Xiaomi unveils custom 3nm chip for intelligent driving

Xiaomi unveiled the Xring D100, an in-house 3nm intelligent-driving chip intended to replace Nvidia hardware starting in 2027. The company has invested over 21bn yuan (~$3.1bn) since restarting its chip program and is “designing Nvidia out” of its vehicles, signaling growing supply-chain and cost-control strategy. Near-term impact is likely more incremental, but it modestly improves the long-term competitive outlook for Xiaomi’s auto roadmap.

Analysis

The market should treat this less as an immediate Nvidia revenue event and more as evidence that China’s premium EV stack is localizing faster than expected. The first-order P&L hit to NVDA is likely negligible, but the second-order effect is more important: once one high-profile OEM proves it can own the ADAS compute layer, procurement teams at other Chinese automakers get a template to dual-source or design-out foreign silicon over the next 12-36 months.

For Xiaomi, the economics are about control, not just chip cost. Owning the silicon roadmap can widen vehicle gross margin and reduce dependency on U.S. vendors, but only if the chip clears automotive qualification, yield, and functional-safety hurdles; the failure mode is a delayed launch that burns R&D without changing BOM. The 2027 timing makes this a long-dated option on vertical integration, not a near-term earnings lever.

Contrarian view: consensus may overreact on NVDA while underappreciating the broader China localization trend. Automotive is too small to move Nvidia’s model by itself, but repeated design wins disappearing in China would gradually compress the market’s willingness to pay for its auto growth narrative. The thesis is falsified if Xiaomi slips the program, retains Nvidia as a fallback, or if other Chinese OEMs keep buying Nvidia despite the precedent.

Key Decisions for Investors

• Do not short NVDA on this headline alone; any knee-jerk 1-2% gap-down is a better opportunity to add to longs than to initiate a structural short, because the auto segment is too small and the cash-flow impact is years away.
• Buy XIACY on weakness over the next 1-3 months as a long-duration vertical-integration option, but keep size modest; upside is margin control and stack ownership, while the key risk is a 2027 delay or failed qualification.
• Set a catalyst watch on Xiaomi’s next two earnings calls: if management reiterates 2027 SOP and names additional vehicle platforms, the strategic value rises; if not, treat this as an expensive R&D project rather than a monetizable advantage.
• If the market starts pricing a broader China OEM migration away from Nvidia, consider a small relative-value long XIACY / short NVDA basket, but only after confirmation from another Chinese automaker or a guide-down to NVDA auto-design-win commentary.
• Falsifier for the China-localization thesis: any public indication that Xiaomi will keep Nvidia as backup compute, or that the new chip is limited to lower-end driver-assist functions rather than full-stack intelligent driving.