China’s 2026 Beijing Auto Show opened on April 24 with a clear message: the next battlefield for Chinese EVs is no longer just electrification, but AI-defined mobility. Geely used the event to unveil Eva Cab, described as China’s first natively developed Robotaxi prototype, while autonomous driving supplier Zhuoyu Technology introduced what it calls the industry’s first native multimodal foundation model for “mobile physical AI.” Alongside them, German supplier ElringKlinger underscored how global component makers are deepening their China strategy, betting that the country will remain the center of EV, battery, and thermal management innovation.
Geely Unveils Eva Cab, Its Native Robotaxi Prototype
Geely’s biggest headline in Beijing was the launch of Eva Cab, a purpose-built Robotaxi prototype developed around the company’s All-Domain AI 2.0 architecture. Rather than adapting an existing passenger EV, Geely is framing Eva Cab as a vehicle conceived from the ground up for autonomous mobility services.
Key highlights include:
- L4 autonomous driving target for public-road shuttle operations
- More than 3,000 TOPS of onboard computing power
- A chip stack combining NVIDIA SuperChip, NVIDIA Thor U, and Qualcomm Snapdragon 8797
- A 2160-line digital lidar with up to 25.92 million points per second
- Claimed maximum detection range of 600 meters
- Planned 2027 launch of a customized CaoCao Mobility version
Geely said Eva Cab will run its Qianli Haohan G-ASD L4 software stack, which is intended for scenarios such as hotel transfers, airport shuttles, and business mobility services. The company also emphasized that Robotaxi validation is already underway: through CaoCao Mobility, Geely says it has been trialing Robotaxi services in Hangzhou and Suzhou for over a year.
That matters because many Robotaxi announcements remain concept-heavy. Geely is trying to present a more complete stack: vehicle, software, chips, operating platform, and ride-hailing ecosystem.
The AI Architecture Behind Geely’s Strategy
Behind Eva Cab is a wider effort to turn Geely from a carmaker into what it describes as a global intelligent mobility technology company. The company said its Xingrui computing center has reached 23.5 EFLOPS, which it claims is the highest among Chinese automakers.
Geely’s AI push includes:
- All-Domain AI 2.0 spanning powertrain, chassis, cockpit, and driver assistance
- The WAM world action model, designed as a unified “vehicle brain”
- A 1+2+N agent architecture linking cockpit, assisted driving, and multiple domain controllers
- Super Eva, a cockpit-driving fusion agent already slated for production use
- Qianli Haohan G-ASD 4.0, which Geely says will roll out to Zeekr, Lynk & Co, and Geely Galaxy models
In practical terms, Geely is signaling a trend now visible across the Chinese EV industry: the software-defined car is evolving into an AI-native vehicle, where autonomous driving, voice, perception, cloud connectivity, and even cybersecurity are developed as one integrated system.
Zhuoyu Technology Wants to Build the “Mobile AI Base” Layer
If Geely represents the OEM side of this transition, Zhuoyu Technology is pitching itself as a core enabler across multiple vehicle categories. At the show, the company launched its native multimodal foundation model, positioning it as the next step beyond current end-to-end autonomous driving systems.
Zhuoyu’s central argument is that intelligent driving must become both:
- Mass-market affordable, across different price bands
- Platform-agnostic, spanning passenger cars, trucks, buses, logistics vehicles, and Robotaxis
The company says it has already:
- Mass-produced solutions across more than 50 passenger vehicle models
- Won design nominations for more than 100 models
- Partnered with 34 customers
- Reached cooperation on more than 130 vehicle programs
This scale matters because it suggests Zhuoyu is no longer a niche ADAS developer. It is trying to become a horizontal AI mobility platform supplier.
From End-to-End ADAS to a Multimodal Foundation Model
Zhuoyu divides the autonomous driving evolution into three stages:
- Small-model systems relying on perception models, HD maps, and rule-based planning
- Mid-sized end-to-end models, now common in advanced Chinese ADAS stacks
- Large native multimodal models with stronger zero-shot transfer and cross-domain generalization
According to the company, its current end-to-end 4.0 model will begin OTA rollout from April 2026 to vehicles using Qualcomm 8650 and 8775 chips. For lower-compute platforms based on TI TDA4-VH, Zhuoyu plans upgrades toward its end-to-end 3.0 system.
The more important long-term move, however, is the new multimodal foundation model. Zhuoyu says it can process:
- Video
- Text
- Actions
- Voice
- Maps
It also claims the model is pre-trained not only on intelligent driving data, but also on broader internet and robot mobility datasets. The goal is ambitious: reduce the amount of localization and re-engineering required when moving across countries, road rules, and vehicle types.
That could be a major competitive advantage if it works in production. One of the biggest bottlenecks in autonomous driving today is not proving a demo in one city, but scaling safely across many geographies and use cases without rewriting the stack each time.
Commercial Vehicles and Robotaxi Are Becoming Core Battlegrounds
One of the most interesting signals from Beijing was how quickly Chinese autonomous driving firms are moving beyond passenger EVs.
Zhuoyu says it has already partnered with all top six Chinese heavy truck brands, with production deliveries of its commercial-truck end-to-end 4.0 system due to begin in June 2026. It has also signed a strategic cooperation agreement with Yutong Bus, with bus programs scheduled for mass production in September 2026.
For commercial vehicle applications, Zhuoyu highlighted:
- A cabin-mounted Jimu 2.0 lidar-vision fusion system
- Variable zoom sensing for urban and highway environments
- Highway NOA for heavy trucks
- Coverage for highway NOA, urban NOA, and autonomous parking in buses
Meanwhile, Zhuoyu’s unmanned mobility roadmap extends to:
- Urban logistics vehicles, with trial operations planned for July 2026
- L4 Robotaxi, with pilot operation expected in H2 2026
- A triple-redundant L4 controller based on dual NVIDIA Thor chips
This broader expansion is important. In China, heavy trucks, buses, logistics fleets, and Robotaxis increasingly serve as testbeds for autonomous systems because they offer repeatable routes, commercial ROI cases, and fleet data loops that private passenger cars often cannot match.
Supplier Spotlight: ElringKlinger Bets on China’s EV Scale
While AI stole much of the spotlight, ElringKlinger offered a reminder that the EV race is also being shaped by materials, battery components, and thermal management. The German supplier said China is no longer just a sales market for the company, but a critical platform for developing and industrializing next-generation vehicle technologies for global use.
ElringKlinger highlighted its China footprint in:
- Changchun
- Suzhou
- Qingdao
- Tianjin
At Auto China 2026, the company showcased:
- Cell contacting systems for EV battery packs
- Lightweight thermoplastic and hybrid material components
- An integrated instrument panel crossbeam combining metal forming and injection molding
- ElroForm TP-ECO solutions for battery thermal insulation and safety
- A new-generation thermal management seal for EVs, hybrids, and ICE platforms
The battery connection system is particularly notable. ElringKlinger said it began large-scale production in Europe in 2025, and is now accelerating global deployment. In an era where battery pack efficiency, thermal stability, and manufacturing precision are becoming key differentiators, these “invisible” components are increasingly strategic.
Key Technologies Compared
| Company | Main Reveal | Key Data Points | Strategic Focus |
|---|---|---|---|
| Geely | Eva Cab Robotaxi prototype | 3,000+ TOPS, 2160-line lidar, 600 m range, 2027 rollout target | AI-native Robotaxi and mobility ecosystem |
| Zhuoyu Technology | Native multimodal foundation model | 50+ mass-produced models, 100+ nominated models, 34 customers, 130+ vehicle programs | Scalable autonomous driving across multiple vehicle classes |
| ElringKlinger | EV battery and thermal solutions | 2025 large-scale production of cell contacting systems in Europe; China R&D/manufacturing hubs in 4 cities | Localized development for China and global EV markets |
Why This Matters for the Chinese EV Market
Several deeper trends emerged from these announcements.
1. China’s EV competition is shifting from electrification to AI stack depth
Battery range and 800V architectures still matter, but automakers are increasingly competing on:
- Compute power
- Foundation models
- Sensor fusion
- OTA iteration speed
- Ecosystem integration
2. Robotaxi is moving from concept to industrial planning
Geely’s 2027 commercialization target and Zhuoyu’s H2 2026 pilot timetable show that Chinese companies are treating L4 not as a distant moonshot, but as a near-term product category.
3. Commercial vehicles may become the fastest real-world autonomy market
Heavy trucks, buses, and logistics fleets offer clearer economics and more controlled deployments than consumer cars. That makes them attractive stepping stones for Chinese autonomous driving suppliers.
4. Global suppliers are localizing around China, not just selling into it
ElringKlinger’s positioning reflects a broader reality: international auto suppliers increasingly see China as the place where EV technologies are refined, scaled, and then exported globally.
Global Implications
For global automakers and suppliers, the Beijing Auto Show news offers a sharp signal. China is no longer merely leading in EV volumes; it is trying to define the next architecture of intelligent mobility.
Geely’s push suggests Chinese OEMs want to own the full AI stack, from compute infrastructure to Robotaxi operations. Zhuoyu’s approach shows how Chinese tech suppliers are building cross-segment autonomous driving platforms that could eventually travel well beyond China. And ElringKlinger’s strategy confirms that even established European suppliers see Chinese EV development as central to future competitiveness.
If these efforts succeed, the global market may soon face a new kind of Chinese auto challenger: not just low-cost EV exporters, but AI mobility system companies with integrated hardware, software, data, and operations.
What Comes Next
The next 12 to 18 months will be critical. Geely must prove that Eva Cab can move from a headline-grabbing prototype to a safe, economically viable Robotaxi platform. Zhuoyu must show that its multimodal foundation model can scale across regions and vehicle categories with lower adaptation costs than today’s end-to-end systems. And suppliers like ElringKlinger will need to keep pace as EV thermal management, battery packaging, and lightweighting requirements grow more demanding.
What is already clear, though, is that Beijing Auto Show 2026 marked a turning point. China’s EV industry is entering a phase where the car is no longer just electric and connected—it is being rebuilt as an intelligent mobile robot, supported by a rapidly maturing supply chain and an increasingly global technology ambition.
