Summary of Key Points
The competition in the AI industry has shifted from chips and model parameters to power supply capabilities: AI data centers (AIDCs) consume significantly more electricity due to their higher power output per cabinet, which is 5-10 times that of traditional data centers. This demand is comparable to that of small cities, leading to restrictions on new data center developments in states like Texas and New York in the United States due to power shortages. In China, a strategy called "computing-power synergy" has been adopted to coordinate the layout of computing power and electricity resources (e.g., direct connection to green energy sources and integration with solar and storage systems). Although tech giants are investing billions in AIDCs, they face challenges such as shortages of power equipment like transformers and gas turbines. Meanwhile, the explosive growth in computing power consumption has created significant opportunities for new energy industries like photovoltaics (PV) and energy storage, with "computing-power synergy" moving from a policy concept to practical implementation.
I. Why Has the AI Race Shifted to Power? Behind the Restrictions in the US Is an Imbalance Between Supply and Demand
The training and inference of large AI models require immense computing power, resulting in much higher electricity consumption compared to traditional data centers. While a typical data center cabinet has a power output of 2-4 kW, an AIDC requires 20-100 kW, meaning the electricity usage of a single AIDC is equivalent to that of a small city with tens of thousands of residents.
Texas in the US has suspended the approval of new data centers due to the total capacity of projects waiting to be connected exceeding the state's peak power demand by five times, with 90% of these being data centers. New York State has even issued a ban on all new data center construction, and at least 75 data center projects across the US were delayed or canceled in 2026 due to insufficient power supply.
II. China's Response: What Exactly Is "Computing-Power Synergy"?
Simply put, it involves planning and constructing computing power and electricity infrastructure together to ensure that data centers do not face power shortages and that electricity is not wasted.
For example, when building a data center, wind and PV power plants along with energy storage facilities are constructed simultaneously. Green energy is directly connected to the data center (known as "direct connection to green energy"), ensuring both power supply and compliance with policy requirements (e.g., requiring that more than 80% of the electricity used in new data centers at key nodes comes from green sources).
For instance, the Goldwind Technology project in Ulanqab includes 200,000 kW of wind power, 100,000 kW of PV power, and 45,000 kW of energy storage. After its operation began, the cost of computing power was reduced by 25%. The "Xinghe Base" project by Envision Technology uses electricity from its own wind farm, eliminating the need to rely on the grid and improving efficiency.
III. The Dilemmas Faced by Tech Giants
Despite investing billions, tech giants (such as Google, Microsoft, and Amazon) still face critical issues:
1. Shortages of Power Equipment: The delivery time for transformers can take 2.5-5 years, while gas turbine orders are backlogged until 2030, with prices tripling.
2. Inadequate Computing Power Supply: Even with investments of $2.4 trillion, these companies still struggle to meet their needs. Microsoft, for example, is short of computing power and has to purchase cloud capacity from competitors because the speed of data center construction cannot keep up with the demand for computing power.
As a result, tech giants are beginning to directly enter the energy industry: Google has acquired clean energy companies, and Meta has built its own natural gas power plants and energy storage facilities alongside its data centers to solve its power supply problems.
IV. The Impact of Computing Power Growth on New Energy Industries
The surge in computing power consumption has created substantial opportunities for new energy businesses:
- Surging Orders for Energy Storage: In the first half of 2026, the scale of energy storage procurement in China increased by 121.8% year-over-year, with some battery manufacturers receiving orders that were 30 times their usual volume, leading to production bottlenecks.
- New Markets for PV Companies: Companies like Jinko Energy see "computing-power synergy" as a breakthrough for the PV industry, predicting an additional 30%-50% demand in the next 3-5 years. Sungrow Power predicts that energy consumption for intelligent computing will be the fastest-growing segment of the PV market.
- Solving Energy Disposal Issues: New energy companies previously faced the problem of having excess electricity that they couldn't sell. Now, by building "computing power stations," they can use the generated electricity directly in data centers, both consuming it and generating additional revenue.
V. The Implementation of Computing-Power Synergy: From Policy to Practical Operations
In China, computing-power synergy has moved from a theoretical concept to practical action:
- Policy Support: The 14th Five-Year Plan includes computing-power synergy as part of the new power system construction, with targets for developing 160 million kW of new energy storage and transmitting 420 million kW of electricity from western regions to eastern areas.
- Practical Successes: Shanghai has successfully coordinated the centralized use of computing power, reducing the load on 16 data centers by 97,800 kW within two hours, saving nearly 100,000 households' electricity consumption.
- Large Market Potential: The combined investment in computing power networks and grids during the 14th Five-Year Plan period is expected to reach $4 trillion, making computing-power synergy a key driver of a trillion-dollar market.
Conclusion
The competition in the AI industry has changed. No longer does it depend on who has the largest model parameters or the strongest chips; instead, it focuses on who can provide stable and low-cost green energy supply. The US's restrictions represent a barrier, while China's approach to computing-power synergy represents a strategic advantage. Whoever succeeds in the energy sector will gain a competitive edge in the AI race. After all, even the most advanced models cannot function without sufficient power.