AUSTIN, TEXAS. Sometime in early 2026, for the first time anywhere, a person standing on a sidewalk in a major American city had a real choice between two driverless cars. Open Uber and get a Waymo. Open the Tesla app and get a robotaxi.
That perimeter — a geofenced swath of the Austin metro — is the front line. Not because it's where the technology works, but because it's the first place where the technology is a business, and two companies with completely opposite theories about how to make it one are standing on the same curb.
THE HISTORY NOBODY WANTS TO REMEMBER. In 2013, Nissan, Mercedes-Benz, Toyota and Volvo all announced autonomous driving goals or demonstrated capabilities. Audi, BMW and Bosch did the same in 2014. Then the Society of Automotive Engineers published its six levels of automation — three through five capable of driving themselves, four and five requiring no human at all, five being the ability to drive anywhere, anytime, in any condition.
The industry read the chart and made promises. Ford said level four by 2021. BMW said level five by 2021. Toyota believed multiple manufacturers would have level four vehicles before 2027. Those dates came and went, and nearly everyone retreated to lane assist and adaptive cruise.
What's left standing is two companies, two architectures, and one question: can either of them run a taxi service that makes money?
THE TORTOISE. Waymo started in 2009 inside Google X, where the question was whether the technology was possible — not whether it was commercially viable. That origin story is the whole company. When it spun out as an Alphabet business in 2016, the press release framed the mission around the 1.2 million lives lost a year to tired, drunk and distracted driving. Safety first, business model later.
You can see that in the hardware. A Waymo Jaguar I-Pace runs 29 cameras, four lidar sensors and six radar sensors. The lidar on the roof fires laser pulses to build a 3D picture; cameras read the color of the streetlight; radar tracks motion. Then the company goes out and 3D-maps each service area by hand, calibrating for local weather and traffic laws, so every vehicle always has a blueprint of where it is.
The safety numbers are real: studies of Waymo's operations in Arizona and California found an 80 percent reduction in injury-reporting crashes and a 55 percent reduction in police-reported crashes versus human drivers. And because there's no supervisor in the car, even a "disengagement" is unacceptable — if a Waymo can't handle something, it calls a human for help, which is a remarkable admission for a driverless car and the reason it works.
THE BILL. A Waymo-equipped I-Pace is estimated at $150,000 to $200,000 a unit. That number is the entire problem, and it's also the reason the tortoise might win anyway.
| VEHICLE | EST. COST PER UNIT |
| Waymo Jaguar I-Pace (retrofitted) | $150,000–$200,000 |
| Waymo Zeekr van (3,200 imported, fitted in Arizona) | ~$32,000 |
| Hyundai Ioniq (deal for ~50,000 units) | ~$50,000 |
| Lidar, 2014 | ~$75,000 |
| Lidar, today | hundreds |
That's the whole thesis in one column. Lidar followed a Wright's Law curve — the kind where cost drops as cumulative production rises — from $75,000 a system to hundreds of dollars. Waymo imported 3,200 Zeekr electric vans from China and retrofitted them in Arizona for roughly $32,000 apiece, a fraction of the Jaguars. It signed a multi-year deal with Hyundai for around 50,000 Ioniqs at a reported $50,000 each. Every year, the same service costs less to deliver.
Waymo's business only has to survive long enough for the hardware to get cheap. Tesla's only has to be safe enough before the software gets good.
THE HARE. Tesla's plan is from 2016, the same year Waymo spun out: Master Plan Part Deux. Take mass-produced electric cars, make them autonomous, let them be hailed and shared — a self-driving Uber. Scale the fleet first; the technology catches up.
On the ground in Texas, per the video, that's about 140 commercial robotaxis, mostly Model Ys with a portion of purpose-built Cybercabs. The service launched in 2025. Rides are described as more mechanical than Waymo's and cheaper. The Cybercabs arrived in September 2026 and were immediately the subject of an NHTSA audit — a vehicle with no pedals or steering wheel raises the small matter of how Tesla self-certified it as meeting federal safety standards.
The technology bet is the opposite of Waymo's in every way: eight cameras, no lidar, no radar, ultrasonic sensors removed in 2022. The argument is unnervingly simple — if humans drive with vision alone, a car can too, and this is a software problem, not a hardware one. Recent software merged perception, route planning and vehicle control into one end-to-end neural network instead of separate modules, which is the most consequential change in the stack: it stops being a set of rules and starts being a learned behavior.
Tesla says its full self-driving software is involved in seven times fewer major and minor crashes and five times fewer off-highway collisions. It still doesn't operate at night. It hasn't begun California's level-four certification, which is the rigorous one.
THE LATENT FLEET. Here's what makes Tesla dangerous despite being behind: roughly four to five million vehicles have already been delivered with capable hardware. If the software crosses the line, the fleet is already built and already on the road. Waymo has to manufacture or retrofit every car it operates. Tesla just has to ship an update.
That's a genuinely different business. It's also unfalsifiable on any timetable — Musk has predicted full autonomy "a few years out" for two decades, and there's no way to distinguish what's feasible, what he says is feasible, and what he believes.
THE PART EVERYONE IGNORES. Uber had its own autonomous program, worked with Volvo, and shut it down after a fatal crash during testing in 2018. Now Uber is the distribution layer: it partners with Waymo in Phoenix, Austin and Atlanta — until that deal ends in 2028, when Waymo intends to serve those riders through its own app. Uber is hedging with Rivian and Baidu; Lyft is working with Tensor and Nvidia on what its CEO calls the coming "hybrid network"; Amazon-backed Zoox is live in Las Vegas and expanding. Waymo is mapping London, Tokyo and Singapore.
SPLASH'S READ: The race isn't about autonomy. It's about which company can survive the gap between a working product and a cheap one.
Waymo is running a hardware business disguised as a software company. Its cost structure is terrible today and improving on a schedule it can predict: lidar down from $75,000 to hundreds, vans at $32,000 instead of Jaguars at $175,000, a Hyundai order of 50,000 that turns a fleet into a purchase order. If the cost curve keeps going, the tortoise doesn't need a breakthrough. It needs patience, which is the one resource Alphabet has in abundance.
Tesla is running a software business disguised as a car company, and its cost structure is already excellent — it builds the vehicles — but its product is unproven in exactly the conditions that matter. No night service. No California certification. An NHTSA audit on day one. And a camera-only stack that is either the correct long-term answer or a bet that machine learning will paper over a missing sensor, a question that has been "one year away" since 2016.
The asymmetry is what makes this interesting. Waymo can lose slowly. Tesla can win suddenly. One is a company that has to keep getting cheaper, the other is a company that has to be right once. And the thing that will actually decide it isn't technology at all — it's which one convinces a nervous public to get in. A quarter of people in one 2024 survey said a driverless car should still have a driver in it, and in 2025 a plurality still said humans are the safest option. Nobody is winning that argument with a press release.
Meantime, the most honest sentence in the whole documentary is about the past: in the 2010s this was going to be solved by 2021. What actually happened is that two companies found a way to make a car drive itself inside a mapped perimeter in one Texas city, and now they're arguing about the fare.