When a massive blackout hit San Francisco on Saturday, Waymo's fully autonomous robotaxis faced an unexpected test they weren't ready for. Videos across social media showed multiple driverless cars stranded in traffic as traffic signals went dark and the city descended into gridlock. The outage forced Waymo to pause operations, raising uncomfortable questions about whether self-driving cars can handle real-world crises as well as the industry claims.
Waymo, owned by Alphabet, hit pause on its driverless ride-hailing service in the San Francisco Bay Area Sunday morning after the previous day's blackouts exposed some uncomfortable truths about autonomous vehicle resilience. The temporary shutdown came after viral videos showed multiple driverless cars stuck in traffic while the city's power grid failed.
Saturday's outages were no minor event. A fire at a PG&E substation around 1:09 p.m. knocked out power to roughly 130,000 customers in San Francisco, with the peak hitting about two hours later. By Sunday morning, over 21,000 customers remained without electricity, mostly in neighborhoods like the Presidio, the Richmond District, and parts of downtown.
When traffic signals go dark, human drivers adapt through instinct and experience. They treat intersections like four-way stops, negotiate with other drivers through eye contact and body language. Waymo's software is supposedly designed to do the same thing, treating non-functional signals as four-way stops. In theory, that should work. In practice, San Francisco's blackout revealed the gap between theory and real-world chaos.
"The sheer scale of the outage led to instances where vehicles remained stationary longer than usual to confirm the state of the affected intersections," Waymo spokesperson Suzanne Philion told CNBC on Sunday. In other words, the autonomous vehicles didn't know what to do and stayed put, unable to confidently determine whether it was safe to proceed. "This contributed to traffic friction during the height of the congestion," she added.
San Francisco resident Matt Schoolfield witnessed the failure firsthand. He saw at least three Waymo vehicles stopped dead in traffic around 9:45 p.m. on Saturday. "They were just stopping in the middle of the street," he said. One vehicle he photographed was wedged between Parker and Beaumont on the north side of Turk Boulevard, apparently unable to figure out its next move.
Waymo claims it "closely coordinated with San Francisco city officials" and proactively paused service Saturday evening and into Sunday morning. "The majority of active trips were successfully completed before vehicles were safely returned to depots or pulled over," Philion noted. The company framed this as responsible behavior. It's also a tacit admission that their fully autonomous vehicles couldn't handle the situation on their own.
Then came Elon Musk's Saturday night X post: "Tesla Robotaxis were unaffected by the SF power outage." It sounds impressive until you look at what Tesla actually operates in San Francisco. Spoiler: they don't run autonomous robotaxis at all. Tesla's ride-hailing service uses "FSD (Supervised)," a driver assistance system that requires a human operator behind the wheel at all times. According to California regulators, Tesla hasn't even obtained permits for driverless operations in the state without human supervisors ready to take control.
So Musk's claim that his robotaxis were unaffected is technically true only because Tesla doesn't have any. It's the kind of move that works great in a tweet but collapses under five seconds of scrutiny. Meanwhile, Waymo, which actually operates fully driverless vehicles and genuinely is the industry leader in the West, had to admit its cars couldn't handle a crisis.
Bryan Reimer, a research scientist at MIT's Center for Transportation, didn't mince words about what the incident reveals. "Something in the design and development of this technology was missed that clearly illustrates it was not the robust solution many would like to believe it is," he told CNBC. He's right. Power outages aren't some unforeseeable act of God. They're predictable, recurring events that cities should be prepared for. That Waymo wasn't prepared is troubling.
Reimer went further, calling for a shift in how cities think about autonomous vehicles during crises. "In the foreseeable future, we will need to mix human and machine intelligence, and have human backup systems in place around highly automated systems, including robotaxis," he said. He also argued that regulators need to set limits on how many autonomous vehicles can operate in any given area at once, and that AV companies should be held accountable for creating gridlock the same way human drivers would be.
The timing of this failure matters because robotaxi services are expanding into other major U.S. cities. Waymo operates in multiple markets, and competitors like Baidu's Apollo Go are scaling in China. Public trust in these vehicles is already fragile - a 2025 survey by the American Automobile Association found that two-thirds of U.S. drivers fear autonomous vehicles. Episodes like the San Francisco blackout don't help.
The San Francisco blackout exposed a critical vulnerability in autonomous vehicle technology: real-world crises are messier and more complex than the controlled environments where most AV testing happens. Waymo made the right call by pausing service rather than letting confused robots clog the streets, but the incident raises serious questions about whether fully autonomous vehicles are ready for widespread deployment in cities without robust backup systems and strict regulatory guardrails. As the robotaxi industry scales, regulators need to catch up and demand that AV companies prove they can handle infrastructure failures as well as sunny-day driving. Until they do, episodes like this will keep reminding us that the autonomous vehicle revolution is still very much a work in progress.
