.9
| Score | Stage |
|---|---|
| 0.0–0.9 | Research, private roads and mapping |
| 1.0–1.4 | Testing on public roads with a safety driver |
| 1.5–1.9 | Carrying the public with a safety driver |
| 2.0–2.9 | Driverless testing, or driverless public service limited to a fixed route or schedule |
Within a band, points measure how close a company is to its next step, not how many cities it is in: a permit to carry riders, tests with the driver's seat empty, a dated launch, or running all day across an area. Permits and dates can add points but never move a company to the next band. A date that passes without a launch earns nothing.
Definitions
ADS provider: credit goes to the company whose system drives the car, not the app that books the ride.
Driverless: no one aboard, or in a chase vehicle, who can take control. Remote assistance or remote control doesn't make a ride supervised.
Commercial: driverless rides open to the public, paid or free.
Restricted ODD: a fixed route with set stops, or service on fewer than 7 days a week or 12 hours a day. These services stay below 3.0 and carry a note.
Coverage: global. Companies that operate only in China are excluded; for the rest, every deployment counts, including China.
Transparency
Where the data comes from
Trips, in order of preference:
1. Trips the company reported for the last 12 months.
2. A recent reported weekly or quarterly run rate, annualized.
3. Reported driverless miles for the latest quarter, annualized and divided by 6.7 miles per trip (Waymo's ratio, including empty miles).
4. Vehicles in driverless public service × 5,000 trips a year.
If the latest reported figure is more than three months old, we scale it by the company's fleet growth since then, counting new cars at half the utilization of the existing fleet.
Figures from steps 3 and 4, and scaled figures, are estimates.
Sources, in order of weight:
1. Company disclosures: earnings releases, SEC filings, investor materials and company announcements.
2. Regulators: permits, registries and required reports, such as the California CPUC and DMV and the Texas DMV.
3. Credible press: two independent reports are needed when there is no primary source.
4. Independent trackers: used to find and cross-check evidence, never as the only source.
Rules. Evidence must be from the last 12 months. A permit or announcement never sets a stage. A fact we can't confirm earns nothing. There are no analyst adjustments. At most two companies share a score; larger ties are separated within their band.
Regulator and tracker sources we monitor
- California CPUC quarterly AV passenger-service reports ↗
California-only; AVFox does not extrapolate this reporting without an explicit model.
Regulatory reporting
- Texas DMV automated motor-vehicle registry ↗
A registry entry is not evidence of active commercial driverless service.
Regulatory registry
- Platform Aeronaut AV Rideshare Tracker ↗
May combine VIN-observed and estimated counts; used for discovery and cross-checking.
Independent tracker
- Robotaxi Tracker ↗
Used for discovery and cross-checking, not as a substitute for primary evidence.
Independent tracker
AVFox distinguishes announcements, permits, testing, and commercial operation. Coverage does not imply that a company has deployed or operates commercially in every market.