Xiaomi is expanding its in-house chip development with three new Xring processors, including the Xring O100, a dedicated AI chip built to handle advanced AI tasks directly on devices without relying heavily on cloud servers.
Xiaomi founder, chairman and CEO Lei Jun introduced the new chips during the company’s Autumn flagship product launch event.
Built for High-Speed AI
The Xring O100 offers up to 1.22TB/s of memory bandwidth. Xiaomi describes it as the world’s first 6nm 3D wafer-level stacked advanced packaging solution.
The chip uses hybrid bonding technology with an extremely small connection pitch. Xiaomi says this allows the O100 to deliver 16 times more memory bandwidth than conventional smartphones and reach AI inference speeds of up to 330 tokens per second.
3D Design Improves Data Transfer
The O100 uses 3D stacking technology, placing processing logic and high-speed DRAM directly on top of each other through wafer-on-wafer technology and hybrid bonding.
This design shortens the distance between memory and processing components, allowing data to move faster. Xiaomi says this helps solve one of the biggest challenges of running large language models directly on devices.
Even when a device has enough processing power, AI performance can be limited by slow communication between memory and the processor. The O100’s high bandwidth is designed to reduce this limitation and improve local AI response speeds.
Multi-Core NPU for AI Workloads
The Xring O100 features a multi-core NPU designed to handle different stages of AI processing efficiently.
It also includes dynamic data routing, which helps distribute AI workloads based on the type of task being performed.
Xiaomi has already tested the chip in prototype tablets and the Xiaomi AI Cube mini-PC. These devices combine the O100 with other Xring processors to support long-running AI workloads locally.
Commercial Devices Coming in 2027
Xiaomi expects the Xring O100 to appear in commercial products starting in 2027.
Possible applications include smartphones, tablets, robots and other AI-powered devices that require powerful local processing.
The O100 is part of Xiaomi’s broader Xring chip strategy, which also covers high-end mobile processors and automotive chips.
Instead of focusing only on smaller chip manufacturing processes, Xiaomi is also investing in higher memory bandwidth and advanced chip packaging.
The company aims to make generative AI faster, more private and available even without an internet connection.
| Feature | Xring O100 Details |
|---|---|
| Chip Type | Dedicated AI Accelerator |
| Manufacturing Process | 6nm |
| Memory Bandwidth | Up to 1.22TB/s |
| AI Speed | Up to 330 tokens per second |
| Memory Technology | 3D wafer-level stacked DRAM |
| Architecture | Multi-core NPU |
| Data Technology | Dynamic data routing |
| Target Devices | Smartphones, tablets, robots and AI devices |
| Prototype Devices | Tablets and Xiaomi AI Cube mini-PC |
| Expected Launch | 2027 |
| Main Goal | Faster and more private on-device generative AI |