Tesla Walks The Chinese Tightrope

Plus: swimming Waymos, and Uber’s plan for Waymo-independence

This is the Ride AI newsletter: The most comprehensive weekly digest of news and intelligence at the intersection of technology and transportation.

ICYMI, early bird tickets for Ride AI 2027 are $500 through May 31. If you’d like to apply for an early spot at the show, please do so here.

Top Story

PC: EV-Tony on YouTube

Tesla has officially confirmed that FSD Supervised is now available in China. The announcement came one week after President Trump’s state visit to Beijing, where Musk attended as a business delegate.

For context, Tesla has been trying to get FSD into China since at least 2023. Musk has made multiple predictions about imminent approval, only to be contradicted by Chinese government sources. Tesla’s partnerships with Baidu (for mapping) and a local Shanghai data center were part of its effort to comply with China’s strict data laws. How Tesla will resolve training data export controls (all data must stay within China’s borders) while US chip export bans prevent building large-scale compute centers there will be the key question going forward. China’s rapidly advancing ADAS market, dominated by BYD and Huawei who already offer ADAS as standard even on their cheapest vehicles, means Tesla can’t afford to wait much longer.

Along with the announcement, Tesla has once again renamed the FSD feature on its Chinese website to “Tesla Assisted Driving.” This seems to be the result of continued Chinese state-mandated name changes meant to align feature names with actual feature capabilities.

Outside China, Tesla continues FSD’s European expansion. Lithuania became the second EU country to approve FSD Supervised, recognizing the Dutch RDW certification through mutual recognition rather than independent testing. Belgium, Greece, and Ireland are moving toward approvals as well, though EU-wide approval faces resistance from Scandinavian regulators over FSD’s tendency to exceed speed limits and its performance on icy roads. A formal vote on approval isn’t expected before July.

Domestic News

PC: Luke Younts/Storyful

Waymo paused service in four cities, Atlanta, San Antonio, Dallas, and Houston, after its robotaxis kept driving into flooded roads despite a software recall issued the previous week. This isn’t Waymo’s first rodeo with stubborn behavioral issues. The company is already under two separate NHTSA and NTSB investigations, one for robotaxis illegally passing stopped school buses in Austin, and another from a January incident where a Waymo vehicle struck a child in Santa Monica. NHTSA says it is in communication with Waymo and will take appropriate action if necessary.

ECARX and May Mobility announced a strategic framework agreement to scale an autonomous ride-hail fleet together, in a deal estimated at approximately $750 million. Under the deal, ECARX will develop up to thousands of autonomy-enabled vehicles with a custom L4 central computing platform and full sensor suite (LiDAR, radar, cameras, and IMUs) for May Mobility’s next-generation system. The two companies are targeting a 50% cost reduction by 2028, with commercial scale-up also in 2028. May Mobility has already completed more than half a million commercial autonomous rides across the US and Japan with partners including Toyota, Lyft, Uber, and Grab, so training data is certainly not lacking. It seems May has also hit the point in the hardware dev cycle where the technology works, but cost has become the main hurdle.

PC: Balaji Krishnamurthy, Uber CFO

Uber’s CFO has revealed that its new AV Labs effort has a long-term vision to outfit its millions of drivers with sensor kits, turning Uber’s driver network into a rolling data collection grid for AV companies. Right now, AV Labs runs a small dedicated sensor fleet separate from Uber’s driver network, but the intent is to eventually scale to the full driver base. Uber currently has 25 AV company partners and is building an “AV cloud,” a library of labeled sensor data that partners can query to train their models. Partners can also run trained models in “shadow mode” against real Uber trips to simulate AV performance without deploying a vehicle.

More data collection for AV training is all well and good, but how much of it will actually be quality data that can be used in the training dataset? Data collection technicians are usually trained to follow a specific set of rules when driving for data gathering. Absent those rules, the gathered data may not be usable. I don’t know about you, dear reader, but I’ve experienced my fair share of “yikes” moments in a large amount of my Uber rides. And beyond that, sometimes the particular driver’s driving style is just plain uncomfortable. I’m not so sure “uncomfortable”, being a very subjective condition, is that easy to filter…

International News

PC: Wayve/Stellantis

Stellantis and Wayve announced a strategic technology partnership to integrate Wayve’s AI Driver into the STLA AutoDrive platform, targeting Level 2++ hands-free, door-to-door automated driving across highway and urban environments, with a first vehicle integration planned in North America in 2028. Notably (and similarly to Nuro’s project with Lucid), Wayve says it brought up a prototype on Stellantis vehicles in under two months, which certainly reinforces the company’s claims of how well its end-to-end AI generalizes across different vehicle platforms. The partnership builds on a prior Stellantis investment in Wayve and is part of Stellantis’ broader strategy to partner with third-party AI companies rather than develop its own autonomy stack, a strategy a lot of traditional vehicle OEMs seem to be taking.

PC: Xpeng

Xpeng CEO and company namesake He Xiaopeng doubled down on the company’s camera-only strategy, arguing lidar is no longer necessary even as the proportion of Chinese vehicles above 150,000 yuan equipped with lidar continues to rise. Xpeng’s new GX SUV, which racked up 24,863 firm orders in the first 12 hours, launched without lidar. The technical team argues that end-to-end large model technology requires the richness of visual signals, not the sparse point clouds lidar provides. Meanwhile, on the other side of the debate, Hesai reported its fourth consecutive quarter of GAAP profitability in Q1 2026, with revenues up 29.6% year-over-year and lidar shipments surging 141% to nearly 472,000 units. Hesai also announced it has become the strategic lidar supplier for Mercedes-Benz’s L3 autonomy programs in both Europe and China, and unveiled Picasso, what it calls the world’s first 6D full-color ultra-sensitive lidar chip, targeting mass production in the second half of 2026. All this having just unveiled the world’s first RGB color sensing lidar, a sensor package meant to replace RGB cameras. Make of that what you will.

Baidu’s Apollo Go delivered 3.2 million fully driverless rides in Q1 2026, a year-over-year increase of more than 120%. Baidu’s broader AI-powered business revenue grew 49% year-on-year to RMB 13.6 billion (about $2 billion USD), now 52% of its general business total. The robotaxi numbers are the clearest indication yet that fully driverless commercial operation is scaling in China at a pace that deserves serious attention, despite the Chinese government having just recently suspended issuing new operation permits.

Robot Heartwarmers

The New York Times has published a piece in which blind Waymo users explain how delightful it is to ride alone in a driverless vehicle. I encourage you to give it a read, not just because of how uplifting it is, but because it describes precisely how driverless technology can enable the less able-bodied in our society to regain their independence, and how empowering that experience can be.

Alright, that’s it from me… until next week. I’m back on the newsletter from here on out. Much thanks to Mike for taking over while I was out 🙏 If you enjoy this newsletter, share it with your friend, colleague, or boss. Thank you for reading; Sophia out!