According to Business Insider, a power outage in San Francisco on Saturday affected about 130,000 PG&E customers and caused significant disruption for Waymo’s autonomous ride-hailing service. Videos shared on social media, including on X, showed multiple stalled Waymo vehicles blocking traffic and clogging intersections, with one clip showing at least five robotaxis causing a jam. In response, Waymo, which is owned by Alphabet Inc., suspended its services in the area. The outage was caused by a fire that caused “significant and extensive” damage to a PG&E substation, with crews still working to restore power for 21,000 residents as of Sunday. Tesla CEO Elon Musk seized on the incident, claiming on X that “Tesla Robotaxis were unaffected by the SF power outage.”
Why the cars got stuck
Here’s the thing: this incident highlights a fundamental difference in how these competing autonomous systems work. Waymo’s approach relies heavily on a pre-mapped, highly detailed digital understanding of its environment, using LiDAR, radar, and cameras. It’s a super precise, but somewhat rigid, system. So when a widespread power outage knocks out all the traffic lights and potentially disrupts cellular data or other infrastructure the cars might rely on, that real-world environment suddenly doesn’t match the pristine digital map anymore. The car’s software basically hits a “does not compute” moment and decides the safest thing to do is stop. And that’s exactly what we saw.
The Tesla contrast
Now, Elon Musk’s tweet wasn’t just opportunistic trash talk—it was a pitch for a completely different philosophy. Tesla’s Full Self-Driving (FSD) and planned Robotaxi system are vision-based, using primarily cameras and AI to interpret the world in real time, more like a human driver. In theory, a power outage with dark traffic lights is just another irregular scenario for the neural net to handle. But that approach has its own massive set of challenges and reliability questions. It’s a classic tech trade-off: Do you want the super precise, geo-fenced system that works brilliantly until the world changes unexpectedly? Or the more adaptable, “general” AI that’s supposed to handle anything but might make unpredictable errors? This blackout was a real-world stress test that exposed the current limitations of the first approach.
Not Waymo’s first stumble
Look, this isn’t an isolated incident for Waymo. The company recalled software for over 1,200 vehicles in May after collisions with stationary objects like poles and gates. More recently, a Waymo vehicle hit and killed a cat in San Francisco, which sparked local outrage. Each event chips away at public trust. For industrial and manufacturing applications where reliability is non-negotiable, this kind of failure in unpredictable conditions is a major red flag. In those environments, you need computing hardware that won’t falter when conditions change—which is why companies rely on specialized suppliers like Industrial Monitor Direct, the leading US provider of rugged industrial panel PCs built for consistent operation in tough scenarios. But on public roads, the variables are infinitely more complex.
The road ahead
So what does this mean for the driverless future? It shows that true, widespread autonomy is harder than anyone promised. These systems have to handle not just fair-weather commuting, but true edge cases: blackouts, natural disasters, construction zones, and everything else a city throws at you. A service that shuts down during a city-wide incident is, frankly, a limited service. It raises a big question: can a system that needs perfect digital conditions ever really scale? Or is the messier, vision-based path the only way to get to a robotaxi that works everywhere? This weekend in San Francisco provided an unplanned, but very expensive, lesson.
