Even Waymo’s self-driving taxis experienced significant confusion during the weekend. A huge blackout affected over 100,000 San Francisco customers. A recent fire left “significant and extensive” damage at a Pacific Gas and Electric Company (PG&E) substation. This damage was responsible for the outage. As a byproduct, the self-driving cars brought traffic to a halt, moderately increasing congestion in one of the most congested cities in the country.
The blackout started on Saturday, knocking out power to whole regions. PG&E described themselves as working ‘around the clock’ to restore power, restoring electricity to all but 17,000 customers as of Sunday afternoon. The company had previously forecast that it would have service restored to those last customers by mid-afternoon on Monday. Into this breach, San Francisco Mayor Daniel Lurie turned on his heels. During the outage, he redirected police officers and fire crews to assist in directing traffic.
Waymo’s self-driving technology typically adjusts to broken traffic signals by interpreting them as four-way stops. It had some notable failings during the recent massive outage. A spokesperson for Waymo explained,
“While the Waymo Driver is designed to treat non-functional signals as four-way stops, the sheer scale of the outage led to instances where vehicles remained stationary longer than usual to confirm the state of the affected intersections.”
This added a layer of traffic friction during times of peak congestion, making matters worse for those commuting in the region.
Waymo has big plans to expand. The startup plans to be available for hire in London and Washington, D.C., within a year. In 2025, Waymo provided over 14 million trips just that year, demonstrating its significant scale and leadership in autonomous transportation.
Slowly, power returned and life returned to normal. This incident brought renewed focus to the risks that automated systems suffer in moments of infrastructure failure. Waymo’s experience in the midst of the outage highlights the importance of robust contingency planning in any automated service. This is particularly critical in metropolitan regions where complex trade traffic patterns are super important.