The "Cooling Revolution" Driven by Surging AI Computing Power: How a Traditional Valve Manufacturer Seized the Opportunity?
Hello everyone, I'm your financial journalist. Today's news might seem overwhelming at first glance, with terms like "EFLOPS," "Capex," and "CDU," but beneath the jargon lies a very practical and even counterintuitive business story:
When the demand for AI computing power surges, it's not the chips that become the bottleneck, but the heat generated; and the solution to this problem comes not from high-tech companies, but from traditional manufacturing giants that have been producing water pipes and valves for decades.
Let me summarize the key points of the article for you, and then we'll break down this story in five easy-to-understand parts.
---
[Summary of Key Points]
With the explosion of AI technology, the world is racing to build more data centers (computing power centers). Data shows that the scale of computing power in China has nearly doubled in a year, and global cloud companies' budgets for building data centers have increased by 90%.
However, as chips become more powerful, they also generate more heat (from 400 watts to 2300 watts), and traditional air cooling methods are no longer effective. As a result, "liquid cooling technology" (using water or coolant to dissipate heat) has become a necessity, with its adoption rate rapidly increasing.
This presents an opportunity for traditional manufacturers. For example, Guanlong Energy Saving, a company that has been producing valves for decades, realized that liquid cooling systems require high-precision valves to control water flow and pressure. Instead of blindly entering the software industry, the company leveraged its expertise in fluid control. It focused on developing liquid cooling valves for AI data centers (a new growth area) while upgrading its traditional water valves with intelligent algorithms (upgrading its existing business).
In simple terms, AI generates a lot of heat, which requires liquid cooling; liquid cooling needs high-quality valves; traditional valve manufacturers have thus reaped the benefits of AI while using AI to transform their old business, adopting a robust transformation strategy that combines two approaches.
---
[In-Depth Analysis: Five Aspects of the Transformation]
1. **Why the sudden shift to liquid cooling? Because chips are getting extremely hot**
Many think the bottleneck in AI is insufficient computing power or a lack of chips. In reality, the biggest physical limitation is heat dissipation.
To illustrate: Traditional chips were like small light bulbs that could be cooled with a breeze; today's AI chips (such as NVIDIA's latest Rubin platform) are like small furnaces, consuming up to 2300 watts per card, with entire racks consuming hundreds of kilowatts. Traditional air cooling methods (using fans) have a physical limit of around 15-20 kilowatts of heat dissipation. It's like trying to extinguish a large fire with a small fan—impossible.
Therefore, liquid cooling is essential. It's like adding coolant to an engine to directly remove heat from the chips. Data predicts that by 2026, the use of liquid cooling in AI chips will exceed 50%. This means liquid cooling will no longer be an optional feature but a necessity. Whoever masters liquid cooling will control the infrastructure of AI.
2. **How can traditional valve manufacturers enter the AI market? Through technology transfer**
You might wonder: What does manufacturing water valves have to do with building AI servers?
This is where the concept of "technology transfer" comes into play. Although AI liquid cooling systems are advanced, they still rely on basic principles of fluid control—how water flows, the pressure, and when to open or close valves. Experts like Jiang Qifa mentioned in the article that the core components of liquid cooling systems, such as CDUs (Cold Distribution Units), cold plates, and pipelines, require high precision, leak-free performance, and corrosion resistance.
Traditional manufacturers like Guanlong Energy Saving, with decades of experience working with water, have a deep understanding of fluid properties, sealing techniques, and pressure control. This is like a chef who has been cooking Sichuan cuisine for 30 years; even if the tools change, their knowledge of cooking techniques remains valuable. They can quickly adapt their skills to new applications.
3. **Guanlong Energy Saving's strategy: A balanced approach with two tracks**
Many companies make the mistake of betting everything on AI when they see the opportunity. However, Guanlong Energy Saving's chairman, Li Zhenghong, is cautious. He realized that although the liquid cooling business is promising, it will only account for 20%-30% of the company's revenue in the future. The more competitive the market and the faster AI technology evolves, the greater the uncertainty.
Therefore, Guanlong adopted a dual-track approach:
- New track (offensive): Developing liquid cooling valves for AI data centers to tap into the growing market.
- Old track (defensive + upgrading): Modernizing traditional municipal water services, which, although less profitable, provides stability.
This strategy ensures that the company doesn't rely solely on one market and retains its core business.
4. **How Guanlong Energy Saving is upgrading its old business**
Guanlong's transformation of its old water business is also noteworthy. Previously, selling valves meant delivering a physical product and ending the transaction. Now, they've added sensors to valves to monitor pressure, flow, and switch states in real-time and transmit this data to systems. This is similar to how cars have evolved from having just steering wheels and accelerators to including autonomous driving and connectivity.
- Pain points in the industry: High leakage rates, energy consumption, and unstable pressure in urban water supply networks.
- Solution: By using smart hardware, algorithms, and digital twins, the company can optimize operations. For example, the system can automatically adjust valve openings to save water and electricity.
Under the "dual carbon" goals, energy efficiency is crucial. Guanlong has transformed from selling products to providing long-term services (operation and maintenance, data analysis, energy optimization), enhancing customer loyalty and stabilizing its revenue.
5. **A lesson for traditional manufacturers: The right approach to embracing AI**
This story offers a valuable advice for all traditional manufacturers: Don't blindly try to enter new fields like large platforms or AI models. Instead, focus on your familiar physical contexts and redefine your products with hardware, data, and services.
Many traditional companies, attracted by AI, try to develop their own chips or software but often fail due to a lack of technical knowledge and market understanding. Guanlong's approach is to:
- Build on expertise: Leverage their core skills in fluid control.
- Reapply knowledge: Transfer expertise from water management to AI liquid cooling and data analysis.
- Take small steps: Start by solving customers' most pressing problems and gradually expand.
In summary, AI is not meant to replace traditional manufacturing but to empower it. Companies that can effectively apply AI to real-world problems will build stronger barriers in the AI era. Guanlong Energy Saving's success shows that for traditional industries to thrive, it's about adding the right "fertilizer" (data and algorithms) and "pruning" (focusing on core areas) rather than completely changing their foundations.