Xiaomi is making its most aggressive bet yet on artificial intelligence, committing more than 60 billion yuan (US$8.8 billion) to AI investments over the next three years. The announcement, made at an investor day in Beijing on April 27, positions the company as a serious contender in the AI race, not just as a hardware manufacturer but as a full-stack technology company with ambitions spanning software models, custom silicon, and electric vehicles.
MiMo-V2.5-Pro Reaches the Top of the Agentic Benchmark
The investment is already producing results. Xiaomi’s open-source AI model, MiMo-V2.5-Pro, has been ranked by the third-party benchmark platform Artificial Analysis as the world’s leading open-source model for agentic capabilities, the ability to autonomously plan and execute complex, multi-step tasks. In overall intelligence and coding evaluations, the model placed second and third respectively, trailing only DeepSeek and Moonshot AI.
This is a striking achievement for a company that only began releasing open-source AI systems approximately one year ago. The rapid progress reflects both the scale of Xiaomi’s engineering resources and the broader acceleration of China’s open-source AI ecosystem, which has been gaining ground on Western counterparts. Xiaomi’s entry into the top tier of open-source model rankings also signals that the competitive landscape for AI models is no longer defined solely by dedicated AI labs, hardware companies with large engineering teams and proprietary data from their device ecosystems are becoming formidable players.
XRing O1: Competing With Apple on Silicon
Beyond software, Xiaomi is making significant strides in custom chip design. The company’s self-developed XRing O1 processor, built on a 3-nanometer lithography process, integrates 19 billion transistors on a single system-on-chip. Xiaomi has stated that the XRing O1 matches Apple’s A18 Pro in single-core and multi-core tests and outperforms it in certain benchmark categories. The chip is currently deployed in the Xiaomi 15S Pro smartphone and the Xiaomi Pad 7 Ultra tablet.
Developing a competitive mobile chip from scratch is an undertaking that very few companies globally have achieved. Qualcomm, MediaTek, and Apple are the dominant players, and each has spent decades building the engineering expertise and supply chain relationships required to produce leading-edge silicon. Xiaomi’s success with the XRing O1 reduces its dependence on third-party chip suppliers and gives the company greater control over the performance and power efficiency of its flagship devices — a strategic advantage that becomes more valuable as AI inference workloads increasingly run on-device rather than in the cloud.
The Human × Car × Home Ecosystem
Xiaomi’s AI strategy is inseparable from its broader hardware ecosystem. CEO Lei Jun described the company’s “Human × Car × Home” framework as the ideal foundation for AI integration, envisioning a world where AI seamlessly connects personal devices, vehicles, and home appliances. The company delivered more than 350,000 electric vehicles in 2025, including the SU7 sedan, starting at 219,900 yuan, and the YU7 SUV at 253,500 yuan, and has been building out its automotive software capabilities to match. This builds on earlier moves, such as recruiting top Tesla China executives to accelerate delivery goals.
Over the next five years starting in 2026, Xiaomi has committed to investing 200 billion yuan in core technology research and development, a figure that underscores the company’s ambition to compete not just as a hardware assembler but as a technology creator. Phate Zhang, founder of CnEVPost, credited China’s EV supply chain as “the most mature in the world,” a foundation that Xiaomi is leveraging to integrate AI into its vehicles at a pace that Western competitors are struggling to match.
Hayden Hou, a principal analyst at Canalys, framed the stakes clearly: “AI is the most critical factor for all tech consumer goods companies.” For Xiaomi, the 60-billion-yuan AI commitment is not a luxury, it is a prerequisite for remaining competitive in a market where the hardware-software boundary is dissolving and where the companies that control the full stack will increasingly set the terms of competition.
Open-Source as a Competitive Strategy
Xiaomi’s decision to release MiMo-V2.5-Pro as an open-source model reflects a deliberate strategic choice rather than an act of altruism. By making its models publicly available, Xiaomi builds developer goodwill, attracts external contributors who improve the model, and establishes a presence in the global AI ecosystem that purely proprietary approaches cannot achieve. The strategy mirrors that of Meta with its Llama series, which has become the dominant open-source model family in the West, and of Alibaba with Qwen, which has achieved similar dominance in Asia.
For a hardware company like Xiaomi, open-source AI models also serve a specific commercial purpose: they lower the barrier for third-party developers to build applications that run on Xiaomi devices, expanding the ecosystem that makes the company’s hardware more valuable. As competition among Chinese AI labs intensifies and model capabilities converge, the ability to build and sustain a developer community may be as important as benchmark performance in determining which companies emerge as long-term winners.