A Critical Analysis Behind the Surge in Computing Power: The Shift from Quantity to Quality in the Competitive Landscape
Hello everyone, I’m your financial journalist. Recently, the news about the 2026 China Computing Power Conference has gone viral, with impressive figures: as of June this year, the scale of intelligent computing in our country has reached 2185 EFLOPS, a year-on-year increase of 177%.
At first glance, this might seem like a story of “excess computing power” or “crazy competition.” But if you dig deeper into the news, you’ll realize that the focus has shifted. In the past, the competition was about who could build the most; now, it’s about who uses the power efficiently, who saves the most money, and who operates most stably.
It’s similar to how people used to compare the size of houses, but now they’re more interested in how well they insulate, how low their utility costs are, and how fast their internet connections are.
Below, I’ll break down this news into five key points to explain the underlying trends in plain language.
---
1. The Truth About Supply and Demand: Not “Overcapacity,” but “Structural Shortage”
Many people, seeing the 177% increase in computing power, might think, “Wow, there’s so much power that we won’t need to worry in the future!”
That’s completely wrong.
The news mentions a crucial concept: structural shortage.
- What is a structural shortage?
It’s like a restaurant where fast-food (general-purpose computing power) is easily available and even in surplus, but the “customized dishes” (high-quality computing power for intelligent agents) are in high demand, with long waiting lists.
- Who is competing for computing power?
Big tech companies like ByteDance, Alibaba, and Tencent, as well as telecom operators and AI model developers, are all investing heavily in computing power.
- Big tech’s logic: We have models and applications; we need to meet our own needs first, and only then sell the excess to others.
- Operators’ logic: We have existing data centers and networks, but our software ecosystems are weak, so we mainly take on big tech’s outsourcing projects.
- Small and medium-sized companies’ logic: They face the greatest pressure, lacking capital, power, and customers, and may end up working as subcontractors or targeting niche markets.
Conclusion: Computing power isn’t overabundant; high-quality power is still in short supply. If you only have general-purpose power, it might be hard to sell at a high price. But if you can provide stable, efficient power for complex AI tasks, you’ll be in high demand.
---
2. The New Metrics: From “Number of Servers” to “Effective Computing Power”
In the past, the quality of a data center was measured by the number of servers and its size. Now, these metrics are outdated. The news emphasizes a new term: effective computing power.
- What is effective computing power?
Imagine a supercar with a great driver: the car’s actual value is much higher if the driver is skilled, can handle tasks efficiently, and avoids problems. Similarly, a server’s actual computing power depends on the quality of its maintenance and management.
- The current situation: Only 7 provinces have data centers with an utilization rate of over 75%. This means much of the built-up computing power is underutilized.
- Why the emphasis on effectiveness?
Building a large-scale cluster (e.g., with tens of thousands of GPUs) is like managing a factory with millions of workers. Poor management can lead to frequent failures and inefficiencies, wasting huge investments.
The future competition will be about getting the most out of the hardware.
---
3. Hidden Champions: Power and Cooling Become Critical Barriers
In the past, building a data center meant finding a location and connecting it to the internet. Now, that’s not enough.
The news highlights key factors such as power supply, the use of renewable energy, power density per server, and liquid cooling capabilities.
- Power: The Lifeline of Computing Power
Electricity costs are a major expense: AI training is energy-intensive. Hebei Unicom charges 0.62 yuan per kilowatt-hour, while in Inner Mongolia, Shanxi, and some western regions, it can be as low as 0.3–0.4 yuan per kilowatt-hour. This difference directly affects profitability.
The trend towards green energy: The government is promoting the use of renewable energy, which not only benefits the environment but also offers policy incentives and potential market advantages for exporters or high-end customers.
- Cooling: Liquid cooling is essential
Air cooling is no longer enough: Traditional servers use fans, with a PUE (Power Usage Efficiency) of around 1.3, meaning 30% of the energy is used on cooling. Liquid cooling reduces this to around 1.15, saving significant costs.
Leading companies like Ruize Technology and China Mobile are upgrading to liquid cooling to maximize efficiency.
---
4. The Reshaping of the Industry Chain: Who Will Win, Who Will Lose?
The upgrade in the computing power industry is reshaping the entire supply chain:
- Winners: Companies with both resources and technology
- Top cloud providers: They have models, computing power, and the ability to manage it efficiently, making them the biggest beneficiaries.
- Operators with power advantages: Those with existing data centers and networks can take on more outsourcing projects.
- Companies specializing in cooling and power solutions: As liquid cooling becomes standard, their demand will increase.
- Losers/Risk Players: Pure “middlemen”
- Small and medium-sized companies: Without power advantages or strong software, they’ll struggle in the market as big tech and operators build their own infrastructure.
- Data centers in high-cost areas: High electricity costs will put them at a disadvantage in price competition.
- New Opportunities: Maintenance and management software
The news highlights the importance of software and services that manage computing power efficiently. Companies that understand AI, scheduling, and maintenance will be highly valued.
---
5. Policy Directions: The Country’s Priorities: Quality, Green Energy, and Independence
The Ministry of Industry and Information Technology’s “15th Five-Year Plan” for the information and communications industry sends three clear signals:
1. Orderly deployment, no blind expansion: The country aims to build large-scale clusters but only where there’s real demand.
2. Adaptation to domestic chips: There’s a push to use domestic chips, which will provide more support and market opportunities.
3. Green computing power certification: The government will issue certifications for green computing power, which will influence business decisions based on ESG (environmental, social, and governance) criteria.
Summary for You:
For individuals and investors, the key is to understand that the growth in computing power is more about efficiency and cost-effectiveness. Focus on utilization rates and power efficiency. Companies that can provide stable, efficient power for AI tasks will be the leaders. Pay attention to liquid cooling and renewable energy technologies, as these are the fastest-growing areas. Be cautious of companies that rely solely on hardware; without strong software and maintenance capabilities, their infrastructure may become useless.
The computing power industry is moving from a phase of rapid expansion to one of focused, high-quality development. The future winners will be those who understand hardware, software, and energy management.