China built hundreds of AI data centers to catch the AI boom. Now many stand unused (AI column)
Interesting. article Overproduction of data centers in China from MIT Technology Review. The point is,
- China’s tech sector is facing serious challenges overcapacity crisis before 80The percentage of newly built computing resources for artificial intelligence remain unused. A massive initiative to build infrastructure for AI, backed by both government and private investors, is experiencing a dramatic decline.
- The scale of the crisis is great - for 2023-2024 Over the years, China has announced more 500 new data center projects, including at least 150 Completed facilities are now desperately looking for customers. The state invested $6,1 Billion in construction of large computing clusters, and private companies added another $28 billion
- The crisis is caused by several Key factors:
- Technological upheaval. The emergence of the open-source DeepSeek R1 model, comparable in capabilities to ChatGPT, but requiring significantly less resources for training, radically changed market demand. Focus shifted from large-scale model training to inference Strategic miscalculations, where many data centers are built: Companies without sufficient experience in AI infrastructure In remote regions to save on electricity and land, although modern AI tasks require proximity to technology hubs to minimize delays Overestimating long-term demand for computing power
- The crisis led to economic consequences Equipment rental prices have collapsed: clusters of eight Nvidia H100 GPUs are now being rented out 75 000 yuan ($10 333) in a month, although previously 180 000 Yuan.
- Such a development is likely will lead to the following effects
- Government intervention: The Chinese government will likely be forced to intervene to stabilize the situation.
- Market consolidation: A wave of mergers and acquisitions is expected among data center operators, as smaller players will not be able to withstand financial pressure.
- Reorientation of infrastructure: Existing facilities will be adapted to new market realities – from massive training clusters to distributed systems for inferencing.
- Environmental challenges: Even with the current level of data center load, their environmental impact is increasing.
This news shows that large-scale Chinese investments in fast-moving technologies are accompanied by great risks, and not always the cards are right:) Interestingly, I’ve written about books that primarily describe the successes of China and its companies.
- The book.Tech Titans of China (The world in the dragon's shadow)An overview of the venture history of China from the early 2000s to the twentieth
- The book.Attention Factory: The Story of TikTok and China's ByteDance (TikTok. Attention Factory. Takeoff history)" - History of the rise of ByteDance Smart Business: What Alibaba's Success Reveals about the Future of Strategy (How Alibaba is using artificial intelligence) An interesting book by Zeng Ming, a key strategist at Alibaba Group (2006–2017)Former Chief of Staff Jack M. My review in two parts: 1 and 2 Alibaba: The House That Jack Ma Built (Alibaba. The history of the world ascent from the first person) An interesting story about the rise of Alibaba in several parts: 1, 2 and 3 The Digital Transformation of China ("The Digital Economy") From Ma Huateng, the founder of Tencent -- I thought the book would feature more Tencent and Wechat, but it turned out not, but I'm still talking about it. written
- The book "Baidu" ("Artifical Intelligence Revolution. How AI Will Change our Society, Economy, and Culture") - book 2017 About Baidu by Robin Lee, CEO and Chairman of the Board of Directors. Interesting topic, translation of G.
- The book "Haier purpose." The real story of China's first global super company ("Evolution of Haier.") - interesting book The evolution of this technological giant from China
#Management #Processes #Leadership #BusinessStory #Engineering #Software #Engineering