In the first half of 2026, China threw away approximately 360 billion kilowatt-hours of wind and solar power. The electricity was generated, delivered to the grid, and then discarded because there was nowhere to put it. Meanwhile, the country‘s total new electricity demand during the same period was only 258 billion kilowatt-hours. The wasted clean energy exceeded the net new demand by more than 100 billion kilowatt-hours.
At the same time, in the United States, data centers are pushing residential electricity bills to levels that have become a midterm election issue. PJM’s independent market monitor estimates that data centers have cost the grid‘s 67 million ratepayers approximately $29 billion over roughly two years. Residential electricity prices rose from 13.15 cents per kilowatt-hour in 2020 to 17.30 cents by 2025. A federal analysis found rates jumped 7.3 percent between April 2025 and April 2026 alone.
China has a curtailment problem. America has a cost-shifting problem. The two countries are solving them — or failing to solve them — in opposite ways.
The Scale of the Waste
China’s curtailment figures are enormous. In the first two months of 2026, national wind utilization fell to 91.5 percent, down 2.3 percentage points year over year. Solar utilization dropped to 90.8 percent, down 3.1 points. In the northwest, the numbers are worse. Gansu‘s solar utilization rate fell to 82.5 percent. Qinghai’s dropped to 78.7 percent. Xinjiang‘s wind curtailment rate reached 11.3 percent, with solar curtailment at 14.4 percent. Inner Mongolia’s western region curtailed 15.6 percent of its wind generation.
The root cause is a mismatch between where the power is generated and where it is consumed. Western China holds more than 78 percent of the country‘s wind and solar resources but carries only about 30 percent of its electricity load. New energy installations grew 22 percent in 2025 to 430 gigawatts, while grid investment grew only about 5 percent.
The power plants are in one place. The factories are in another.
The Data Center Solution
China’s answer to this mismatch is called “computing-power coordination” — a policy framework that aligns data center construction with renewable energy generation. The goal is to place data centers in the regions where renewable energy is abundant and curtailment is highest.
The first large-scale green power direct-supply project began operation in May 2026 at China Datang‘s cloud data center in Zhongwei, Ningxia. A 500-megawatt photovoltaic station was connected directly to the data center via a dedicated transmission line, bypassing the main grid entirely. During periods of solar generation, electricity flows directly from the PV plant. When solar is unavailable, wind power compensates. The first phase is designed for 2,000 MW of total capacity — 500 MW solar and 1,500 MW wind — with an investment of 8.7 billion yuan ($1.28 billion). A second phase will add 4,600 MW, bringing total investment to nearly 20 billion yuan. Upon full operation, the first phase will meet the data center’s annual power demand of 2.29 billion kilowatt-hours.
In Gansu‘s Qingyang “East Data, West Computing” industrial park, a different model is being tested. Instead of building dedicated generation, the project aggregates power from multiple distributed wind and solar stations, packages it into a stable green electricity stream, and sells it directly to the data center cluster. The park’s delivered electricity price is capped at 0.398 yuan per kilowatt-hour. The green power supply rate exceeds 80 percent. The project is expected to reduce electricity costs for park enterprises by more than 30 million yuan annually and to supply approximately 4 billion kilowatt-hours of green power to the park each year.
The policy framework reinforces the pattern. New data centers at the eight national hub nodes are required to meet a green power share of over 80 percent and a PUE of ≤1.2. In Inner Mongolia, wind power direct supply is locked at 0.26 yuan per kilowatt-hour for 10 years. In Ningxia, the base electricity price is 0.2734 yuan per kilowatt-hour. These are not market-clearing prices. They are policy outcomes designed to make curtailment elimination profitable.
The mechanism is not complicated. AI training workloads — model training, data backup, offline inference — are time-tolerant. They can run when the wind blows and the sun shines, and scale down when the grid is stressed. China is building data centers that treat renewable energy as a scheduling input, not an afterthought.
The American Cost-Shifting Problem
The United States has the opposite problem. Its data centers are not solving curtailment. They are creating a new cost burden that is being distributed across the entire ratepayer base.
The mechanism is structural. When a data center connects to the grid, the utility builds transmission lines, substations, and generation to serve it. Those costs are recovered across all customers. Ari Peskoe, director of the Harvard Electricity Law Initiative, summarized it: “Everyone pays those costs. They‘re spread across the region to every business and resident that has an electricity meter.”
The numbers are visible in the capacity market, too. PJM’s capacity auction cleared at $28.92 per MW-day for the 2024/2025 delivery year, $269.92 for 2025/2026, and $329.17 — the price cap — for 2026/2027. The grid’s independent market monitor attributes about 63 percent of the 2025/2026 increase to data centers.
The tax credit cliff has made it worse. With the Inflation Reduction Act‘s renewable tax credits expiring, clean energy power purchase agreement prices are projected to rise 40 to 120 percent. In Texas, PPA prices could jump from $55 per megawatt-hour to $121. Data centers, desperate for power and willing to pay premium rates, are squeezing manufacturers out of the clean energy contract market. As one industrial buyer put it: “Hyperscalers are willing to pay a premium, which makes it difficult for manufacturers.”
The problem is not that data centers use electricity. It is that the American regulatory framework has no mechanism for ensuring they pay the full cost of the infrastructure built to serve them. The Ratepayer Protection Pledge — signed by seven major AI companies in March 2026 — is voluntary and nonbinding. The House passed a Ratepayer Protection Act in September 2026, but the bill only requires states to consider cost-allocation standards. It does not mandate them.
The Structural Difference
The contrast between the two systems is not about technology. It is about institutional design.
In China, the grid is state-owned and policy-directed. When the government decides that data centers should be built in western provinces to absorb curtailment, the grid operator builds the transmission line, the renewable generator signs a direct-supply contract, and the data center connects. The price is set to make the project work. The green power share is mandated. The PUE threshold is enforced through the ISP licensing process.
In the United States, the grid is regulated but privately owned. When a data center connects, the utility builds the infrastructure and recovers the cost from all ratepayers. There is no national policy directing data centers to locate near curtailed renewables. There is no mechanism for ensuring that the entity creating the demand pays for the infrastructure that serves it. The market allocates costs, and the market has no incentive to protect households from subsidizing hyperscalers.
The results speak for themselves. China’s curtailment rate is high, but the country has a mechanism for reducing it — and that mechanism is producing results. Gansu‘s Qingyang park, which was built specifically to absorb local curtailment, now supplies 80 percent green power at 0.398 yuan per kilowatt-hour. The United States has no comparable mechanism. Its curtailment persists, its data centers cluster in population centers, and its ratepayers absorb the cost.
What the Comparison Reveals
China’s curtailment problem is real. In the first half of 2026, the country wasted 360 billion kilowatt-hours of clean electricity — more than its net new demand. That is a failure of planning and grid capacity.
But the failure is being addressed. The “East Data, West Computing” initiative is not a slogan. It is a construction program. The 500-megawatt solar station at Zhongwei is not a pilot. It is a commercial facility supplying a commercial data center. The green power direct-supply model is not theoretical. It is operational, priced below grid rates, and expanding.
The American problem is different in kind. The U.S. does not have a curtailment crisis of the same scale. It has a cost allocation crisis. The data center boom is producing enormous economic value — and enormous costs that are being distributed to people who did not ask for the infrastructure. The political system has recognized the problem. It has not solved it.
China solves curtailment by moving the compute to the power. The United States moves the power to the compute, and sends the bill to the ratepayer. One approach wastes electricity. The other wastes money. Both are choices — and both are being made right now.
Sources: Centre for Research on Energy and Clean Air (CREA) and Global Energy Monitor (August 2026); National Energy Administration via 163.com (March 2026); Monitoring Analytics via Maryland Matters (2026); Consumer Reports (May 2026); EIA (2026); SASAC via China News Service (May 2026); Xinhua (January 2026); People’s Daily (2026); National Development and Reform Commission (May 2026); Cailianshe (July 2026); Harvard Electricity Law Initiative via Canary Media (2026).
Disclaimer
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations
This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.
Curtailment and utilization rates come from published Chinese energy statistics; provincial reporting methods vary.
Capacity market prices reflect PJM auction results and are subject to change in future auctions.
Project costs and timelines are point-in-time announcements from developers and state media.
Comparative claims rely on public reporting, not audited operational data.
Sources
- Centre for Research on Energy and Clean Air (CREA) and Global Energy Monitor (August 2026)
- National Energy Administration via 163.com (March 2026)
- Monitoring Analytics via Maryland Matters (2026)
- Consumer Reports (May 2026)
- EIA (2026)
- SASAC via China News Service (May 2026)
- Xinhua (January 2026)
- People’s Daily (2026)
- National Development and Reform Commission (May 2026)
- Cailianshe (July 2026)
- Harvard Electricity Law Initiative via Canary Media (2026).
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations: This analysis is based on reporting and public data available as of the article date; figures may be revised as sources update.; Curtailment and utilization rates come from published Chinese energy statistics; provincial reporting methods vary.; Capacity market prices reflect PJM auction results and are subject to change in future auctions.; Project costs and timelines are point-in-time announcements from developers and state media.; Comparative claims rely on public reporting, not audited operational data.