On May 2, 2026, a 500-megawatt solar plant in the Tengger Desert began sending electricity directly to a data center cluster one kilometer away. No detour through the public grid. No waiting for transmission capacity. The power flowed through four 110-kilovolt lines built specifically for the project, straight into the server racks of the Zhongwei cloud base in Ningxia.
It was China's first large-scale "computing-power coordination" green direct-supply project. By June, the facility was operating as the country's first AI data center powered entirely by a direct green electricity connection. Its PUE — the ratio of total facility power to computing power — met the national requirement of below 1.2 for new hub-node data centers.
The project cost 8.7 billion yuan. It took months, not years, to connect. And it is one of at least half a dozen similar projects now operating or under construction across western China.
The United States, by contrast, has an interconnection queue that exceeds 2,600 gigawatts — nearly twice the capacity of the entire installed U.S. generating fleet. The median wait time is above four years, and in PJM, the largest grid operator, it is 40 months. Active projects in data center load-growth zones wait 36 to 48 months. Google's global head of sustainable development and climate policy has said that connecting to the U.S. transmission system is now "the biggest problem" for powering its data centers, with wait times exceeding a decade in some parts of the country.
China built the connection first. America is still in line.
What the Zhongwei Project Did
The Zhongwei project is not a pilot. It is a commercial facility with a commercial power contract.
Its structure is what makes it work. A 500-megawatt solar station was built adjacent to the data center cluster. A 1,500-megawatt wind farm is under construction 40 kilometers south, scheduled for full connection in September 2026. Together, they will generate 4.3 billion kilowatt-hours annually — sufficient to meet the data center cluster's demand.
The key is the "physical direct supply + bilateral trading" dual-track system. During solar generation hours, electricity flows directly from the PV plant to the data center via dedicated lines. When solar stops, wind power fills the gap. The transmission lines were built specifically for the project, reducing line losses from the typical 5 percent to under 3 percent.
The cost impact is measurable. The delivered electricity price for the Zhongwei data center cluster is approximately 45 percent of the industrial rate in eastern China. Electricity accounts for over 60 percent of a data center's operating cost. Cutting that cost by more than half changes the economics of AI training entirely.
The policy framework reinforces the model. China's "East Data, West Computing" initiative directs computing workloads to western provinces with abundant renewable energy. New data centers at the eight national hub nodes are required to meet a green power share of at least 80 percent and a PUE below 1.2. The "computing-power coordination" policy — written into the government work report for the first time in 2026 — makes the alignment of data center construction and renewable generation an explicit national priority.
The Scale Beyond Zhongwei
Zhongwei is the most visible example, but it is not the only one.
In Ulanqab, Inner Mongolia, a 300-megawatt renewable project (200 MW wind, 100 MW solar) with 45 megawatts of battery storage became China's first "source-grid-load-storage integration" project in July 2025. Source-grid-load-storage integration is a model that coordinates generation, transmission, load, and storage at the regional level. The project operates on a "load determines source, source follows load" principle. The city now hosts more than 100 signed data center projects with an operational computing scale of 172,000 PFlops.
In Helingeer, Inner Mongolia, a 360-megawatt wind-solar-storage project supplies green power directly to a data center cluster, with a green power share exceeding 80 percent and a computing scale of 125,000 PFlops.
In Chifeng, Inner Mongolia, Envision Group and Tencent built what they call the world's first 100 percent green power direct-supply data center. The project uses a self-built wind farm, self-laid dedicated lines, and physical direct supply into the equipment room. Its comprehensive energy cost dropped by more than 40 percent, with annual carbon reduction of 180,000 tons. Envision has since scaled the model to a 2-gigawatt "super node" in Ulanqab, designed for million-card parallel computing.
In Zhangjiakou, Hebei, four computing center projects have been approved for green power direct connection.
The pattern is consistent. China is not building data centers and then hoping the grid can supply them. It is building the generation and the connection at the same time, in the same place, under a single integrated plan.
The American Interconnection Trap
The U.S. has the opposite problem. It has abundant renewable energy resources, abundant capital, and abundant demand for AI compute. What it does not have is a mechanism for connecting them on a relevant timescale.
The national interconnection queue held more than 2,200 gigawatts of generation and storage capacity at the end of 2025 — well over the country's entire installed generating fleet. The Federal Energy Regulatory Commission targets an 8-to-11-month timeline for interconnection agreements. The actual average is 25 months nationally, 40 months in PJM, and 36 to 48 months in the data center load-growth zones of both PJM and ERCOT.
Most of those projects will never be built. Of ERCOT projects that began screening by 2020, only 40 percent have reached interconnection agreement status or become operational. In PJM, the conversion rate is even lower.
The result is a structural mismatch. Data center demand is arriving on a timeline the grid cannot match. Since October 2025, more than 10 gigawatts of new data center projects have been announced in PJM and ERCOT alone, with approximately 30 percent planned online by 2027. The generation to serve them is competing for space in a queue that moves at a fraction of the required speed.
The U.S. does not lack renewable energy. Texas, the Great Plains, and the Mountain West have some of the best wind and solar resources in the world. What they lack is a transmission system that can move that power to where the data centers are — and a regulatory process that can approve new transmission lines in less than a decade.
The Structural Difference
The contrast between the Zhongwei project and the American interconnection queue is not about technology. It is about institutional design.
China's "computing-power coordination" model treats data centers and power generation as a single planning problem. The grid operator builds the transmission line. The renewable generator signs a direct-supply contract. The data center connects. The price is set to make the project work. The entire process is coordinated at the provincial and national level.
The U.S. model treats them as separate problems. A developer builds a data center. A utility builds the generation. A regional grid operator manages the interconnection queue. The Federal Energy Regulatory Commission sets the rules. State public utility commissions approve the rates. The process is fragmented across jurisdictions, each with its own timeline, its own priorities, and its own incentives.
The result is that China can build a 500-megawatt solar plant and connect it to a data center in months. The U.S. can take years just to reach the front of the queue — and then years more to build the transmission line that the queue was waiting for.
What the First 100% Project Proves
The Zhongwei project does not prove that China's model is universally superior. It operates in a political system with a level of centralized coordination that would be difficult to replicate in the United States. It benefits from geography — western China has abundant land, cheap labor, and some of the best renewable resources in the world — that not every region can match.
What it does prove is that the bottleneck the U.S. is struggling with is not inevitable. The power was waiting for the compute in China. In the U.S., the compute is waiting for the power.
The first 100 percent green power data center is a planning breakthrough rather than a technological one. And it is the kind of breakthrough that the U.S. interconnection queue — with its 2,600 gigawatts of waiting projects and its four-year median wait — is not designed to produce.
Sources: CGTN (June 26, 2026); China Energy News (May 3, 2026); Economic Information Daily (July 29, 2026); Xinhua Finance (April 1, 2026); China News Inner Mongolia (August 6, 2026); Ulanqab Municipal Government (September 11, 2026); China Energy News (June 30, 2026); Carbon Direct (May 14, 2026); Lawrence Berkeley National Laboratory via Taskade (August 4, 2026); JLL Global Data Center Outlook 2026 via ECMag (August 17, 2026); Reuters (January 15, 2026); People's Daily (May 11, 2026); National Development and Reform Commission (August 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.
Project capacity and cost figures come from public reporting; independent verification is limited.
Queue and wait-time statistics are point-in-time estimates; rules and backlogs change.
Comparative claims rely on public reporting, not operational data.
Forward project timelines are plans, not completed construction.
Reporting standards differ across jurisdictions; figures may not be directly comparable.
AI infrastructure is changing quickly; claims can become outdated soon.
Sources
- CGTN (June 26, 2026)
- China Energy News (May 3, 2026)
- Economic Information Daily (July 29, 2026)
- Xinhua Finance (April 1, 2026)
- China News Inner Mongolia (August 6, 2026)
- Ulanqab Municipal Government (September 11, 2026)
- China Energy News (June 30, 2026)
- Carbon Direct (May 14, 2026)
- Lawrence Berkeley National Laboratory via Taskade (August 4, 2026)
- JLL Global Data Center Outlook 2026 via ECMag (August 17, 2026)
- Reuters (January 15, 2026)
- People's Daily (May 11, 2026)
- National Development and Reform Commission (August 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.; Project capacity and cost figures come from public reporting; independent verification is limited.; Queue and wait-time statistics are point-in-time estimates; rules and backlogs change.; Comparative claims rely on public reporting, not operational data.; Forward project timelines are plans, not completed construction.; Reporting standards differ across jurisdictions; figures may not be directly comparable.; AI infrastructure is changing quickly; claims can become outdated soon.