Freeway service paused across four markets
Waymo has suspended robotaxi service on freeways in San Francisco, Los Angeles, Phoenix and Miami as the company works through problems linked to construction zones. The pause leaves its self-driving vehicles operating on surface streets in those cities, but removes one of the most important pieces of capability the service added in late 2025.
The company told TechCrunch it is integrating recent technical learnings into its software and expects to resume freeway routes soon. Even without a timeline, the decision is significant. Freeways are not just another road type for an autonomous taxi service. They are essential to improving trip times, connecting distant neighborhoods and reaching airports efficiently in large metropolitan areas. Pulling that capability back, even temporarily, indicates that the edge cases involved are still difficult enough to affect commercial operations.
Why freeway driving matters so much
In dense metro areas, self-driving cars on local streets alone can offer useful transportation, but the network is less competitive when it cannot use high-speed roads. Waymo’s own rollout of highway rides had helped cut travel times in the Bay Area, particularly for cross-peninsula trips that could otherwise take 45 minutes to more than an hour on surface roads. Freeway access can therefore shape whether robotaxis feel like a premium convenience or a serious substitute for human-driven rides.
That makes the new suspension more than a technical footnote. Waymo is trying to expand into more cities around the world this year and has set a goal of offering as many as one million paid rides per week by the end of 2026. Scaling to that level requires more than safe operation in ideal conditions. It requires reliable performance in the messier parts of real transportation systems, including road closures, lane shifts, temporary barriers, emergency activity and weather-related disruptions.
Construction zones remain a hard autonomy problem
Waymo did not cite a single incident as the direct trigger for the pause, but the company’s robotaxis have been seen struggling in highway construction zones. Those environments are particularly difficult for automated systems because they can override the normal logic that maps, road markings and routing software depend on. Lanes move unexpectedly, cones redirect traffic, police or construction workers may give hand signals, and visual cues often conflict with the permanent road geometry encoded in a vehicle’s understanding of the environment.
Human drivers can also find construction zones confusing, but they tend to improvise based on context. A self-driving system must interpret ambiguous signals in real time and choose a safe action with very high confidence. If it hesitates, takes the wrong lane, or misreads a temporary boundary, the result can quickly become hazardous at freeway speed. That combination of ambiguity and speed is one reason construction zones continue to test the limits of autonomous driving systems.
The report also referenced a May 19 social-media post claiming a Waymo ride drove through cones and was then followed by police. While the company did not tie its current action to that episode specifically, the public visibility of such incidents adds pressure. For robotaxi operators, reputation and regulatory trust are tightly linked to software performance, especially in rare but high-risk road situations.
Part of a broader pattern of service interruptions
The freeway suspension follows other recent operational pullbacks. Waymo has also paused service in Atlanta and San Antonio after vehicles drove into flooded streets. Last week, the company announced a software recall intended to help its fleet avoid flooded areas in San Antonio while it worked on a more permanent solution. At least one vehicle was also reported stuck in Atlanta this week.
Viewed together, these pauses show a company still advancing rapidly but encountering the same reality that has defined autonomous driving for years: the hardest problem is not ordinary driving, but unusual conditions that arrive suddenly and break assumptions. Flooded roads and construction zones differ technically, yet both expose how difficult it is for robotaxis to generalize safely when infrastructure stops looking or behaving normally.
What this means for the robotaxi market
The setback does not erase Waymo’s lead in commercial autonomy, but it does narrow the gap between proof of concept and full transportation utility. The company is still operating on city streets in major markets and is testing its Zeekr-built robotaxi, known as Ojai, ahead of expected service in the coming months. Those are signs of continued momentum. Still, the new suspensions demonstrate that scaling a robotaxi network is not simply a matter of adding more vehicles to more places.
Instead, expansion depends on solving scenario-specific weaknesses one by one without letting them undermine public confidence. For riders, a freeway pause may register mainly as longer trips. For the industry, it is a reminder that autonomy remains a software business defined by edge cases. A fleet can appear mature until a new road configuration, weather pattern or infrastructure failure exposes behavior that is not ready for routine deployment.
A pragmatic retreat rather than a collapse
Waymo’s decision to suspend freeway rides can be read as a conservative operational move rather than a sign of broad system failure. Companies that deploy autonomous vehicles commercially need to narrow their operating domains when conditions reveal a weakness, then widen them again once software improves. That is the practical rhythm of deployment in a field where safety expectations are extremely high.
But the timing matters. Waymo is trying to scale quickly, broaden its geographic footprint and normalize robotaxi use in everyday transportation. Every temporary retreat complicates that story. The company may well resolve the construction-zone issue and restore freeway service soon, but the episode reinforces a central truth about autonomous mobility in 2026: the technology is real, commercially active and useful, yet still constrained by the hardest corners of the real world.
This article is based on reporting by TechCrunch. Read the original article.
Originally published on techcrunch.com
