Li Auto, BYD lead China EV makers into custom AI chip race
Synopsis
China's top EV makers — Li Auto, BYD, Nio, and Xpeng — are all building proprietary AI chips for autonomous driving, with Li Auto's new 5nm Mach M100 hitting 1,280 TOPS and BYD's 4nm Xuanji A3 already in mass production, threatening the dominance of Nvidia and Horizon Robotics in the world's largest auto market.
Key Takeaways
Li Auto unveiled the Mach M100 , a 5nm AI chip on 16 June 2026 , delivering 1,280 TOPS at an 82% utilisation rate for its L9 Livis SUV .
BYD debuted the Xuanji A3 , a 4nm chip already in mass production, capable of supporting Level 3 and Level 4 autonomy; three units in tandem exceed 2,100 TOPS .
Nio has introduced its own 5nm NX9031 chip for intelligent-driving applications.
Xpeng is developing the proprietary Turing chip for next-generation smart-driving systems.
The parallel in-house chip push across multiple Chinese automakers signals a structural shift away from reliance on external suppliers such as Nvidia and Horizon Robotics .
China's electric-vehicle industry has opened a new technological front, with leading carmakers designing proprietary smart-driving chips to reduce dependence on foreign silicon and sharpen their competitive edge in the world's largest auto market. The shift signals that custom AI silicon is fast becoming a defining battleground in the EV sector.
Li Auto's Mach M100 sets a new benchmark
On Monday, 16 June 2026, Li Auto unveiled the Mach M100, a 5-nanometre artificial-intelligence chip purpose-built for autonomous driving. Designed to power the company's new L9 Livis SUV, the chip delivers a single-unit computing output of 1,280 trillion operations per second (TOPS) — a standard measure of how rapidly an AI processor can handle incoming sensor data. Li Auto also said the chip achieves an 82 per cent utilisation rate, a figure the firm described as highly efficient.
BYD's Xuanji A3 already in mass production
The Li Auto announcement followed a significant move by world-leading EV maker BYD, which debuted its Xuanji A3 chip weeks earlier. Built on a 4-nanometre process node, the Xuanji A3 is already in mass production and is capable of supporting Level 3 and Level 4 autonomous-driving functions. When three Xuanji A3 chips are deployed in tandem, the combined processing power exceeds 2,100 TOPS, according to the company.
Why TOPS matter for autonomous vehicles
Higher TOPS ratings allow a vehicle's onboard computers to make split-second driving decisions by simultaneously processing data streams from multiple sources, including radar and lidar sensors. As autonomous-driving systems grow more sophisticated, raw compute throughput has become a critical differentiator between competing platforms.
Nio and Xpeng join the in-house silicon push
Nio has introduced its own 5nm NX9031 chip, while Xpeng is actively developing its proprietary Turing chip for next-generation intelligent-driving systems. The parallel efforts across multiple Chinese automakers suggest the move toward custom silicon is an industry-wide strategic pivot rather than an isolated bet by any single player.
What's next
With BYD, Li Auto, Nio, and Xpeng all advancing in-house chip programmes, the pressure on external suppliers — including Nvidia and domestic players such as Horizon Robotics — is set to intensify. Analysts and investors will be watching whether these proprietary chips can match third-party solutions on performance, cost, and reliability at scale, and whether the trend accelerates consolidation among China's smart-driving chip ecosystem.
Point of View
Li Auto, Nio, and Xpeng are not merely engineering projects — they are a direct response to the US-led semiconductor export controls that have made access to cutting-edge foreign chips unpredictable and expensive. What mainstream coverage underplays is that vertical integration of AI silicon gives these automakers a structural cost advantage and a data moat: proprietary chips can be tuned to each company's unique sensor stack and training data, compounding over time. This mirrors the playbook Apple used with its M-series chips — once you control the silicon, you control the performance roadmap. The biggest losers in this trend are not just Nvidia but also domestic Chinese chip vendors like Horizon Robotics, who now face commoditisation pressure from the very customers they once supplied.
NationPress
1 Aug 2026
Frequently Asked Questions
What is the Li Auto Mach M100 chip?
The Mach M100 is a 5-nanometre AI chip developed by Li Auto specifically for autonomous driving, unveiled on 16 June 2026 . It delivers 1,280 TOPS of computing power and operates at an 82% utilisation rate, powering the company's L9 Livis SUV .
How does BYD's Xuanji A3 chip compare to Li Auto's Mach M100?
BYD's Xuanji A3 is built on a 4nm process — one node smaller than Li Auto's 5nm Mach M100 — and is already in mass production. Three Xuanji A3 chips working together can exceed 2,100 TOPS , compared with the 1,280 TOPS delivered by a single Mach M100 .
Why are Chinese EV makers building their own AI chips?
Designing proprietary chips allows Chinese automakers to optimise performance for their specific sensor setups, reduce costs, and limit exposure to supply-chain risks tied to foreign chip suppliers. It also gives them greater control over their autonomous-driving technology roadmaps.
Which other Chinese EV companies are developing custom smart-driving chips?
Beyond Li Auto and BYD , both Nio and Xpeng are building in-house silicon. Nio has introduced the 5nm NX9031 chip, while Xpeng is developing the Turing chip for next-generation intelligent-driving systems.
What does this mean for Nvidia and Horizon Robotics?
As China's top EV makers shift to custom silicon, external chip suppliers — including Nvidia and domestic player Horizon Robotics — face the prospect of losing major automotive customers. The trend could intensify competition and pricing pressure across the Chinese smart-driving chip market.
