History & approaches
Waymo vs Tesla: Two Bets on Autonomy
Waymo and Tesla are making opposite bets on how to reach self-driving. Waymo runs driverless robotaxis that combine lidar, cameras, radar and detailed maps, expanding city by city. Tesla relies on cameras and neural networks alone, across millions of customer cars and a growing robotaxi service, betting that scale and data will let it drive anywhere.
What are the two bets?
Waymo's bet is to build a complete driver first, with several kinds of sensors and a detailed map of each service area, and then scale it one city at a time. It sells rides rather than cars. Tesla's bet is to reach autonomy through scale: cheap cameras on millions of customer cars collect driving data, one software system learns from all of it, and the same system is meant to grow from supervised driving into a driverless service.
Each approach has a strength the other lacks, and each has a well-known criticism. Neither is proven to be the only path. The comparison below sets out what each company has stated and deployed as of September 2026.
| Waymo | Tesla | |
|---|---|---|
| Sensors | Lidar, cameras and radar (6th generation: 13 cameras, 4 lidar, 6 radar) | Cameras only (radar removed from 2021, ultrasonic sensors from 2022) |
| Maps | Detailed pre-built maps of each service area | Says it does not rely on HD maps |
| Software | Large AI models combined with a structured driving system | A single end-to-end neural network since FSD v12 (2024) |
| Who is responsible while driving | The Waymo Driver; no human driver aboard | FSD (Supervised): the human driver. Robotaxi and Cybercab: Tesla, in its service area |
| Scale so far | About 15 US metro areas; about 500,000 paid rides a week | About 12 billion supervised FSD miles; Robotaxi in seven metro areas |
| Business model | Runs its own robotaxi fleet, in some cities through Uber | Sells cars and FSD software; runs its own Robotaxi network |
| Cost per vehicle | Not disclosed; analyst estimates vary widely | Cybercab target under $30,000, according to Elon Musk |
| Main criticism | Costly vehicles and city-by-city mapping slow expansion | Cameras alone may not be enough in glare, fog or low light |
Sensors: two eyes and a brain, or many kinds of sensors?
Tesla's argument starts from how people drive. Humans drive with two eyes and a brain, so cameras, which play the part of the eyes, and a neural network, which plays the part of the brain, should be enough. In Elon Musk's words in January 2025, humans drive with eyes and a neural net, and the entire road system was designed for passive optical neural nets. Andrej Karpathy, then Tesla's head of AI, called lidar a shortcut in 2019, and when Tesla removed radar in 2021 he argued it was better to double down on vision than to fuse sensors that disagree. Cost is the other half of the argument: cameras are cheap enough to fit to every car, and Musk has called lidar expensive and unnecessary.
Waymo's argument starts from what a driverless car must never miss. In its words, cameras are incredible, but they aren't enough: lidar captures 3D geometry precisely, and radar can see through heavy rain, fog or dust that blind a camera. The sixth-generation Waymo Driver combines 13 cameras, 4 lidar units and 6 radar units. Co-CEO Dmitri Dolgov has said cameras alone are a very reasonable way to match human drivers or to build an assistance product, but that weak sensing leads to a safety curve that flattens out too early.
Each side faces a matching criticism. Camera-only driving is questioned in poor visibility: NHTSA is investigating FSD crashes in sun glare, fog and dust, and said in March 2026 that the system did not detect conditions impairing its cameras until immediately before the crash. A multi-sensor car is questioned on cost and complexity, which Waymo says its sixth generation reduces.
Maps: detailed prior maps or driving without them?
Waymo maps each service area in detail before launching, recording features such as curb heights, lane lines and traffic light positions, and says it uses these HD maps to jump-start its validation in each new place. Its cars report changes back so the maps stay current.
Tesla says it does not rely on HD maps. In 2019 Musk called high-precision maps a really bad idea that makes a system extremely brittle when the world changes, and said Tesla had briefly tried high-precision lane lines before deciding it was a mistake.
The trade-off cuts both ways. Detailed maps give a car strong prior knowledge of the road, but must be built and kept up to date for every place it drives, which critics say slows expansion. Driving without them lets a system go anywhere in principle, but it must understand every road from scratch, with no prior knowledge to fall back on.
Software: end-to-end or structured?
With FSD v12 in 2024, Tesla replaced its city-driving code with what its release notes described as a single end-to-end neural network trained on millions of video clips, replacing over 300,000 lines of explicit C++ code.
Waymo describes its system as combining fast sensor-fusion models with the reasoning of large vision-language models, presented as an alternative to purely end-to-end or purely modular designs. It has also published research on end-to-end driving (EMMA, 2024) without using it as its production driver.
Scale: who drives more, and who is responsible?
Waymo offers public driverless rides in about 15 US metro areas, averaging about 500,000 paid rides a week in September 2026, with more than 271 million rider-only miles through June 2026.
Tesla's fleet is far larger. Its customers had driven about 12 billion miles on FSD (Supervised) by mid 2026, with about 1.48 million active FSD subscriptions, and its Robotaxi service was live in seven metro areas. On customer cars, Tesla states that FSD (Supervised) requires active driver supervision and does not make the vehicle autonomous, which places it at SAE Level 2. Its Robotaxi service began some rides without a human safety driver in January 2026, and on 3 September 2026 Tesla began commercial rides in a small number of Cybercabs, which have no steering wheel or pedals. The same day, NHTSA opened an audit of how Tesla certified the Cybercab against federal safety standards.
The mileage figures are not like for like. Tesla's supervised miles have a human ready to take over, which Waymo's rider-only miles do not, while Waymo's miles come from far fewer cities.
Which is safer?
No independent study compares the two directly, and both companies publish their own figures. Waymo reports that through June 2026 its cars had 95 percent fewer serious-injury crashes and 82 percent fewer injury crashes than human drivers on the same roads, and a 2024 study co-authored with the insurer Swiss Re found 92 percent fewer bodily-injury claims. Tesla reports that with FSD (Supervised) engaged, its cars had one major collision every 5.3 million miles in early 2026, against a US average of one every 660,000 miles.
Both have drawn scrutiny. Researchers have questioned Tesla's comparison, which sets newer cars against the older US fleet, US senators asked NHTSA in June 2026 to review Tesla's safety data, and NHTSA's reduced-visibility investigation covers about 3.2 million vehicles. Waymo has issued several software recalls: in 2024 after a car hit a pole, in December 2025 over how its cars behaved around stopped school buses, and in 2026 over standing water and freeway construction zones. The NTSB opened an investigation into the school bus cases in January 2026, and NHTSA is investigating a Santa Monica crash in which a Waymo struck a child, who was slightly injured.
Watch
Frequently asked
- Is Tesla FSD self-driving?
- Not on customer cars. Tesla says FSD (Supervised) requires active driver supervision and does not make the car autonomous. Its Robotaxi service does run some rides without a driver in the car.
- What is the main difference between Waymo and Tesla?
- Waymo runs driverless robotaxis with lidar, cameras, radar and detailed maps, expanding city by city. Tesla relies on cameras and neural networks across millions of customer cars that a driver supervises, plus a growing robotaxi service.
- Does Tesla run a robotaxi service like Waymo?
- Yes, on a smaller scale so far. Tesla's Robotaxi service launched in Austin in June 2025 and was live in seven metro areas by mid 2026, with some rides running without a safety driver. Waymo offers public driverless rides in about 15 US metro areas.
- Which is safer, Waymo or Tesla?
- No independent study compares them directly. Both publish favorable figures from their own data, both have critics of their methods, and both have faced federal investigations and recalls.
- Can I buy a Waymo?
- No. Waymo sells rides, not cars, through its own app and through Uber in some cities.
Keep learning, once a week
The weekly digest: the stories worth your time from the autonomous vehicle industry, hand-picked by people who work in it. No marketing gloss.