Xiaomi unveils 3nm Xring chips, challenging Qualcomm and Nvidia

Xiaomi unveiled three in-house chips — a 3nm mobile SoC, a 6nm AI accelerator, and a 3nm smart-driving processor — a move that cuts reliance on Qualcomm and Nvidia silicon across its phones, cars, and AI devices.

"The Xring O3 is a dedicated project completed over 459 days," Lei Jun, founder, chairman and CEO of Xiaomi, said at the launch event, calling it the first flagship SoC to surpass 5 million points on AnTuTu benchmarks.

The Xring O3, built on TSMC's 3nm process with 24 billion transistors, packs a 10-core all-big-core CPU and a 16-core G2-Ultra NX GPU, delivering 200 TOPS of tensor performance from a rebuilt 4-core NPU and supporting LPDDR6 RAM. The chip scored 5.22 million on AnTuTu, a 26 percent transistor increase over the prior generation, and debuts next month in the Xiaomi 18 Fold foldable and Xiaomi Pad 9 Pro Max. The Xring O100, a 6nm AI accelerator, uses Wafer-on-Wafer packaging with vertical chip-and-memory stacking to reach 1.22 TB/s bandwidth for on-device AI workloads across phones, vehicles, and robotics. The Xring D100, a 3nm smart-driving chip with a 20-core CPU and 16-core NPU, supports 160GB of unified memory and runs large language models up to 200 billion parameters locally; it is scheduled for commercial deployment in vehicles by 2027.

The announcement escalates a race among Chinese EV makers to build in-house silicon. Nio's 5nm Shenji NX9031, already in production with over 300,000 units shipped, delivers more than 1,000 TOPS; Li Auto's 5nm Mach M100 hits 1,280 TOPS per chip; and BYD's 4nm Xuanji A3, in mass production, reaches 2,100 TOPS across a three-chip cluster. Xiaomi has invested more than 21 billion yuan ($3.1 billion) in chip development since restarting the program and now employs nearly 3,000 chip engineers. Xiaomi shares fell 4.1 percent on the announcement day, though the drop coincided with a gross margin miss in its Q2 results.

Xring O3 targets Qualcomm's flagship turf

The O3's 5.22 million AnTuTu score and 200 TOPS NPU put it in direct competition with Qualcomm's Snapdragon flagship line and Apple's A-series chips. The 3nm node (which packs more transistors per square millimeter, improving performance per watt) gives Xiaomi a process advantage over rivals still on 4nm. The chip's 10-core all-big-core CPU architecture departs from the traditional big.LITTLE design, prioritizing sustained multi-core performance over power efficiency. Multi-core benchmark scores exceeded 15,000 points for the first time, a 60 percent improvement, according to Lei Jun.

D100 enters a crowded smart-driving field

The D100's 3nm process gives it a node advantage over Nio's 5nm Shenji NX9031, Li Auto's 5nm Mach M100, and BYD's 4nm Xuanji A3. Xpeng has also deployed its Turing chip in its own vehicles and some Volkswagen models in China. But Xiaomi has not disclosed the D100's TOPS rating, making a direct performance comparison difficult. The company's EVs currently rely on Nvidia Thor chips for smart-driving systems, having previously used Nvidia Orin. The D100's 2027 commercial timeline means Xiaomi will continue depending on Nvidia for at least another year. Xiaomi's EV and AI business posted a 2.6 billion yuan operating loss in Q2, driven partly by higher core component costs and increased AI spending.

The AI Cube Prototype, an engineering mini PC integrating all three chips, runs 120 billion and 3 billion parameter models simultaneously while switching between fast and slow operating systems, delivering up to 150W of sustained performance. The demonstration shows Xiaomi's ambition to create a unified silicon platform across its smartphone, automotive, and AI hardware businesses — a strategy that could reduce procurement costs and improve hardware-software integration.

Xiaomi trades as an OTC ADR (XIACF, XIACY) and on the HKEX (1810.HK). The chip program's 21 billion yuan investment and 3,000-person team represent a significant bet that in-house silicon will lower costs and differentiate products. But the 2027 timeline for the D100 and the lack of disclosed TOPS data mean investors cannot yet quantify the competitive impact. The Q2 operating loss of 2.6 billion yuan in the EV and AI segment shows the cost pressure the chip program adds in the near term.

This article is for informational purposes only and does not constitute investment advice.