ZYT autonomous Driving AI has reached a milestone that its own CEO describes with a mix of professional pride and personal humility the system his company built now drives better than he does on the crowded, unpredictable streets of Shenzhen. Shen Shaojie, 39, founder and chief executive of ZYT, a spin-off from Chinese drone giant DJI, told Reuters that the company's soon-to-be-unveiled mobility foundation model navigates narrow roads with oncoming traffic, manages school zones with children nearby, and handles the dense urban complexity of one of China's most congested cities with a confidence and precision that he cannot match behind the wheel himself. The system will make its public debut at the Beijing auto show in April.
The claim would sound like marketing hyperbole from almost any other source. From a CEO who has spent years building the system and whose engineers delivered a sobering reminder during a test drive telling him directly that they do not know what the car is thinking it reads more like an honest technical assessment of how far AI-powered driving has come. ZYT's foundation model does not work the way most autonomous driving systems work, does not depend on the geographic and market-specific training data that conventional systems require, and has already been deployed in partnerships covering five of China's six largest truck manufacturers. This is not a prototype chasing a dream it is a commercial system with real customers, real revenue logic, and a Hong Kong listing targeted as early as 2027.
The broader context in which ZYT is operating makes its ambitions more significant than a single company story. China is in the middle of a national push to embed artificial intelligence across every sector of its economy under President Xi Jinping's drive to develop what the government calls new productive forces a strategic priority designed to build technological dominance in fields where the United States is simultaneously trying to limit Chinese access. Autonomous driving sits at the centre of that competition, and ZYT is one of the most technically interesting competitors in a race that includes Tesla, Huawei's smart driving unit, Xpeng, and Momenta. The outcome will shape not just who builds the cars of the future but who controls the AI systems that drive them.
How ZYT Was Built and Why Its Approach Is Fundamentally Different
ZYT did not emerge from the automotive industry it emerged from the world of aerial robotics, as a spin-off from DJI, the Chinese company that built the global consumer drone market from scratch and in doing so developed some of the most sophisticated spatial computing and real-time navigation software in the world. That origin matters because it shaped the technical philosophy that ZYT brought to the autonomous driving problem. DJI's engineers had spent years teaching machines to navigate three-dimensional space using visual data, making real-time decisions in environments where errors have immediate physical consequences. That capability translated directly into the approach ZYT took when building its driving AI.
The conventional approach to building an autonomous driving system involves creating dedicated modules separate software components trained to detect cars, identify pedestrians, read traffic lights, and interpret road markings and then assembling those modules into a system that navigates based on their combined outputs. The limitation of this architecture is that it must be retrained and recalibrated for each new geographic market, each new road type, and each new traffic pattern. A system trained on California highways performs poorly on Shenzhen urban streets. A system tuned for European roundabouts struggles with Indian intersections. The modularity that makes these systems manageable to build also makes them expensive and slow to scale across different markets.
ZYT's foundation model takes a fundamentally different approach. Rather than training separate modules on specific detection tasks, the system was trained on an extraordinarily diverse dataset of visual experience not just dashcam footage from road driving but video from drones, robots, household vacuum cleaners, motorcycles, and people carrying moving cameras. By exposing the AI to navigation and spatial reasoning challenges across all those different platforms and perspectives, ZYT built a system with a generalised understanding of how to move through physical space that does not depend on market-specific calibration. Shen describes this as the key departure from how the industry has built and trained autonomous systems to date.
Training on Everything to Be Ready for Anything
The decision to feed the ZYT foundation model with video from vacuum cleaners and drones alongside conventional road footage is not an engineering curiosity it reflects a deliberate strategy to build the kind of broad, transferable spatial intelligence that allows the system to operate competently in environments it has never specifically been trained on. A vacuum cleaner navigating a living room is solving a fundamentally similar problem to a truck navigating a warehouse moving through a space while avoiding obstacles and completing a task efficiently. A drone managing airspace around buildings and people is processing spatial relationships and making real-time adjustments in ways that translate to urban driving at ground level.
This training philosophy produced a system with capabilities that conventional autonomous driving architectures struggle to match. ZYT's AI adapted its passenger car training data for heavy-duty truck operation in approximately six weeks a timeline that reflects the transferability of the foundation model approach rather than any shortcuts in safety validation. The same underlying spatial intelligence that learned to drive passenger cars on urban roads transferred to a fundamentally different vehicle class operating in different conditions without requiring months of dedicated retraining from scratch. For commercial customers evaluating autonomous driving technology, that adaptability represents a significant practical and economic advantage over systems that require full retraining for each new application.
The application potential extends beyond conventional vehicles to the broader category of autonomous machines that the robotics industry is building. Shen noted that the foundation model's generalised spatial intelligence could make it useful for controlling the movement of future autonomous robots and other devices a market that is developing rapidly as manufacturing, logistics, and service industries explore automation at scale. ZYT is positioning itself not just as an autonomous driving company but as a provider of generalised machine navigation intelligence, a framing that gives it a much larger addressable market than automotive alone and aligns it with China's broader ambitions in AI-powered robotics.
The DJI Sanctions Problem and How FAW Resolved It
DJI, which retained a stake in ZYT through an affiliate company called New Territory, has been operating under U.S. sanctions due to national security concerns raised by American government agencies. For ZYT, that DJI association created a compliance problem with potential customers outside China who needed to ensure their supply chains were free of sanctioned entities. The problem was not merely theoretical any major international automaker considering ZYT technology for vehicles sold in markets where U.S. sanctions have extraterritorial implications needed assurance that ZYT's ownership structure did not create regulatory exposure for their own operations.
The resolution came through FAW Group, a state-owned Chinese automaker and one of China's largest vehicle manufacturers, which purchased a 35.8 percent stake in ZYT from New Territory late last year. FAW now holds the largest single ownership stake in ZYT, with New Territory retaining 34.85 percent. Shen told Reuters that the transaction resolves compliance concerns for customers outside China because ZYT is no longer majority-owned by the DJI-linked holding company. The restructuring transforms ZYT's ownership profile from a DJI spin-off with sanctions adjacency into a state-backed Chinese automotive technology company a meaningfully different proposition for international commercial partners navigating their own regulatory environments.
Commercial Partnerships, Truck Markets, and the Road to a Hong Kong IPO
ZYT's most commercially significant current deployment is not in the passenger car market that generates the most public attention around autonomous driving it is in the Chinese trucking industry, where the financial case for advanced driving assistance is more immediate and more measurable than in consumer vehicles. Shen told Reuters that ZYT has partnerships with five of the six largest Chinese truck manufacturers, a group that collectively controls more than 98 percent of the domestic market. In January, ZYT announced highway truck driving system plans with three of those manufacturers XCMG, SHACMAN, and SINOTRUK for deployment in the first half of this year.
The commercial logic of the truck market is straightforward and compelling. Advanced driving systems in trucks can generate immediate, quantifiable savings on fuel consumption ZYT's system delivers low single-digit percentage savings on fuel costs according to Shen, a figure that translates into significant annual savings across large commercial fleets where fuel is one of the largest operating expenses. Unlike consumer autonomous driving, which faces complex questions about liability, public acceptance, and regulatory approval before generating revenue, truck automation on defined highway routes offers a cleaner commercial proposition with faster ROI and less regulatory friction. ZYT recognised this dynamic early and built its truck market position while competitors focused on the more visible but commercially harder passenger car segment.
The scale of ZYT's truck market penetration partnerships covering 98 percent of domestic Chinese truck manufacturing gives it a commercial foundation that many better-known autonomous driving companies lack. Revenue from fuel efficiency savings and driver assistance subscriptions across millions of Chinese trucks provides the financial sustainability to fund the continued development of the foundation model and the chip compression work required to bring the system to mass-market passenger vehicles. It also gives ZYT a dataset of real-world highway driving at commercial scale that feeds back into the foundation model's continuous improvement a virtuous cycle that makes the system better as it deploys more widely.
Volkswagen, FAW, and the Path Into European Markets
ZYT's first automotive customer was Volkswagen, whose primary China manufacturing partner is FAW the same state-owned automaker that now holds the largest stake in ZYT. That ownership and commercial relationship creates a direct channel for ZYT technology to reach Volkswagen vehicles globally, and ZYT has already established an engineering and compliance presence near Volkswagen's Wolfsburg headquarters in Germany. The company has been testing a prototype from FAW's Hongqi brand on European roads as part of its compliance and validation work for the European market, building the real-world data and regulatory documentation that European type approval requires.
The European market represents a significant strategic opportunity for ZYT precisely because it is one of the most technically demanding and regulatory complex markets in the world the kind of environment where a foundation model's geographic generalisation capability is most valuable. A system that requires market-specific retraining for every new geography would face significant barriers to European deployment given the continent's diversity of road types, traffic patterns, and driving conventions. ZYT's foundation model approach, if it delivers the cross-geography performance Shen describes, would give it a structural advantage over conventional autonomous driving architectures in exactly the markets where those architectures struggle most.
The United States, by contrast, is explicitly not on ZYT's current roadmap. Shen was direct with Reuters: ZYT will keep itself away from the U.S. market for now, while noting that the rest of the world is already picking up. That framing reflects both the regulatory and geopolitical realities of operating a Chinese AI technology company in the current U.S. policy environment and a strategic choice to build global market presence through Europe, Asia, and other markets before engaging with the complexity of U.S. regulatory approval, data sovereignty requirements, and the political scrutiny that Chinese autonomous driving technology would inevitably attract in Washington.
The Chip Problem and the 2027 Passenger Car Target
ZYT's foundation model currently runs on the kind of high-powered, expensive computing hardware found in robotaxis and research prototypes systems that cost tens of thousands of dollars per vehicle and are entirely unsuitable for mass-market passenger cars where consumers expect advanced driving features to be included in vehicles at accessible price points. The work of compressing a foundation model sophisticated enough that its engineers cannot fully explain its internal reasoning onto chips affordable enough for production vehicles is one of the central technical challenges ZYT is working to solve, and Shen acknowledged that this compression work is still ongoing.
The target is a first passenger car using the ZYT system in 2027, aligned with the timeline of the EU's DHC-515 aircraft delivery programme and ZYT's own Hong Kong IPO ambition a convergence of milestones in that year that suggests 2027 represents ZYT's moment of commercial maturation across multiple dimensions simultaneously. The chip compression challenge is not unique to ZYT every autonomous driving company building systems on high-performance hardware faces the same fundamental problem of making those systems economically viable at mass-market scale. But ZYT's foundation model architecture, which achieves broad capability through generalised training rather than dedicated module complexity, may give it advantages in compression that more architecturally complex systems do not have.
The Hong Kong listing targeted for 2027 would give ZYT access to public capital markets at a moment when its commercial partnerships, passenger car product, and international market presence are expected to be substantially more developed than they are today. Shen described the timeline as the potential quickest, indicating that 2027 is an optimistic target rather than a committed date but the combination of FAW backing, VW partnership, trucking revenue, and foundation model technology gives ZYT a more credible IPO story than most autonomous driving startups have been able to assemble. The question is whether the six-month competitive advantage Shen describes as already huge will still be intact when the Beijing auto show debut translates into production deployments at scale.