Chinese Automakers' In-House Chip Development Reshapes Supply Chain, but Regulatory and Manufacturing Dependencies Remain
Miles Bennett
BYD unveiled the 4 nm Xuanji A3, billing it as China's first 4 nm automotive-grade autonomous-driving chip, while NIO, Li Auto, and XPeng push parallel in-house efforts — yet advanced-node fabrication still depends on TSMC, leaving a structural gap that caps both cost savings and geopolitical exposure.
Why are automakers insisting on designing their own chips?
The most immediate reason is cost: NIO previously spent roughly $300 million a year buying chips from Nvidia. In-house design burns cash upfront, but unit cost drops as volume scales.
This means → chip self-design is essentially a bet on economies of scale — the more cars sold, the cheaper each chip becomes.
The second driver is supply-chain security: a single smart vehicle uses thousands of chips. Relying on Nvidia, Mobileye, and other external suppliers exposes automakers to geopolitical tension, price swings, and shipping delays.
In plain terms = designing your own chip keeps the "blueprint" in your hands, so no one else can cut you off overnight.
What deeper motives lie behind in-house chip efforts?
Customization: off-the-shelf chips don't always fit a carmaker's proprietary software stack. BYD says the Xuanji A3, tightly integrated with its own algorithms, doubles computing efficiency and cuts energy consumption per unit of compute by 20%.
This means → it is not just "more powerful" — it is "same workload, less power drawn." For an EV, saving watts is saving range.
Physical-AI platform play: autonomous driving and humanoid robots share heavy overlap in AI chips, sensors, motors, and batteries. Automakers treat chip R&D as a platform strategy spanning multiple hardware categories.
XPeng's Turing AI chip (7 nm) has received mass-production certification from Volkswagen Group. This reflects a potential path beyond internal use — selling the technology externally.
Where does each company's chip stand today?
BYD Xuanji A3: 4 nm, positioned as China's first 4 nm automotive-grade AD chip, currently fabricated by TSMC.
NIO NX9031: 5 nm; Li Auto M100: 5 nm; XPeng Turing AI chip: 7 nm.
This means → Chinese automakers are approaching the design frontier, but sub-5 nm advanced-node manufacturing remains a weak link — they can draw the blueprint, but the production line cannot yet keep up.
What does "can design but can't fabricate" really mean?
The core contradiction: advanced-node wafer fabrication still depends on external foundries such as TSMC. This means → automakers achieve design autonomy while manufacturing-side geopolitical risk stays intact.
In plain terms = in-house chip design solves the question of "whose blueprint," but the question of "who runs the factory" remains unanswered.
This reflects a structural gap in China's semiconductor industry: design capability is catching up fast, but manufacturing capability is still constrained by equipment and process technology.
Can these chips actually reach overseas markets?
BYD's overseas sales already account for 43% of volume (July figure), but the Xuanji A3 targets L3-to-L4 autonomous driving — and regulatory approval for that level varies widely by country.
Most BYD models currently sit at L2+. The Xuanji A3 is expected to support the "L3 Ready God's Eye 5.0" system with urban navigation, offered with a one-year safety guarantee in the Chinese market.
This means → hardware capability ≠ market access: no matter how powerful the chip, a car cannot take the road without a driving permit. Regulatory pace — not chip performance — is the binding variable for BYD's overseas AD rollout.
Content is for reference only, not financial advice.