ZYT, a Chinese autonomous driving startup spun from DJI, announces its AI foundation model can drive better than its CEO on Shenzhen’s crowded streets, trained on diverse real-world data from drones to vacuum cleaners. This development signals a potential shift in autonomous driving technology amid China’s national AI push and intensifying global competition.
In a bold claim that could reshape autonomous driving, ZYT—a Shenzhen-based startup spun from drone giant DJI—says its upcoming “mobility foundation model” AI system already outdrives its own CEO on the notoriously complex streets of Shenzhen. CEO Shen Shaojie, 39, told Reuters that during test drives, the AI navigates narrow roads with oncoming traffic and school zones more competently than he does, with engineers remarking, “We don’t know what the car is thinking.”
The AI That Learns From Everything
Unlike conventional autonomous systems that rely on specialized modules for detecting specific objects like traffic lights or pedestrians, ZYT’s foundation model learns from a vast, heterogeneous dataset. It was trained on video not just from road-driving cars but also from drones, robots, household vacuum cleaners, motorcycles, and even humans carrying moving cameras. This approach, Shen says, enables the AI to operate across vehicle types and geographies without being finely tuned for specific road conditions.
“The model is thinking in its own internal brain,” Shen explained, highlighting a departure from modular architectures. This could allow the system to control future autonomous robots or diverse devices beyond passenger vehicles. The technology is set for a public demonstration at the Beijing auto show in April.
Context: China’s AI Ambitions and Global Competition
ZYT’s unveiling coincides with China’s state-backed push to embed AI across its economy under Xi Jinping’s “new productive forces” initiative—a strategic effort to counter U.S. technology restrictions that target dual-use technologies. The startup enters a fiercely competitive arena dominated by Tesla‘s Full Self-Driving team and Chinese rivals like Xpeng, Huawei’s smart driving unit, and Momenta.
Shen acknowledges the breakneck pace: “If you can get six months of advantage, that’s already a huge thing.” ZYT’s edge stems from its DJI heritage; the drone maker retains a stake via an affiliate but faces U.S. sanctions over national security concerns. A recent deal saw state-owned automaker FAW Group acquire a 35.8% stake, becoming ZYT’s largest shareholder and resolving compliance issues for international customers by distancing ZYT from DJI.
From Trucks to Global Expansion: Strategic Moves
ZYT is prioritizing commercial trucking first, where immediate cost savings are clearer. The AI adapted from passenger car data to heavy-duty trucks in just six weeks, offering “low single-digit (percentage) savings” on fuel, according to Shen. The startup has partnerships with five of China’s six largest truck manufacturers—controlling over 98% of the domestic market—including recent deals with XCMG, SHACMAN, and SINOTRUK for highway-based systems.
Internationally, ZYT has established engineering and compliance operations near Volkswagen’s Wolfsburg headquarters, testing a FAW Hongqi prototype on European roads. VW, partnered with FAW in China, was ZYT’s first customer. However, the United States is explicitly excluded from ZYT’s roadmap: “We will keep ourselves away from the market at this moment,” Shen said, noting global uptake elsewhere.
Hardware Hurdles and the 2027 Passenger Car Launch
The current foundation model runs on expensive, high-powered computing hardware akin to robotaxis—unsuitable for mass-market cars. ZYT is actively compressing the AI for cheaper, mass-produced chips, though Shen admits this work is “still ongoing.” The first passenger car integration is projected for 2027, aligning with ZYT’s target for a Hong Kong listing as early as that year, leveraging commercial partnerships and the FAW investment.
Why This Matters: Beyond a CEO’s Test Drive
Shen’s personal anecdote isn’t just PR; it exemplifies a broader shift toward AI models that generalize from diverse, real-world data rather than curated simulations. If ZYT’s system scales reliably, it could lower development costs and accelerate deployment across logistics, robotics, and transportation. For China, successful domestic AI integration in critical infrastructure like autonomous driving would advance the “new productive forces” agenda while reducing reliance on foreign technology.
However, challenges remain. The leap from controlled demos to widespread safety validation is formidable, and ZYT’s avoidance of the U.S. market limits its global scale. Competitors like Tesla and Xpeng are also iterating rapidly, with Tesla recently unveiling its own vision-based approach. ZYT’s claim of a six-month lead is significant but fragile in a sector where software updates and regulatory approvals can alter landscapes overnight.
The Road Ahead
ZYT’s story encapsulates the volatile fusion of AI innovation, geopolitical strategy, and automotive disruption. By training on data from everyday devices, the company bets on a more adaptable, perhaps more human-like, driving intelligence. Whether this translates into safer, cheaper autonomous fleets will depend on rigorous testing and regulatory acceptance—but the message to the industry is clear: the next leap may come from unexpected data sources, not just bigger neural networks.
For the fastest, most authoritative analysis on breaking technology and business news, trust onlytrustedinfo.com to deliver immediate depth and clarity on the developments that matter.
