Xiaomi Escalates AI Price War with 99% Cut — Chinese LLMs Massively Undercut Western Rivals
Summary
On May 27, 2026, Xiaomi announced a permanent API price reduction of up to 99% for its MiMo-V2.5 large language models. This move marks a significant escalation in the ongoing AI price war among Chinese tech giants and poses a major challenge to the economic models of Western AI companies like OpenAI and Anthropic.
What happened?
The Xiaomi MiMo platform announced that following the conclusion of the “100 Trillion Token Creator Incentive Plan,” API prices for the MiMo-V2.5 series would be permanently reduced by as much as 99%. This follows a broader trend in China, where companies like DeepSeek recently implemented permanent price cuts of 75%. Xiaomi’s move is currently the most aggressive pricing strategy in the LLM market.
Why it matters
The implications of this pricing strategy are profound:
- Agent Economics: The cost of running complex coding agents and agentic frameworks is set to drop dramatically.
- Western Budget Pressure: Enterprises using premium-priced Western models face increasing pressure to justify costs against significantly cheaper Chinese alternatives.
- Market Share Strategy: Chinese providers appear willing to operate at thin or negative margins to secure global market dominance in AI infrastructure.
Evidence
- Xiaomi MiMo Official Announcement (May 27): Permanent price cuts of up to 99%.
- AASTOCKS Reporting: Confirmed the massive reductions for XIAOMI-W (01810.HK).
- Market Sentiment: Discussions on platforms like Hacker News suggest that US-based AI companies may struggle to defend their high valuations in the face of such extreme price competition.
Analysis
Xiaomi’s strategy aims to lower the barrier to entry for its LLMs in global technology stacks. While Western providers focus on high-margin premium models, Chinese companies are effectively turning AI compute into a low-cost commodity. This could lead to a shift where open-source agent frameworks like Hermes or OpenCode increasingly rely on these ultra-cheap backends for large-scale operations.
Practical Takeaways
- Benchmarking: Developers should evaluate MiMo-V2.5 performance on tasks like SWE-bench to assess its viability compared to Western models.
- Cost Reduction: For compute-heavy agent workflows, these models offer unprecedented savings potential.
- Strategic Diversification: Companies should assess the feasibility of integrating Chinese models as cost-effective backends for specific, non-sensitive tasks.
Open Questions
- Will Western providers like OpenAI or Anthropic respond with their own price reductions?
- How will regulatory factors, such as export controls and data sovereignty, affect the adoption of these models in international markets?
- Is this level of aggressive pricing sustainable in the long term?