Zijin Semiconductor, a company incubated by Great Wall Motor (GWM), has crossed the milestone of shipping over 10 million automotive-grade chips, according to Chinese outlet First Electric. The achievement is notable because it puts the open-source RISC-V instruction set into mass production, challenging ARM's long-standing dominance in the automotive chip market.
The most symbolic product in this shipment is the M100, described as China's first automotive-grade MCU (microcontroller unit) designed on an open-source RISC-V core. It entered mass production and was installed in vehicles on December 18, 2025, handling body control functions such as combination headlights, ambient lighting, and combination switches.
At the recent Sixth Intelligent Vehicle Chip Ecosystem Conference, Xu Xingkui, chief strategy officer and assistant to the president at GWM Hunan Zijin Semiconductor, highlighted the transformation of cars from mechanical devices to mobile intelligent terminals. He noted that a single vehicle now contains around a thousand chips, with chip costs rising from about 5% to 20% of total vehicle cost. However, the industry has long been based on the ARM architecture, creating a de facto closed ecosystem. Xu compared this to renting a house versus building your own: with a fixed instruction set, companies can only do 'interior decoration,' while open-source RISC-V allows 'independent construction.'
Domestic chips have historically lacked opportunities to be used in vehicles, and a lack of trust has further limited their application, creating a vicious cycle. The collaboration model is also changing. Previously, automakers dealt with Tier 1 suppliers, who in turn dealt with chipmakers in a chain structure. Now, automakers are deeply involved, working directly with chipmakers to co-define chips around specific scenarios and metrics, turning the chain into a network. Xu believes this gives RISC-V an opportunity, as carmakers seek differentiation rather than limited fixes on someone else's instruction set.
Technology and localization
Zijin's approach can be seen on two levels. Technically, the company says it has developed more than 15 core IP blocks, with complete capabilities from design and simulation verification to tape-out and mass production. It has built a software toolchain covering MCAL, low-level drivers, real-time operating systems, and development debugging environments, and has passed ISO 26262 ASIL-D process certification, ASIL-B product certification, AEC-Q100, ISO 9001, and ASPICE certifications.
In terms of implementation, Xu said the M100's IP, manufacturing, and packaging and testing are all localized in China. GWM and Mande Optoelectronics participated in its definition and development. The chip underwent more than 800 tests at the chip level, over 1,400 tests by Mande Optoelectronics, and nine major vehicle-level tests. Zijin has more than 15 cooperative customers, and the customizable features have shortened development cycles by about 40% and brought mass production forward by nearly 12 months.
The product lineup is expanding. The M200 is planned for tape-out in the fourth quarter of 2026, using a 40nm process, with applications extending to combination instruments, T-BOX, gateways, BCM, shock absorbers, and stabilizer bars. The M300 targets regional controllers with high computing power, large capacity, and multiple I/O. The A100 series of power and driver chips will complement the lineup.
Currently, the volume products are concentrated in lower-barrier segments like body control, power, and drivers, while high-value chips for intelligent driving and cockpit main control remain dominated by ARM. RISC-V in automotive is still early-stage. Xu acknowledged that technical maturity, software ecosystem, and architecture fragmentation are challenges, with standard-setting being an urgent priority. He predicts RISC-V and ARM will coexist long-term, noting that Infineon, NXP, and Renesas have all made moves in this area. In May, Zijin, together with GWM, established a RISC-V industry innovation center to fill gaps in standards and testing specifications.
Next stop: embodied intelligence
After ramping up automotive business, Zijin has set its next RISC-V target on embodied intelligence. This direction leverages GWM's ecosystem. In 2025, GWM signed a strategic cooperation agreement with Unitree Robotics, covering robot technology R&D, smart manufacturing upgrades, and intelligent vehicle innovation, focusing on embodied intelligence underlying technology, 'vehicle + robot' scenario innovation, and vehicle manufacturing upgrades. The initial focus is on 'off-road' scenarios, exploring collaboration between vehicles and robots in complex environments.
At the chip level, Zijin is developing its first 'cerebellum control' chip, the E100. Based on the RISC-V architecture with a modular, scalable design, its computing power ranges from low-power joint control to high-performance main control, aiming to provide a computing foundation for robots. The logic is that automotive-grade functional safety and reliability experience can be reused. Robots come in many types, iterate quickly, and are sensitive to cost and power consumption, which aligns with RISC-V's open-source, free, and customizable characteristics.
The 'cerebellum' manages movement: multiple joints need real-time coordination, balance must be maintained, and latency between sensing and execution must be kept extremely low. This parallels the real-time and functional safety requirements of automotive chassis and body control. In off-road scenarios, vehicles and robots could potentially divide tasks, and such collaboration is precisely the entry point for automotive-grade chipmakers to cross into robotics.
The market growth is indeed substantial. Industry data shows the global RISC-V-based automotive MCU market was about $291 million in 2024, projected to grow to $1.232 billion by 2031. In 2025, global embodied intelligence chip shipments were about 800,000 units, with the market expected to approach 20.84 billion yuan (about $2.94 billion) by 2032, a compound annual growth rate of 42.2%.
Automotive MCUs are a proven replacement market, while embodied intelligence chips are still pre-explosion: 2025 global shipments were only about 800,000 units, and the E100 is not yet in mass production. How much synergy 'vehicle + robot' can generate and when orders will materialize remain to be seen. Xu said at the conference that RISC-V's open sharing, open-source free nature, high customizability, and support for AI-native custom instructions are the basis for entering intelligent driving, cockpits, and robotics. But software tools, compatibility, and reliability verification systems are still inadequate. He expects more RISC-V automotive chips to emerge in the next three to five years. In the short term, robot chips are unlikely to contribute significant revenue, being more of a future-oriented layout.
The broader context is the gap in domestic substitution. Xu cited data that last year China's automotive production accounted for about 70% of the global total, yet domestic chip share was only around 10-15%. Zijin hopes to help push this share from about 10% to over 35%, and plans to move toward a platform-type supplier through integration and mergers. From one company achieving mass production to the entire industry achieving self-reliance, the standards, ecosystem, and trust still need to be built step by step.