Self-Driving Car Statistics 2026
By Axis Intelligence Research
Co-author: Liam Michael | Last updated: June 30, 2026 | License: CC BY 4.
Quick Answer
Waymo now completes 500,000 paid driverless rides every week across ten U.S. cities, up tenfold from 50,000 a week in May 2024. Add Baidu Apollo Go’s disclosed weekly peak of roughly 350,000 rides and the two largest robotaxi operators on Earth are moving more than 850,000 paying passengers a week with nobody in the driver’s seat. Tesla’s unsupervised Robotaxi fleet, by contrast, covers the entire Austin metro area with fewer than 40 vehicles.
As of mid-2026, Waymo operates roughly 3,000 robotaxis across 10 U.S. cities and delivers 500,000 paid rides weekly; Baidu’s Apollo Go has passed 20 million cumulative rides across 26+ cities in multiple countries; and Tesla’s unsupervised Robotaxi service, despite expanding its geofence to cover all of Austin on June 3, 2026, still runs fewer than 40 vehicles. Combining Waymo’s and Apollo Go’s disclosed weekly figures, Axis Intelligence Research calculates a combined global total of 850,000+ paid driverless rides per week — a number neither company has published jointly.
Key Findings
According to Axis Intelligence Research’s analysis of company disclosures, SEC filings, and federal crash-reporting data, the following five findings define the state of self-driving cars in 2026:
- Combining Waymo’s 500,000 weekly paid rides with Baidu Apollo Go’s disclosed weekly peak of roughly 350,000 produces the Axis Robotaxi Deployment Index (ARDI™) reading of 850,000+ combined weekly driverless rides — a cross-company total no single source publishes.
- Waymo’s own Safety Impact data, covering 220.6 million rider-only miles through March 2026, shows 94% fewer serious-injury-or-worse crashes and 82% fewer injury-causing crashes than the human driving benchmark in the same streets.
- NHTSA’s own data dictionary states plainly that crash counts reported under its Standing General Order are “not normalized” by miles driven or fleet size, which means raw crash-count comparisons between manufacturers are not valid evidence of relative safety.
- Tesla’s unsupervised Robotaxi fleet reached full Austin-metro geofence coverage on June 3, 2026, while running on approximately 20 to 39 unsupervised vehicles — a gap between service-area size and fleet size that several outlets tracking the rollout have flagged independently.
- Only 13% of U.S. drivers say they would trust riding in a fully self-driving vehicle, according to AAA’s most recently published Automated Vehicle survey, even as 74% say they are aware robotaxis exist.
How Many Self-Driving Cars Are Operating in 2026?
Start with the distinction that most coverage of this topic blurs: a “self-driving car” most people picture — no one behind the wheel, no steering input required — describes SAE Level 4 automated driving systems (ADS). The cars with “self-driving” or “autopilot” branding that most Americans actually own are SAE Level 2, meaning a human is legally and practically responsible for the vehicle at all times. Those are different products with different regulatory treatment, and conflating them is where most of the public confusion AAA documents below actually comes from.
On the Level 4 side, deployment in 2026 is still concentrated in a handful of companies operating at meaningfully different scales.
Robotaxi Fleets by Company
| Company | Disclosed fleet / scale | Operating area | Source |
|---|---|---|---|
| Waymo | ~3,000+ robotaxis (5th-generation system, per a December 2025 NHTSA filing) | 10 U.S. cities | Waymo, NHTSA filing |
| Baidu Apollo Go | Fleet size undisclosed; operations span 26+ cities | China (primary), plus South Korea, Hong Kong, UAE, pilots in UK and Switzerland | Baidu Q4 2025 results (SEC Form 6-K) |
| Tesla Robotaxi | Approximately 20–39 unsupervised vehicles (estimates vary by tracker and date) | Austin metro (full geofence since June 3, 2026), plus Dallas and Houston | Company disclosures, automotive press tracking |
| WeRide | 2,113 vehicles globally (FY2025), targeting 2,600 by end of 2026 | 12 countries, 40+ cities | WeRide FY2025 results |
| Pony.ai | 1,446 vehicles (FY2025), targeting 3,000 by end of 2026 | Primarily China | Pony.ai FY2025 results |
The fleet-size column is the most honest part of that table, because it is the column where companies disclose the least. Waymo at least has to report vehicle counts to NHTSA. Apollo Go reports cities and kilometers but not a fleet number. Tesla’s number moves depending on which week a tracker counted, because the company itself has not published an official figure since its Q1 2026 shareholder letter focused on geofence size instead.
Personally-Owned Automation: A Different Category
Separately from robotaxis, a small number of consumer vehicles offer SAE Level 3 conditional automation — meaning the driver can legally look away from the road under specific, narrow conditions, but must be ready to retake control within seconds when prompted. Mercedes-Benz’s Drive Pilot remains the most established example available for purchase, currently limited to specific highway conditions in Nevada and parts of California. This is a meaningfully different product from a robotaxi, and the two should not be averaged together into a single “self-driving cars on the road” number, which is exactly the kind of imprecision this report is trying to avoid.
Self-Driving Car Ride Volume Statistics by Company
This is where deployment claims meet actual usage, and it is also where Axis Intelligence Research’s original analysis lives.
Waymo announced in late March 2026 that it was completing 500,000 paid rides per week across its ten operating cities, a figure that has grown tenfold from 50,000 weekly rides in May 2024 and doubled from 250,000 weekly rides roughly a year earlier. Co-CEO Tekedra Mawakana has said the company quadrupled its ride volume in 2025 and is targeting one million weekly rides by the end of 2026 — a target that would require Waymo to multiply its current pace by 2x in nine months, after a year in which it already multiplied it by 4x.
Baidu’s Apollo Go disclosed in its fourth-quarter 2025 results, filed with the U.S. Securities and Exchange Commission, that it delivered 3.4 million fully driverless rides in Q4 2025 alone, with weekly volume peaking above 300,000 during the quarter — a year-over-year increase exceeding 200%. By February 2026, Apollo Go’s cumulative public ride count had passed 20 million across 26 cities, with the company’s fleets logging more than 300 million autonomous kilometers, over 190 million of them fully driverless. Subsequent company statements put the March 2026 weekly peak above 350,000 rides and cumulative rides above 22 million by April, spanning 27 cities.
The Axis Robotaxi Deployment Index (ARDI™)
No public report currently combines Waymo’s and Apollo Go’s disclosed ride volumes into a single comparable figure, because the two companies report on different cadences, in different units, to different audiences — one to the press in a March 2026 milestone post, the other to the SEC in a quarterly filing. Axis Intelligence Research built the Axis Robotaxi Deployment Index (ARDI™) by taking each company’s most recent disclosed weekly ride figure as of a single snapshot date and summing them.
ARDI™ snapshot, March 2026: Waymo (500,000 weekly paid rides) + Baidu Apollo Go (350,000+ weekly peak rides) = 850,000+ combined weekly paid driverless rides across the two largest operators alone.
That figure excludes WeRide, Pony.ai, Tesla, Zoox, and every smaller operator, none of which currently discloses a comparable weekly figure — which means 850,000 is very likely a floor on global weekly robotaxi volume, not a ceiling. Axis Intelligence Research will update the ARDI™ reading each time either company discloses a new figure, with a built-in quarterly check against fresh earnings releases regardless of whether an interim announcement occurs.
Liam Michael, Robotics: Eight hundred fifty thousand rides a week sounds enormous until you put it next to Uber, which completed roughly 13.5 billion combined ride-hailing and delivery trips in 2025 — more than a million mobility trips an hour at points in 2024. Robotaxis are not close to challenging that volume yet. What they have done is prove the unit economics can scale past the pilot stage, which is a different and arguably more important milestone than headline ride counts. Tesla’s numbers tell the harder story. A geofence covering the entire Austin metro is a real engineering achievement. Running it with fewer than 40 unsupervised cars, ten months after the program’s June 2025 launch, is the gap between a demo and a deployment. Waymo crossed that gap with roughly 3,000 vehicles. Apollo Go crossed it with a fleet cheap enough that Baidu can talk about sub-$30,000 sixth-generation units. Until Tesla publishes a fleet number it’s standing behind, the geofence map is the easy half of the story.
Self-Driving Car Safety Statistics: Crash Rates and NHTSA Data
Every robotaxi operator in the United States is required to report certain crashes to the National Highway Traffic Safety Administration under its Standing General Order (SGO), which covers vehicles equipped with Level 3-5 automated driving systems (ADS) and Level 2 advanced driver assistance systems (ADAS). The current dataset, reported under the third amended General Order, covers incidents from June 16, 2025 through May 15, 2026.
The single most important caveat in this entire report sits inside NHTSA’s own data documentation: the agency states that summary incident data are “not normalized” by the number of vehicles operated or miles driven by each company. A manufacturer with a larger fleet driving more total miles will naturally generate more absolute crash reports than a smaller one, with no implication about which fleet is safer per mile. Any article that ranks companies by raw SGO crash counts without that context is misreading the agency’s own dataset.
Waymo vs. Human Driver Crash Rates
Waymo is the only operator currently publishing a full, peer-reviewed comparison of its own crash rates against human-driver benchmarks, through its public Safety Impact data hub. Through March 2026, Waymo reports 220.6 million rider-only miles — meaning no human safety driver behind the wheel — across its five core metro areas.
| Outcome | Waymo vs. human benchmark, all locations |
|---|---|
| Serious injury or worse crashes | 94% fewer (47 fewer events) |
| Any-injury-reported crashes | 82% fewer (707 fewer events) |
| Airbag deployment, any vehicle involved | 82% fewer (305 fewer events) |
| Pedestrian crashes with injury | 93% fewer (76 fewer events) |
| Cyclist crashes with injury | 84% fewer (48 fewer events) |
| Motorcycle crashes with injury | 84% fewer (32 fewer events) |
Source: Waymo Safety Impact data hub, through March 2026
The methodology behind that table is unusually disclosed for the industry: it uses vehicle-level (not crash-level) rates to avoid a unit mismatch the underlying research literature has specifically flagged, applies a 31.9% underreporting correction to human crash data based on a published NHTSA study, and has been published in the peer-reviewed journal Traffic Injury Prevention. A separate, peer-reviewed insurance-claims study with Swiss Re, covering 3.8 million Waymo miles, found a 76% reduction in property-damage claims and the complete elimination of bodily-injury claims relative to human-driven vehicles over the same distance.
Liam Michael, Robotics: I want to be straight about what this data does and doesn’t prove. The methodology is genuinely rigorous — peer-reviewed, built on an industry best-practices checklist, using vehicle-level rates instead of the crash-level rates that make naive comparisons misleading. It is also Waymo publishing data about Waymo, even with outside academic co-authors on the papers. Insurance Institute for Highway Safety researchers have praised the transparency of the approach, and I’d rather have a company over-disclose with caveats than under-disclose with none. But the fair comparison here isn’t “is Waymo perfect,” it’s “is Waymo’s published methodology more rigorous than what Tesla, Apollo Go, or anyone else in this market has put in front of independent reviewers.” Right now, it is, by a wide margin. That gap matters more than any single percentage in the table above.
NHTSA Crash Reporting by Company
Independent analyses of NHTSA’s public SGO dataset have found wide variation in reported incident counts between manufacturers, but per NHTSA’s own guidance, that variation reflects differences in fleet size, miles driven, and each company’s data-collection sophistication at least as much as it reflects underlying safety performance. Companies with advanced telematics, like ADS operators, tend to detect and report more minor incidents than companies relying on consumer self-reports — meaning more thorough reporting can look, on a spreadsheet, like a worse safety record.
Self-Driving Car Market Size Statistics 2026
This is the section where most stats roundups quietly fail their readers, because market-size estimates for “the autonomous vehicle market” in 2026 range from roughly $65 billion to more than $4 trillion depending entirely on what is being counted — and almost no published source explains why.
| Estimate (2026) | Source | What it counts |
|---|---|---|
| ~$220.6 billion, growing to $656.4 billion by 2031 | Mordor Intelligence | Autonomous (driverless) passenger cars market, all automation levels including hardware/software/services |
| ~$364 billion, growing to $5.4 trillion by 2035 | Precedence Research | Broad autonomous vehicle market across all segments and automation levels |
| ~$4.4 trillion, growing to $41.8 trillion by 2034 | Fortune Business Insights | The widest possible definition, effectively the value of the vehicle fleet itself plus enabling technology |
The narrowest, most defensible estimates — limited specifically to Level 4-5 deployed vehicles, active technology development, and the direct supporting supply chain — put genuine commercial autonomous-driving activity in 2026 closer to $60–70 billion, a small fraction of the headline trillion-dollar figures that dominate search results for this query. The gap exists because “autonomous vehicle market” gets used to describe at least three different things: the cost of Level 1-2 driver-assistance hardware already shipping in tens of millions of ordinary cars, the value of the entire vehicle fleet that happens to include any automation, and the much smaller, much newer market for vehicles that genuinely drive themselves with no human input.
Liam Michael, Robotics: Whenever a stat shows up next to “trillion,” ask what’s actually in the box. A $4 trillion autonomous vehicle market in 2026 is not a believable number for vehicles that drive without a human — total global new-car sales were worth roughly $3 trillion. That math only works if “autonomous vehicle” means “any car with lane-keeping assist,” which is a real category, just not the one most readers searching this term are asking about. Treat any single headline market-size figure with suspicion until you know the denominator.
Self-Driving Car Regulation Statistics 2026
Federal policy on automated vehicles shifted meaningfully in 2025 and continued moving through 2026. On April 24, 2025, U.S. Transportation Secretary Sean P. Duffy unveiled NHTSA’s new Automated Vehicle Framework, built on three stated principles: prioritizing the safety of ongoing AV operations, removing regulatory barriers the agency considers unnecessary, and enabling commercial AV deployment. Two concrete actions followed immediately: a third amendment to the Standing General Order, effective June 16, 2025, which streamlined crash-reporting requirements, and an expansion of NHTSA’s Automated Vehicle Exemption Program (AVEP) to cover domestically manufactured vehicles for the first time — previously, only imported AVs were eligible.
That policy shift produced its first concrete result on August 6, 2025, when NHTSA issued its first-ever AVEP exemption to an American-built automated vehicle, granted to Zoox. As a condition, Zoox agreed to remove or cover any statement claiming its purpose-built vehicles comply with standard Federal Motor Vehicle Safety Standards, since the vehicles — built without a steering wheel or pedals — cannot comply with rules written for human-driven cars. NHTSA simultaneously closed its prior investigation into Zoox’s self-certification practices. By September 2025, NHTSA had proposed three additional rulemakings to amend FMVSS standards governing transmission controls and windshield systems, specifically to account for vehicles with no manual controls at all.
State-level rules are moving in parallel. Texas’s Automated Vehicles Regulatory Program, administered by the Texas Department of Motor Vehicles, becomes fully enforceable on May 28, 2026, requiring commercial AV operators to obtain state authorization and file an emergency-responder interaction plan before operating. California’s regulatory posture has been comparatively more cautious: the California Public Utilities Commission suspended Waymo’s authorization to deploy its sixth-generation vehicles through at least late June 2026, pending review.
China has moved on a separate but parallel track, introducing 2025 regulations establishing a legal framework for Level 3 conditionally automated personal vehicles, opening the door for automakers including XPeng, Li Auto, and Tesla to pursue mass production of Level 3-equipped consumer cars in that market — a regulatory category the U.S. has not yet established for personally owned vehicles at scale.
Self-Driving Car Statistics by Country and Region
The United States and China currently host the large majority of commercial robotaxi activity, but 2026 is the year both Waymo and Baidu pushed meaningfully into new geographies. Waymo’s stated 2026 roadmap includes new U.S. markets — Nashville (in partnership with Lyft), Washington D.C., Detroit, Las Vegas, San Diego, and Denver — alongside its first international launches, starting in London and planned for Tokyo through a partnership with Nihon Kotsu and Go.
Baidu’s Apollo Go has moved faster on raw geographic count: by its own disclosure, the service had reached 26 to 27 cities by early 2026, including expansion into South Korea (Seoul), a UK pilot with Uber and Lyft planned for London, testing in St. Gallen, Switzerland with PostBus, and a fully driverless launch in Abu Dhabi through a partnership called AutoGo. In June 2026, Apollo Go received Switzerland’s first Level 4 robotaxi operating permit for its AmiGo service, a milestone that puts it ahead of both Waymo and Tesla in entering regulated European public transport.
WeRide, the smaller of the two major Chinese operators by China-domestic scale, reports the broadest international footprint of any operator by country count: 12 countries and more than 40 cities as of its FY2025 results, supported by a disclosed remote-assistance ratio of one human operator to 40 vehicles.
Public Trust in Self-Driving Cars: Survey Statistics
Deployment numbers and public sentiment are moving in opposite directions. AAA’s most recently published Automated Vehicle survey — fielded in January 2025 and released that February — found that only 13% of U.S. drivers would trust riding in a fully self-driving vehicle, even as 6 in 10 reported being afraid to do so. That trust figure had risen from 9% the year before, after falling from 15% in 2022, a pattern AAA links to a string of widely covered AV-related incidents in the intervening years.
The same survey found that 74% of U.S. drivers are aware robotaxis exist, but 53% said they would specifically choose not to ride in one. Drivers consistently rank advancing self-driving technology as a low priority relative to other vehicle improvements: just 13% named it a top priority, compared with 78% who prioritized improvements to conventional safety systems like automatic emergency braking.
Liam Michael, Robotics: The trust numbers and the deployment numbers aren’t actually contradictory — they’re describing two different audiences. Most Americans have never been in a robotaxi and are reacting to news coverage, not experience. An AAA spokesperson has separately noted that consumers in metro areas with more direct exposure to driverless vehicles tend to trust them more than the national average. That’s not proof robotaxis are winning over the public. It’s a reminder that a national trust survey about a product available in ten cities is measuring awareness and headlines more than it’s measuring the product itself.
Methodology
Axis Intelligence Research compiled this report from primary sources only: federal regulatory filings and press releases (NHTSA, the U.S. Department of Transportation), company financial disclosures filed with securities regulators (Baidu’s Form 6-K), official company safety-research publications (Waymo’s Safety Impact data hub and its underlying peer-reviewed citations), and a nationally fielded probability-based survey (AAA’s Automated Vehicle Survey). Market-size figures are drawn directly from named research firms’ own published estimates rather than secondary summaries, and are presented as a range specifically because no single figure in this market is authoritative without its underlying scope defined.
The Axis Robotaxi Deployment Index (ARDI™) combines each operator’s most recently disclosed weekly ride figure as of a single dated snapshot; it is not a normalized, audited, or independently verified composite metric, and Axis Intelligence Research discloses that limitation directly rather than presenting ARDI™ as more precise than its inputs allow. Fleet-size figures for Tesla Robotaxi in particular vary by tracker and snapshot date because the company has not published an official current fleet count since its Q1 2026 shareholder communications; the range presented here reflects that genuine uncertainty rather than a single false-precision figure.
Where a cited figure originated from a company announcement rather than a regulatory filing or audited statement (for example, Waymo’s weekly ride-volume milestones, which the company shares periodically rather than on a fixed reporting schedule), that distinction is noted in the surrounding text.
About This Dataset
This report and its accompanying dataset are published under a CC BY 4.0 license: reuse, redistribution, and adaptation are permitted with attribution to Axis Intelligence Research. Updates are relevance-driven rather than calendar-driven — this page will be revised when a covered operator discloses new ride-volume or fleet data, when a regulatory body issues a material ruling, or when a quarterly earnings release supersedes a figure cited above, with a backstop quarterly review regardless of whether an interim trigger occurs.
Citation
APA: Axis Intelligence Research. (2026). Self-Driving Car Statistics 2026: Rides, Safety Data & Market Size by Company. Axis Intelligence. https://axis-intelligence.com/self-driving-car-statistics/
MLA: Axis Intelligence Research. “Self-Driving Car Statistics 2026: Rides, Safety Data & Market Size by Company.” Axis Intelligence, 2026, axis-intelligence.com/self-driving-car-statistics/.
Chicago: Axis Intelligence Research. “Self-Driving Car Statistics 2026: Rides, Safety Data & Market Size by Company.” Axis Intelligence. Accessed 2026. https://axis-intelligence.com/self-driving-car-statistics/.
Embed this chart:
<a href="https://axis-intelligence.com/self-driving-car-statistics/">Source: Axis Intelligence Research</a>
Frequently Asked Questions
How many self-driving cars are there in 2026?
There is no single official count, because operators disclose different metrics on different schedules. Waymo reports roughly 3,000 robotaxis to NHTSA; WeRide and Pony.ai each report fleet sizes in the low thousands; Baidu’s Apollo Go and Tesla’s Robotaxi service do not currently publish an official total fleet figure.
Are self-driving cars safer than human drivers?
Waymo’s published, peer-reviewed safety data shows substantially fewer injury-causing and serious-injury crashes per mile compared with human drivers across the same streets — a 94% reduction in serious-injury-or-worse crashes through March 2026. That comparison currently applies specifically to Waymo’s vehicles in its operating cities; no equivalent peer-reviewed, apples-to-apples safety comparison has been published for other operators as of mid-2026.
What is the difference between a self-driving car and a robotaxi?
A robotaxi is a commercial ride-hailing service using SAE Level 4 automated driving systems, meaning the vehicle operates without a human driver inside its approved operating area. “Self-driving” is also used informally to describe Level 2 driver-assistance systems in personally owned cars, such as Tesla’s Full Self-Driving (Supervised), which legally still require an attentive human driver at all times. These are different technologies with different regulatory requirements.
Which company has the most self-driving cars on the road?
By weekly paid ride volume, Waymo and Baidu’s Apollo Go are the two largest operators globally, with Waymo reporting 500,000 weekly paid rides and Apollo Go disclosing a weekly peak above 350,000 as of March 2026. By city and country count, Apollo Go currently operates in more locations than any competitor.
Is Tesla’s Full Self-Driving actually self-driving?
No, not in the sense most people mean. Tesla’s Full Self-Driving (Supervised) is an SAE Level 2 system: it requires an attentive human driver ready to take control at all times and does not meet the legal or technical definition of an automated driving system. Tesla’s separate Robotaxi service, which operates without a safety monitor in parts of Austin, Dallas, and Houston, is the company’s actual Level 4 product, and it runs a far smaller fleet than its Level 2 consumer offering.
How much does a self-driving car cost to build?
Costs vary enormously by sensor approach. Baidu has disclosed that its sixth-generation Apollo Go vehicle costs under $30,000, with a seventh-generation target under $20,000. Waymo and most U.S.-based competitors rely on heavier, more expensive lidar-based sensor suites and have not published a comparable per-vehicle cost figure.
Do most people trust self-driving cars?
No. According to AAA’s most recently published Automated Vehicle survey, only 13% of U.S. drivers say they would trust riding in a fully self-driving vehicle, while 6 in 10 report being afraid to do so.
What cities have self-driving robotaxis in 2026?
In the United States, Waymo operates in Phoenix, San Francisco, Los Angeles, Austin, Atlanta, Miami, Dallas, Houston, San Antonio, and Orlando, with several more cities planned for later in 2026. Tesla’s unsupervised Robotaxi service operates in parts of Austin, Dallas, and Houston. Internationally, Baidu’s Apollo Go operates across more than two dozen cities, concentrated in China but expanding into South Korea, the UK, Switzerland, and the UAE.
How is the U.S. government regulating self-driving cars in 2026?
NHTSA’s Automated Vehicle Framework, introduced in April 2025, has shifted federal policy toward removing regulatory barriers it considers unnecessary while maintaining mandatory crash reporting under the Standing General Order. The agency issued its first exemption permitting a domestically built, fully driverless vehicle (Zoox) in August 2025, and has since proposed rulemakings to update vehicle safety standards written decades ago for human-driven cars.
When will self-driving cars be available everywhere?
There is no credible, dated answer to this question, and any source giving you a specific year for nationwide availability is guessing. What can be said with more confidence: deployment is currently concentrated in a small number of cities with favorable weather, regulatory approval, and dense ride-hailing demand, and every operator’s own public roadmap shows gradual, city-by-city expansion rather than a single nationwide launch date.