baidu-robotaxi-outage-wuhan-china-2026
title: "The Cars Just Stopped" coverStrategy: "logo" coverKey: "baidu" date: "2026-04-05" tags: ["AI"] author: "SiliconFeed Editorial Team" coverKey: "baidu"
The Cars Just Stopped
More than 100 of Baidu's Apollo Go self-driving taxis froze simultaneously in the middle of traffic across Wuhan, China, on April 1st. Passengers found themselves trapped — SOS buttons didn't work, customer service lines were unreachable, and the cars offered no explanation for why they'd stopped.
A college student riding with two friends said their car malfunctioned four or five times during the trip before parking at an intersection. The display told them to stay seated with seatbelts on and wait for a representative "in five minutes." After 30 minutes of fruitless phone calls and another hour of waiting, they simply walked out — the doors weren't locked.
The Bigger Picture
A dashcam video posted to Chinese social media showed one driver passing 16 stranded Apollo Go vehicles in just 90 minutes. Some were stopped in fast lanes on busy highways, forcing other drivers to swerve and reportedly causing minor crashes.
Passengers described a system that was fundamentally unprepared for failure at scale:
- The SOS button was unavailable for some users
- Customer service lines were unreachable
- The app's emergency options didn't connect to anyone who could help
- One passenger had to physically force the door open as traffic behind her came to a complete standstill
Local Wuhan police called it "likely a system malfunction" and confirmed no injuries. Baidu has not publicly explained what caused the outage.
Why This Matters for Autonomous Vehicles
Waymo just hit 500,000 paid rides per week — double its rate from a year ago and halfway to its year-end target of 1 million. It's the most successful autonomous ride-hailing service in the world and proof that robotaxis can work at scale. But the Wuhan incident illustrates the flip side: when an autonomous fleet encounters a systemic failure, the consequences cascade faster and more widely than a human fleet ever could.
A human driver who encounters a mechanical problem pulls over, calls in, and exits the road. If a centralized system fails, hundreds of vehicles stop simultaneously across an entire city. There's no graceful degradation — just a fleet-wide brick event.
The Scaling Problem
Baidu has invested heavily in Apollo Go, deploying hundreds of vehicles across Wuhan and other Chinese cities. The economics are compelling: once the software works, scaling is cheaper than training thousands of human drivers. But the risk profile is different too.
A single point of failure in a centralized autonomous fleet can create city-wide disruption. The Wuhan outage may be the first major incident of this kind, but as robotaxi deployments grow globally, the probability of similar events only increases.
What Autonomous Vehicle Operators Should Watch For
- Graceful degradation protocols — Can a fleet handle partial failures without total collapse?
- Emergency contact redundancy — If the primary system goes down, is there a backup?
- Passenger communication — Even if the system can't fix itself, it should at least tell people what's happening
- Physical override mechanisms — Can passengers get out without forcing doors?
The Apollo Go outage is a wake-up call for every company deploying autonomous services: the harder your system scales, the harder it can fail.
Sources: WIRED, ABC, The New York Times