Waymo is using its 200 million fully autonomous miles to make a blunt argument against Tesla's camera-only bet: better AI can't recover detail a car never sensed in the first place.
Srikanth Thirumalai writes the onboard software that tells Waymo's cars what's around them. In an interview with Axios published August 26, 2026, he made the case against camera-only autonomy in the plainest terms: "the AI can only make sense of what it sees, and if you just can't see it, the AI can't do much."
That's the whole fight. Tesla is trying to prove that cameras and neural nets can do the job. Waymo is saying you don't remove the driver by asking software to guess its way around missing information.
Thirumalai didn't need to name Tesla for the target to be clear. Elon Musk has called lidar "a fool's errand" and said anyone relying on it is "doomed," comments TechCrunch reported from Tesla's 2019 Autonomy Day. Tesla then began removing radar from North American Model 3 and Model Y vehicles in 2021, and Ars Technica reported that it started dropping ultrasonic sensors in 2022 as it pushed deeper into Tesla Vision.
Waymo went the other way. Its cars combine cameras, lidar and radar. The company says the mix earns its keep: the system can carry on when one sensor has a bad day. Fog, dust, glare, a dark street with a pedestrian at the edge of the beam, these are not edge cases to the person sitting in the back seat. They're the ride.
Tesla Will Launch Its No-Steering-Wheel Cybercab in Austin on September 3 Tesla is launching its Cybercab robotaxi, a two-seat vehicle with no steering wheel or pedals, at an invite-only event in Austin on September 3, 2026. Attendees must be 21 or older and pass ID checks, and prediction markets give the company only 17-18% odds of hitting Elon Musk's sub-$30,000 price target this year. - Tesla Cybercab launch event in Austin September 3 - robotaxi with no steering wheel or pedals
Waymo's blog post sharpened the argument #
The Axios interview landed the same day Waymo published a blog post under Thirumalai's name called "10 AI Lessons from Driving 200+ Million Fully Autonomous Miles." The post says cameras are "incredible," then cuts straight through the softer version of the debate: they aren't enough for fully autonomous driving at scale.
One line does even more work. Waymo says turning a Level 2 driver-assist system into full autonomy is a "false summit." That's not a minor technical disagreement. It's a rejection of the path Tesla has sold for years: keep improving Full Self-Driving in customer cars, gather more miles, train bigger models, and eventually remove the human.
Look at the difference in what each company is counting. Tesla can point to a huge installed base and an ocean of supervised driving data. Waymo points to fully driverless miles, with nobody in the driver's seat and no customer expected to save the car when the software gets confused. If you're the rider, that distinction matters more than the size of the training set.
The safety data around Tesla's Austin robotaxi effort doesn't make the camera-only case stronger. Fortune and Electrek reported earlier this year that Tesla's supervised Austin fleet had logged roughly 14 crashes over an estimated 800,000 miles, about one crash every 57,000 miles. Tesla's own benchmark for a minor collision by a human driver was about one every 229,000 miles. The math is ugly, even before you ask how many interventions the safety monitors prevented.
Waymo has had incidents too. Every autonomous company does. But Waymo's public pitch is built around a different threshold: don't scale until the car can take full responsibility for the driving task. That's the standard regulators and riders should care about.
California just gave Waymo more room to prove it #
The timing helped Waymo. On August 14, the California Public Utilities Commission approved Waymo's Advice Letter 0004 and 0004-A, clearing an expanded driverless passenger service area tied to its updated Passenger Safety Plan and its newer Ojai vehicle platform. The CPUC's disposition letter says the approval was effective that day.
Electrek reported that the expansion covers 18 California counties, from Sonoma and Sacramento in the north to Los Angeles, Orange, Riverside, San Bernardino, San Diego and Ventura in the south. The California DMV's public Waymo permit page lists Jaguar I-Pace and Zeekr Ojai as approved platforms, all times of day and night, all speeds, and all rain, fog and other conditions. That's a wide operating box.
Nevada clears Tesla, Uber and Waymo to run up to 8,000 robotaxis in Las Vegas The Nevada Transportation Authority unanimously approved permits letting Tesla, Uber and Waymo run up to 8,000 paid robotaxis in Clark County over the next year. It's the largest coordinated state-approved autonomous vehicle rollout to date, and it puts Tesla's robotaxi bet directly against Waymo for the first time. - Nevada robotaxi deployment regulations and permits approved - Tesla Waymo Uber Las Vegas autonomous vehicles
It also gives Waymo's sensor argument a practical test. A car that has to handle freeways, city streets, rural roads and rail crossings in changing weather needs more than a clean demo route. It needs redundancy when the road gets boring, messy or badly marked. Frankly, that's where a lot of autonomy hype falls apart.
Tesla isn't standing still. Its first production Cybercab rolled off the Giga Texas line on February 17, 2026, and Tesla's quarterly filing says Cybercab was in pilot production in the first quarter, with volume production expected this year. The vehicle has no steering wheel or pedals, which makes the gap between manufacturing and unsupervised driving impossible to ignore.
That's the part you should watch. Tesla can build the cheaper-looking hardware story. Waymo can point to the harder operating record. If Tesla proves cameras alone can clear the bar, the whole industry changes. Until then, Waymo has the cleaner argument: you can't model your way past a sensor you chose not to carry.
Also read: Nvidia Is Closing In On A $12.9 Billion Deal To Buy Hugging Face • ShinyHunters Leaked 13 Million Carhartt Accounts After Insulting Its Negotiator • Marvell says its data center chip business will hit $18 billion by 2028