🔥36氪•Stalecollected in 8m
First 'Token Factory' launched in Wuxi, China
💡China's new 'Token Factory' model integrates power and computing for sovereign AI infrastructure.
⚡ 30-Second TL;DR
What Changed
The project is a collaboration between Suihong Huachuang and Huawei.
Why It Matters
This development highlights the push for sovereign AI infrastructure and the integration of energy-intensive data centers with local power grids.
What To Do Next
Monitor the availability of localized computing clusters if you are building enterprise AI applications in China.
Who should care:Enterprise & Security Teams
Key Points
- •The project is a collaboration between Suihong Huachuang and Huawei.
- •It focuses on the localization of AI computing power and 'computing-electricity' integration.
- •The facility is designed to support the 'National Core, National Model, National Use' strategic initiative.
- •Located in Wuxi, it aims to boost the AI industry ecosystem in the Yangtze River Delta.
🧠 Deep Insight
Web-grounded analysis with 18 cited sources.
🔑 Enhanced Key Takeaways
- •The 'Token Factory' concept involves converting electricity, particularly low-cost green energy from western China, into 'tokens,' which are the fundamental computational units for AI models, enabling the export of AI computing services.
- •This initiative leverages China's abundant and affordable renewable energy resources in its western regions, leading to significantly lower AI inference costs (estimated at one-tenth to one-sixth of overseas models) and positioning China as a competitive global provider of AI computing services.
- •Huawei's involvement includes providing its Ascend series AI chips and developing large-scale computing clusters, such as Atlas SuperPoDs and SuperClusters, which utilize proprietary high-speed interconnects to combine numerous chips, aiming to achieve competitive aggregate AI computing power despite international chip manufacturing limitations.
- •The project is a tangible implementation of China's 'East Data, West Computing' strategy, which aims to balance regional development by shifting energy-intensive data processing to western provinces rich in renewable energy, while eastern hubs like Wuxi focus on low-latency applications and industrial integration.
- •The 'National Core, National Model, National Use' initiative, supported by this factory, is part of a broader national plan to achieve secure and reliable supply of core AI technologies by 2027, foster a world-leading open-source AI ecosystem, and deeply integrate AI into the manufacturing sector.
🛠️ Technical Deep Dive
- Huawei's AI computing power relies on its Ascend series chips, including models like Ascend 910C, 950, and 960.
- The company employs a strategy of 'scaling out' by binding hundreds or thousands of Ascend accelerators into large logical training or inference systems.
- This is achieved using a proprietary high-speed fabric alongside standard PCIe and RoCE networking to create systems like the CloudMatrix 384 and Atlas 950/960 SuperPoDs and SuperClusters.
- This clustering approach aims to achieve competitive aggregate throughput, compensating for potential individual chip performance differences compared to leading Western counterparts.
- 'Tokens' are defined as the smallest information units for AI models, with the generation of one million tokens consuming approximately 0.1 to 0.3 kilowatt-hours of electricity.
- The 'computing-electricity integration' strategy involves strategically locating AI training and batch inference tasks in western regions with abundant, low-cost green power (solar, wind, hydropower), while eastern hubs handle low-latency applications.
🔮 Future ImplicationsAI analysis grounded in cited sources
China will significantly enhance its global competitiveness in AI services by leveraging low-cost green energy.
The 'Token Factory' model, integrated with the 'East Data, West Computing' strategy, converts cheap western renewable energy into AI computing power (tokens), offering a substantial cost advantage (1/10th to 1/6th of overseas models for inference) for AI services globally.
China will achieve greater digital sovereignty and reduce reliance on foreign AI chip technology.
Huawei's strategy of building large-scale computing clusters using domestically controlled components, combined with national initiatives like 'National Core, National Model, National Use,' aims to create a self-sufficient AI infrastructure.
The 'Token Factory' model will drive regional economic rebalancing and industrial upgrading in China's inland provinces.
By converting abundant, low-cost energy in western regions into high-value AI computing services (tokens), the initiative aims to attract investment, create jobs, and foster advanced technology development in historically underdeveloped areas.
⏳ Timeline
2017-08
China issues 'New Generation Artificial Intelligence Development Plan'.
2018-09
Huawei launches its AI strategy and full-stack, all-scenario AI portfolio.
2021-09
Huawei introduces its Pangu models for industrial intelligence.
2022-01
China's 'East Data, West Computing' initiative is first proposed.
2025-08
China's State Council publishes 'Opinions on Deepening the Implementation of the Artificial Intelligence+ Initiative'.
2026-05
First 'Token Factory' launched in Wuxi, China.
📎 Sources (18)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
- Google Search Source
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗