RISC-V, the open-source chip architecture that waives instruction-set licensing fees, is quietly entering Chinese cars through the least glamorous components first: headlight controllers, ambient lights and wiper switches. According to Chinese outlet D1EV, the first mass-produced automotive-grade RISC-V chip is already on the road, and the industry is betting that the low-cost architecture will climb from the edges of the car toward its computing core over the next few years.
From headlights to powertrains
GWM's Zijing Semiconductor, incubated by the automaker's technology center, has put its first RISC-V automotive chip, the M100, into body-control scenarios such as combination headlights, ambient lighting and switch modules, with up to 17 chips per vehicle. Zijing says the M100 entered mass production at the end of 2025, claiming it is the world's first mass-produced automotive RISC-V chip, and GWM plans to fit it in 2.5 million vehicles over five years.
Dongfeng's DF30, a RISC-V multi-core engine-control chip, has completed validation on several models and is scheduled for volume production in 2026. Since August 2024, multiple RISC-V automotive microcontrollers (MCUs) have entered production in China, with another wave due in 2026, focused on body, powertrain and electronic mirrors, according to speakers at an industry event covered by D1EV.
In bigger chips for assisted driving and smart cockpits, RISC-V currently appears mainly as auxiliary cores or dedicated acceleration units. Horizon's Journey 6 series uses some RISC-V cores in production, though the main CPU remains ARM-based. Mobileye's next-generation EyeQ Ultra goes further: all 12 general-purpose CPU cores inside use RISC-V, according to its official materials, but that remains a rare case.
Why are the big chips hard to switch? "Automakers would rather buy expensive overseas chips than risk switching," said Ye Qi, senior manager at SAIC's large passenger-car architecture division, quoted by D1EV. He noted that NXP, Infineon and Renesas control roughly 70% of the high-end MCU market and are deeply tied to AUTOSAR toolchains, while Qualcomm, MediaTek and Nvidia dominate assisted-driving and cockpit chips with high software and algorithm barriers. Swapping a chip means redoing hardware, base software and testing from scratch, he said.
The breakthrough points are low-risk parts: headlights, windows and wipers use many chips per car, are functionally independent, and a single failure does not threaten vehicle control. Validation scope is manageable, pin-compatible replacements are possible, and the low-end control segment is the most price-competitive zone, so the industry is taking a "from the edge to the center" path.
Costs beyond the free instruction set
The free part is only the instruction set itself. Commercial RISC-V cores, development tools and verification environments still cost money, though industry sources cited by D1EV say commercial RISC-V core licensing runs about one-tenth of ARM's, which ranges from millions to tens of millions of yuan depending on tier, plus royalties of 1-2% of chip selling price. The real value for automakers, D1EV argues, is breaking supplier lock-in and gaining a strategic "second option."
The bigger expense is the ecosystem. Infineon, the global leader in automotive MCUs, said it started building its RISC-V ecosystem two to three years before sample chips appeared: compilers, debuggers and virtual prototypes from 2022 to 2024, operating systems and base software in 2025-2026, and application-layer reference designs only after 2027. In March 2026 it released DRIVECORE, a virtual-prototype evaluation kit for RISC-V. A single automotive chip typically takes three to five years to complete functional-safety certification, and engineers who know both functional safety and the new architecture remain scarce, according to Li Weili, cited by D1EV.
That is why Quintauris, the automotive RISC-V alliance founded by Bosch, Infineon, NXP and Qualcomm, exists: to share the cost of maintaining base software and automotive real-time profiles. Automakers in China are also defining chips themselves. GWM's Zijing model lets the automaker set requirements directly, which Zijing says cuts development time by about 40%. Dongfeng's DF30 follows a joint-development route with domestic IP and chip firms. Changan is taking a middle path, using RISC-V's custom instruction extensions to build a low-latency matrix-computation accelerator for in-car AI inference, without replacing the main control chip.
Forecasts cited by Li Weili, from RISC-V International and other bodies, project RISC-V penetration in automotive will rise from about 10% in 2025-2026 to roughly 31% by 2031. Ye Qi sees 2026-2028 as the window when vehicle architectures shift to central computing plus zone control, with new-architecture penetration exceeding 30% by 2027. L3 regulation, urban NOA (city advanced driver assistance) and 800V platforms will pull chip demand in the same period, he said.
Executives at the event agreed the near-term relationship between RISC-V and ARM is complementary, not substitutive. Infineon's next-generation AURIX MCUs will move to RISC-V, but its ARM-based TRAVEO line continues to sell. Qualcomm dominates cockpit chips on ARM while co-founding Quintauris. The risk, D1EV notes, is fragmentation: if Chinese players build mutually incompatible extensions, the open architecture becomes a new kind of closed system. Infineon says Chinese standards should align with international ones, and Quintauris is already cooperating with Chinese core developer Nuclei. If the window closes before control chips reach scale and software ecosystems cross a critical threshold, RISC-V could remain stuck at the edge; if benchmark models and million-unit volumes materialize, its move toward the center may come faster than expected.
