Skip to main content

Xiaomi Unveils Xring O3 and Stacked AI Accelerator in Shift Away from Qualcomm

25 AUGUST 2026·2 MIN READ·5 SOURCES·Independently corroborated

Xiaomi is expanding its custom silicon portfolio with a new 3nm mobile SoC and a 6nm AI accelerator designed for on-device large-model inference, aiming to reduce reliance on third-party chipmakers.

Xiaomi Unveils Xring O3 and Stacked AI Accelerator in Shift Away from Qualcomm

Key takeaways · 3

  • 01

    Xiaomi's 3nm Xring O3 mobile chip will debut in the Xiaomi 18 Fold this September.

  • 02

    The 6nm Xring O100 accelerator leverages 3D stacking to deliver 1.22TB/s of near-memory bandwidth.

  • 03

    The new 3nm XRING D100 chip targets autonomous driving and supports local deployment of 200-billion-parameter models.

The New Xring Lineup

Xiaomi is seeking to reduce its reliance on longtime suppliers Qualcomm and MediaTek by launching the 3nm Xring O3 system-on-chip for its mobile devices. [2] The company introduced the hardware at a chip technology briefing in Beijing on August 24 as part of an effort to extend its in-house silicon design from smartphones to AI computing and intelligent driving. [4] According to Xiaomi, the XRING O3 recorded an AnTuTu score of more than 5.22 million and is set to debut in the Xiaomi 18 Fold in September. [4] Cumulative shipments of devices using the previous-generation XRING O1 have already surpassed one million. [4]

O100 Accelerator Architecture

Xiaomi also unveiled the XRING O100, a 6nm AI accelerator designed for on-device large-model inference. [4] The accelerator is meant to boost its MiMo models when paired with its O3 mobile chip. [3] Unlike a general-purpose system-on-chip, the 6nm accelerator uses wafer-on-wafer 3D stacking and hybrid bonding to vertically integrate a logic die with two layers of AI-focused DRAM. [4] The architecture utilizes face-to-face metal interconnection, a 1.4-micron bonding pitch, and 0.7-micron through-silicon vias to shorten the path between memory and compute. [4] Xiaomi claims the design delivers 1.22TB/s of near-memory bandwidth and supports inference speeds of up to 330 tokens per second when running its MiMo 3B model. [4]

D100 and Commercial Rollout

During the same event, the company introduced the 3nm XRING D100 for autonomous driving. [3] The D100 chip, which is aimed at intelligent-driving applications, integrates a 20-core CPU and a 16-core NPU. [4] Xiaomi says the XRING D100 architecture supports up to 160GB of unified memory and local deployment of models with more than 200 billion parameters. [4] Both the O100 and D100 have completed development and are expected to enter commercial use next year. [4] Meanwhile, the company's new Xring O3 chip reportedly hides a 60-micron 'OK' gesture physically embedded on the silicon. [5]

What it means

Xiaomi’s simultaneous launch of mobile, AI accelerator, and automotive chips signals an aggressive vertical integration strategy aimed at bypassing third-party silicon providers. By explicitly moving away from Qualcomm and MediaTek, the company is positioning its XRING line to control the entire hardware stack for its next-generation devices. The use of advanced packaging techniques like wafer-on-wafer 3D stacking in the O100 demonstrates an ambition to match the high-bandwidth memory architectures typically reserved for data centers, bringing them directly to local smartphone inference. What the sources don't address: How developers will optimize third-party applications for Xiaomi's proprietary high-bandwidth matrix bus and NPU architectures.

Xiaomi's move into advanced, specialized silicon for mobile and automotive platforms demonstrates the growing trend of device manufacturers pursuing vertical integration. The hardware focuses on eliminating memory bottlenecks to enable massive on-device AI inference.

Why it matters
Daily session

Turn this story into practical AI skill after launch.

Get the release link for daily sessions built around your role and industry.

Join the waitlist

How this developed

  1. 25 August 2026

    Xiaomi Unveils Xring O3 and Stacked AI Accelerator in Shift Away from Qualcomm

  2. 25 August 2026

    Event created from source cluster.

  3. 24 August 2026

    Event created from source cluster.

Sources

AI fluency, one session a day, built for your work.