Nuro, the Silicon Valley autonomous-vehicle startup backed by Nvidia, Uber and Softbank, has begun testing its self-driving software on public roads in Tokyo. The trials use Toyota Prius vehicles fitted with Nuro’s autonomy stack and human safety operators riding as a backup while the company evaluates performance in a new driving environment. 🇯🇵🚗
What the Tokyo tests involve
Nuro’s Tokyo operations mark the company’s first overseas expansion. Vehicles equipped with Nuro’s software are running in real-world traffic while engineers collect data and validate system behavior. The company hasn’t disclosed fleet size or a timeline for removing human safety operators, but its public posts indicate these trials are an early step toward broader deployment. 🌏
Why Tokyo presents new challenges
- Left-side driving: Japan drives on the left, which means control logic and routing must adapt to reversed lane patterns.
- Dense urban traffic: Tokyo’s crowded streets, complex intersections and many road users raise edge-case scenarios that differ from U.S. roads.
- Different signage and markings: Road signs, markings and local traffic conventions require regional calibration and validation.
How Nuro’s AI approach works
Nuro builds its autonomy stack on an end-to-end AI foundation model that the company says learns as it drives. The firm calls this strategy “zero-shot autonomous driving,” claiming the model can operate in new environments — like Tokyo — without prior training on local driving data. That approach is similar to other startups using broad, generalizable AI for vehicle control. 🤖
Safety, validation and shadow mode
Nuro emphasizes safety throughout its testing process. Key steps include:
- Closed-course testing for each software release.
- Simulation and targeted edge-case evaluation before public deployment.
- On-road trials with human drivers in the vehicle while the autonomy system runs in shadow mode — the software predicts actions but does not control the vehicle.
- Post-drive analysis comparing what the model would have done versus human inputs to determine readiness for active control.
These layered checks are intended to ensure the system is validated before being allowed to control vehicles on public streets. 🛡️
Business pivot and funding history
Nuro was founded in 2016 by former Google self-driving engineers Dave Ferguson and Jiajun Zhu. The company originally focused on low-speed delivery robots and drew major investment — including a $940 million infusion from SoftBank Vision Fund in 2019. After facing high development costs and industry consolidation, Nuro pivoted in 2024 to licensing its autonomy software to automakers and mobility providers instead of operating its own low-speed fleet.
More recently, Nuro raised $203 million in a Series E round that included backers such as Baillie Gifford, Icehouse Ventures, Kindred Ventures, Nvidia and Pledge Ventures. Uber has also indicated plans for a significant investment tied to a broader strategic arrangement. 💼
What comes next
Nuro’s Tokyo tests are an early indicator of its ambition to deploy globally. The company has suggested further geographic expansion in public statements, but specific timelines and rollout plans remain undisclosed. For now, engineers will continue closed-course and shadow-mode validation while collecting data to refine the AI model for Japan’s unique driving conditions.
Key takeaways
- Nuro is testing its autonomous software on Toyota Priuses in Tokyo with human safety operators aboard.
- The company relies on a foundation AI model and a “zero-shot” claim to generalize to new driving environments.
- Safety validation uses closed-course tests, simulation and shadow-mode on public roads before enabling active control.
- TOKYO trials are Nuro’s first overseas move as it pursues licensing deals and broader deployment. 🌍
Stay tuned: these tests will be a key signal of whether Nuro’s AI-driven strategy can scale across different countries and traffic ecosystems.