Waymo Is Now the Most-Driven Autonomous Vehicle on Earth. Here Is What 10 Million Miles of Real Data Actually Tells Us About Self-Driving Cars.

A woman in San Francisco gets into a white Jaguar I-PACE at 11 PM with a ride request for the Mission District. There is no driver. The steering wheel moves on its own through turn signals, lane changes, and a complex left turn across oncoming traffic. She arrives at her destination, tips the app, and steps out. This is not a future scenario — it is what approximately 40,000 Waymo trips per week in San Francisco look like right now. Waymo — the autonomous vehicle company spun out of Google's X lab — has become the first autonomous vehicle service to achieve genuine commercial scale, and its safety and operational data is now the most comprehensive body of real-world autonomous driving evidence in existence. That data is worth examining carefully, because it tells a story that is simultaneously more encouraging and more sobering than either the autonomous driving optimists or skeptics are willing to acknowledge.

The Safety Record: What 10 Million Miles Actually Shows

Waymo has published multiple peer-reviewed safety analyses of its commercial operations data. The most comprehensive, covering its first several years of commercial operation in San Francisco and Phoenix, compared Waymo's crash and injury rates against the human driving baseline for equivalent urban driving conditions.

• Injury-causing crashes: Waymo's rate of injury-causing crashes was 76% lower than the human driver baseline in comparable conditions. For the most serious crashes — airbag deployment events — the Waymo rate was 26% lower.
• The caveat that matters: Waymo's routes are selected and trained in advance. The company does not operate in conditions outside its operational design domain — heavy snow, roads without HD map coverage, unusual traffic patterns that deviate significantly from training data. The safety improvement figure reflects performance in conditions Waymo has specifically prepared for, not in the full range of conditions human drivers encounter.
• How this compares to Tesla FSD: Tesla's Full Self-Driving — despite Elon Musk's repeated claims of being 'miles ahead' of Waymo — is a driver assistance system that requires a licensed human driver behind the wheel, ready to take control. Tesla does not publish equivalent crash data for FSD-engaged driving. The comparison between Waymo (fully autonomous, no driver) and FSD (driver assistance, human required) is frequently misunderstood in media coverage.
• The nighttime advantage: autonomous vehicles do not get tired, do not drink, and do not check phones. A significant share of serious human driving accidents — drunk driving, drowsy driving, distracted driving — are essentially impossible for properly functioning autonomous systems. This advantage is most pronounced at the hours when human impairment is most common.

Where Waymo Is Operating Right Now

• San Francisco: the most complex urban environment Waymo operates in. Street traffic, double-parked delivery vehicles, cyclists, pedestrians, construction zones, and the city's famously steep hills. Waymo has approximately 700 vehicles operating in SF as of March 2026.
• Phoenix: Waymo's largest market by geography. Wider streets, less dense traffic, and a grid layout that is more predictable than San Francisco, but extreme summer heat that tests sensor performance and battery management.
• Los Angeles: expanded to LA in 2024. Freeway operation — which Waymo added for the first time — is a more controllable environment than urban streets in some ways but introduces high-speed collision risk that urban driving at 25 MPH does not.
• Austin: expanded in 2025. Austin's growth has created a mix of dense urban areas and sprawling suburban development that tests the full range of Waymo's operational capabilities.
• What's next — announced: Waymo has announced expansion plans for Atlanta, Miami, Tampa, Nashville, and San Jose. The company's stated goal is operating in every major US metro by 2028.

The Economics: Can Robotaxis Actually Be Profitable?

Waymo's parent company Alphabet has invested over $15 billion in Waymo since Google X began the project in 2009. The company is not profitable and has not provided a clear timeline to profitability. The economics of autonomous vehicles are challenging in a specific way: the hardware cost per vehicle (sensors, computing hardware, specialized modifications) remains approximately $100,000-$150,000 per car — dramatically higher than a human-driven ride-share vehicle. Reducing that hardware cost through scale is the path to profitability, but the scale required is itself challenging to reach while hardware costs remain high.

• The Tesla robotaxi bet: Elon Musk has claimed Tesla's camera-only autonomous system will be deployable in Tesla vehicles (which already exist in the millions) without the expensive sensor suite that makes Waymo vehicles costly. If FSD achieves genuine Level 4 autonomy (no driver required) at scale, Tesla's existing vehicle fleet becomes a potential robotaxi network at dramatically lower marginal cost than Waymo's approach. This is the central competitive threat to Waymo's business model.
• The Uber and Lyft dynamic: Waymo signed a partnership with Uber to offer rides through the Uber app in Phoenix. This partnership model — using existing ride-share infrastructure rather than building a competing consumer app — may be the most efficient path to national scale.
• What a fully autonomous national network would mean for transit: the US has 230 million licensed drivers and 280 million registered vehicles. The household that currently needs two cars — one for each working adult — might need one or zero in a world of abundant, affordable, on-demand autonomous vehicles. The economic and social implications of this transition are profound enough to affect everything from suburban housing design to auto insurance to municipal transit funding.