Where the Rubber Meets the Road
Ten years ago, self-driving car timelines were wildly optimistic. Elon Musk promised full autonomy by 2017. General Motors said 2019. Ford said 2021. None of that happened. But something interesting has happened in the last two years: autonomous vehicles quietly started working — in limited areas, under controlled conditions, but actually working.
Waymo: The Quiet Leader
Waymo (Alphabet’s self-driving subsidiary) operates the largest commercial autonomous ride-hailing service in the US. As of early 2025, Waymo One operates in Phoenix, San Francisco, Los Angeles, and Austin, with over 100,000 paid trips per week. That’s up from about 10,000 per week a year earlier — a 10x scaling in 12 months.
The safety data is genuinely impressive. Waymo published a study covering 7.14 million miles of fully autonomous driving (without a human driver) in Phoenix, San Francisco, and Los Angeles through March 2024. Compared to human drivers over the same distance, Waymo vehicles were involved in 85% fewer crashes resulting in injury, and 57% fewer police-reported crashes. When you adjust for the fact that Waymo operates mostly in urban areas (where crash rates are higher) and doesn’t drive impaired or distracted, the safety advantage becomes clearer.
But the operational challenges are real. Waymo vehicles still struggle with unusual situations — construction zones with flaggers, emergency vehicles approaching from unexpected directions, and heavy rain or fog that degrades sensor performance. The company handles these with remote assistance: human operators who don’t drive the vehicle remotely but can suggest paths when the AI is uncertain. The rate of remote assistance interventions has been declining, but it’s not zero and may never be.
Cruise: The Cautionary Tale
GM’s Cruise provides the industry’s most sobering lesson. In October 2023, a Cruise vehicle in San Francisco was involved in an accident where a pedestrian was struck by another vehicle and thrown into the path of the Cruise AV, which then braked but dragged the pedestrian about 20 feet before stopping. The incident itself was a difficult edge case, but Cruise’s response — initially not disclosing that the vehicle had dragged the pedestrian, and showing regulators an incomplete video — led to the suspension of Cruise’s California operating permit, the grounding of its entire fleet nationwide, the resignation of its CEO, and the layoff of roughly 25% of its workforce.
Cruise resumed testing with safety drivers in Phoenix and Dallas in mid-2024, but the damage was done. The incident exposed a truth the industry had been avoiding: autonomy isn’t just a technology problem. It’s a trust problem. One bad day can erase years of goodwill. Companies that prioritize scaling speed over safety transparency risk their entire business.
GM has since restructured Cruise, cutting spending by about $1 billion annually and shifting focus to a more conservative deployment strategy. The lesson for the industry: Waymo’s slow-and-steady approach, while less exciting, may be the only viable path.
Other Players
Tesla: Tesla’s “Full Self-Driving” (FSD) — a name that has drawn criticism from safety advocates — has improved significantly with version 12 (neural network-based end-to-end driving released in 2024) but remains a Level 2 driver assistance system that requires constant driver supervision. Tesla claims FSD has driven over 1.5 billion cumulative miles, but those miles include active driver interventions, making safety comparisons with Waymo’s truly driverless miles difficult.
Zoox (Amazon): Zoox is taking a different approach — building a purpose-built autonomous vehicle with no steering wheel or pedals, designed from the ground up for ride-hailing. They’re testing on public roads in San Francisco, Las Vegas, and Seattle, but commercial deployment timelines remain vague.
China: Baidu’s Apollo Go service operates in Wuhan, Beijing, and several other Chinese cities, with a fleet of over 500 autonomous vehicles. Didi and Pony.ai also have active autonomous ride-hailing pilots. Chinese regulators have been more permissive than their Western counterparts in allowing autonomous vehicles on public roads, which is accelerating development but also raising safety questions.
The Regulatory Landscape
Regulation is fragmented. In the US, autonomous vehicles are regulated primarily at the state level, creating a patchwork: California and Arizona are relatively permissive, Texas has minimal state-level regulations, and many states have no AV laws at all. The federal government has been slow to act — the NHTSA has the authority to regulate vehicle safety but hasn’t issued comprehensive AV standards.
In Europe, the UNECE has established global technical regulations for autonomous vehicles, and the EU has set a framework that member states are implementing. Germany, the UK, and France have all passed AV legislation, but deployments are more limited than in the US or China.
The Bottom Line
Autonomous vehicles work in 2025 — in geo-fenced urban areas, in good weather, with remote assistance backup, and at significant cost (Waymo vehicles are estimated to cost $150,000+ each). They’re safer than human drivers in the conditions they operate in, but they still can’t handle edge cases that humans navigate routinely.
The economic equation is challenging. A Waymo ride in San Francisco currently costs roughly the same as an Uber, but Waymo is subsidizing operations — the vehicles need to get a lot cheaper and the technology a lot more robust for autonomy to be profitable at scale.
The industry has made real, measurable progress since the overhyped promises of 2015. What’s less clear is whether the remaining 5% of capability — handling every possible edge case a human driver encounters — will take another 5 years or another 20.