Multiple Waymo robotaxis lost battery power while stuck in San Francisco gridlock tied to highway closures for the Semiquincentennial Fourth of July fireworks display on the Golden Gate Bridge, forcing tow trucks to pull the stranded vehicles from active traffic lanes. The episode exposed a gap between how autonomous vehicle fleets are regulated and how they perform when city streets seize up around large public events. With Caltrans shutting down US-101 in both directions and closing key ramps, electric vehicles that could not escape the resulting congestion drained their batteries well before reaching a charging station.
Golden Gate Bridge closures trapped electric AVs in hours of gridlock
The trigger was specific and predictable. Caltrans District 4 announced that US-101 would temporarily close in both directions for the Semiquincentennial Fourth of July Fireworks Display on the Golden Gate Bridge. Planned ramp closures included the Girard Street on-ramp in San Francisco, funneling traffic off the freeway and onto surface streets that were already packed with spectators heading toward waterfront viewing areas.
For human drivers, the closures meant frustration and delay. For battery-electric robotaxis circulating without a human behind the wheel, the closures created a different kind of problem. Vehicles that entered the affected corridors before ramps shut down had no driver to make a judgment call about pulling over, conserving charge, or rerouting to a charger. Instead, they idled in bumper-to-bumper traffic until their batteries ran dry. Eyewitness accounts from riders and nearby motorists described Waymo vehicles stopped in travel lanes, hazard lights blinking, waiting for flatbed tow trucks.
The mechanics are straightforward. Electric vehicles consume energy while stationary with climate control running, sensors active, and onboard computers processing data. A robotaxi’s compute stack draws more power than a typical passenger EV because lidar, radar, and camera arrays operate continuously. In normal city driving, that overhead is a rounding error. In multi-hour gridlock with no clear exit, it becomes a material drain on range.
In this case, the closures created a kind of trap. Once the ramps and bridge approaches were shut, vehicles already committed to the corridor had few legal options to turn around or exit. Human drivers sometimes respond to that kind of situation by making technically illegal but situationally pragmatic moves: U-turns across double yellow lines, cutting through gas station lots, or diverting onto side streets not marked as official detours. An autonomous vehicle bound by strict mapping and traffic-law compliance will not take those shortcuts unless they are explicitly modeled as permissible behaviors, leaving it to inch forward until its battery management system forces a shutdown.
CPUC deployment filings offer no playbook for event-driven strandings
Waymo operates its San Francisco passenger service under permits governed by the California Public Utilities Commission. The CPUC deployment docket tracks Waymo’s filings and compliance history for its driverless passenger service. Those filings cover operational boundaries, safety reporting, and service conditions, but they do not appear to include protocols for how the fleet should respond when large-scale road closures strand vehicles far from charging infrastructure.
That gap matters because the Caltrans closure was not a surprise. The agency published its notice in advance, listing specific ramps and routes that would shut down and the hours when traffic would be diverted. A fleet operator with access to that information could, in theory, have geofenced the affected area, recalled vehicles before closures took effect, or staged mobile charging units nearby. Whether Waymo took any of those steps before the event is not documented in publicly available CPUC filings or Caltrans records, leaving regulators and the public to infer the company’s planning from what happened on the street.
The regulatory structure around autonomous vehicles in California splits authority between the CPUC, which oversees passenger service permits, and the Department of Motor Vehicles, which handles testing and deployment permits for the vehicles themselves. Neither agency has published rules that specifically require AV operators to pre-position charging resources or withdraw vehicles ahead of scheduled highway shutdowns. The result is a framework built for routine operations that does not fully account for the predictable chaos of a major public celebration, even when that celebration is announced well in advance and involves shutting down a central regional artery.
In practice, that means key questions are left to each company’s internal policies. How conservative should dispatch algorithms be when closures are scheduled? At what state of charge should an AV refuse a new trip that might enter a potentially congested area? When conditions deteriorate faster than expected, who bears responsibility if the vehicle ends up blocking a lane and requiring a tow? Without explicit regulatory expectations, those answers vary by operator and may only come to light after something goes wrong.
Battery range under stress differs sharply from typical city driving
The stranding episode raises a practical engineering question. Waymo’s current fleet uses the Jaguar I-PACE, an electric SUV with a battery designed for mixed urban and highway driving. Under normal conditions, the vehicle’s range is sufficient for a full shift of city pickups and drop-offs. But those “normal conditions” assume the vehicle is moving at reasonable speeds, with opportunities to reach a charger before the pack falls into its reserve band. Stop-and-go traffic at walking speed, combined with the energy demands of a full autonomous driving sensor suite, compresses effective range in ways that standard estimates do not capture.
Heat adds another variable. San Francisco’s summer temperatures are mild compared with other Waymo markets like Phoenix, but a vehicle stuck in traffic with passengers still runs its climate system. Even modest cooling loads, layered on top of constant perception and compute, can significantly increase hourly energy consumption. Each hour of idling with air conditioning and full sensor operation chips away at the battery reserve that would otherwise carry the vehicle to a charger, especially if the car had already completed several trips before entering the jam.
The core tension is that autonomous fleets cannot make the same improvised decisions a human driver would. A person stuck in gridlock with a low battery warning can pull into a gas station, ask a bystander for directions to the nearest charger, or simply abandon the vehicle in a safe spot and walk away. A robotaxi follows its programming. If that programming does not include a low-battery emergency protocol triggered by real-time traffic data and event calendars, the vehicle sits until it dies, transforming a range-management problem into a traffic-safety problem for everyone around it.
Engineers can mitigate some of this risk with more conservative thresholds. An AV could be programmed to begin seeking a charger earlier when it detects severe congestion, or to decline trips that route through zones flagged for closures. But those strategies trade off against availability and passenger convenience. If the system refuses too many trips during busy periods, riders may see the service as unreliable, undermining the business case for large fleets in dense urban cores.
What riders and regulators should watch after the fireworks strandings
The fireworks strandings are unlikely to be the last time an autonomous fleet is tested by a planned disruption. Cities routinely close major corridors for marathons, parades, and construction projects. As AV services scale, the intersection between event planning and fleet management will grow more important, and the stakes will extend beyond inconvenience to questions of safety and public trust.
For riders, the incident is a reminder to pay attention to how AV operators communicate during abnormal conditions. Clear in-app alerts about potential delays, low-battery contingencies, and the possibility of trip cancellations could help passengers decide whether to board a robotaxi heading toward a known choke point. Transparency about how the company plans for closures-and how often vehicles have required tows-would also give users a better sense of real-world reliability.
For regulators, the episode suggests several areas for scrutiny. One is data sharing: requiring AV operators to ingest official closure feeds and demonstrate how that information shapes dispatch decisions. Another is contingency planning: asking companies to submit and periodically update playbooks for dealing with prolonged congestion, including thresholds for withdrawing vehicles, contacting public agencies, and arranging rapid removal of disabled cars from live lanes.
Local transportation agencies, for their part, may need to treat large AV fleets as stakeholders when planning major events. That could mean earlier briefings, dedicated staging areas for tows or mobile chargers, or temporary operating restrictions in especially sensitive zones. Coordinated planning would not eliminate the risk of strandings, but it could reduce the odds that multiple driverless vehicles end up blocking the same corridor at once.
The Golden Gate Bridge closures were a celebration of national history, not a stress test designed for autonomous technology. Yet they functioned as one. How Waymo and its regulators respond-through updated protocols, clearer public communication, or new rules around event-driven operations-will help determine whether robotaxis can be trusted to handle not just the everyday flow of city traffic, but also the extraordinary moments when that flow comes to a halt.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
