In a landmark development reported by China Electric Power News and later by Bloomberg, China’s large-scale data centers have for the first time entered the electricity spot trading market, operating as virtual power plants (VPPs). This milestone, achieved on May 14, 2026, signifies a strategic shift in how massive AI compute facilities manage their electricity consumption, enabling them to dynamically adjust workloads in response to real-time electricity prices. The innovative model, dubbed “computing follows electricity” (算力跟着电走), reflects a growing emphasis on energy efficiency and cost optimization amid China’s rapid AI expansion and its evolving power grid landscape.
The pioneering initiative took place in Guangdong province, where three major data center clusters participated: China Unicom Group’s Shaoguan Data Center alongside China Mobile Group’s Guangzhou and Zhanjiang Data Centers. These facilities connected to the “Yue Nengtou Virtual Power Plant Operation Platform,” operated by Guangdong Power Grid Energy Investment Co., Ltd. Through this platform, the data centers engaged in spot market electricity trading and settlement, a process known as “flexible spot purchases.” This approach allows the data centers to modulate their electricity demand and AI compute workloads based on volatile spot market prices, effectively acting as distributed energy assets within the grid.
The significance of this event cannot be overstated. It marks the first time in China that data centers have officially participated in electricity spot trading as virtual power plants, expanding the scope of VPPs beyond traditional energy producers and consumers. By integrating data centers into spot market mechanisms, China is pioneering an energy-compute symbiosis aimed at improving grid stability, reducing energy costs, and enhancing the sustainability of AI cloud infrastructure.
The mechanism underlying this model is straightforward yet transformative. As electricity spot market prices fluctuate throughout the day, the data centers adjust their AI workloads accordingly, scaling down compute-intensive tasks when prices peak and ramping up during periods of low demand and cheaper electricity. This dynamic scheduling embodies the “computing follows electricity” concept, which aligns with broader government and industry goals to integrate demand-side flexibility into China’s power grid. The data centers’ ability to act as virtual power plants means they can respond rapidly to grid signals, helping to balance supply and demand in real time.
This innovation also addresses a pressing challenge for China’s AI ecosystem: the high and fluctuating cost of electricity, which is a major factor in the operational expenses of AI compute clusters. With the explosive growth in AI model training and inference workloads, data centers consume enormous amounts of power, and optimizing electricity costs is crucial for maintaining competitive cloud pricing and profitability. The Guangdong pilot demonstrates how data centers can leverage market mechanisms to achieve cost savings while supporting renewable energy integration.
Virtual Power Plants and the Future of AI Compute Efficiency
The integration of data centers into virtual power plants is a natural evolution given the increasing scale and sophistication of China’s cloud and AI infrastructure. Traditionally, virtual power plants aggregate diverse distributed energy resources, such as solar panels, batteries, and demand response assets, to provide grid services and optimize energy use. By including data centers, which are both large consumers and flexible loads, the VPP ecosystem gains a powerful new component capable of modulating tens of megawatts of demand almost instantaneously.
This development is especially timely as China grapples with energy overcapacity and grid management challenges. The country’s power grid is increasingly incorporating intermittent renewable sources like solar and wind, creating volatility in supply. Flexible loads such as AI data centers can absorb excess power during peak renewable generation and ease demand during shortages, smoothing the grid’s operational profile. This not only enhances grid stability but also supports Beijing’s ambitious carbon neutrality goals.
From an AI industry perspective, the ability to “follow electricity” means cloud providers and AI service firms can optimize the cost-performance balance of their compute resources. This has implications for the ongoing compute price waves seen across China’s AI cloud market, where major players like Tencent, Alibaba, and Baidu have recently raised cloud rates amid chip supply constraints and rising energy costs. The new spot trading participation model could eventually help moderate compute price inflation by introducing more flexible and efficient electricity consumption patterns.
The Guangdong pilot’s success may encourage expansion across other provinces, particularly those with liberalized electricity markets. It also aligns with China’s broader push for data center decarbonization, including efforts around subsea data centers and offshore wind power integration, as discussed in our China Turns to Subsea Data Centers and Offshore Wind to Ease AI Computing Bottleneck analysis.
Broader Implications for China’s AI and Energy Strategy
China’s move to integrate data centers into spot electricity markets as virtual power plants has several strategic implications. First, it reflects the government’s increasing emphasis on optimizing energy consumption within the AI sector, which is now considered core national infrastructure. This complements recent regulatory measures targeting AI ethics and security, such as Beijing’s draft rules on AI chatbot interactive services and restrictions on AI virtual companions.
Second, the initiative highlights Guangdong’s role as a testing ground for energy and technology innovations. Guangdong is already a major data center hub with robust policy support and advanced grid infrastructure, making it an ideal province to pilot new electricity market mechanisms involving critical AI infrastructure.
Third, this development could reshape the competitive landscape for AI cloud service providers. By lowering electricity costs through spot market participation, companies like China Mobile and China Unicom can improve margins, offer more competitive pricing, or reinvest savings to expand AI compute capabilities. This could exert pressure on other cloud giants like Alibaba and Tencent, which are also navigating rising compute costs due to chip shortages and export restrictions, as detailed in our coverage of China’s chip self-sufficiency ambitions.
For investors, the integration of data centers into VPPs signals a maturing AI compute infrastructure market that is increasingly intertwined with China’s energy transition. Opportunities may arise in companies developing smart energy management systems, AI workload schedulers, and grid-interactive data center technologies.
(Related: China’s AI Compute Price Wave: Tencent, Alibaba, and Baidu All Raise Cloud Rates)
Outlook: Scaling “Computing Follows Electricity” Nationwide
Looking ahead, the success of Guangdong’s pilot program sets a precedent for expanding the “computing follows electricity” paradigm across China’s data center landscape. As more provinces liberalize electricity markets and enhance grid intelligence, data centers are well-positioned to become key players in demand-side management and renewable energy integration. This could accelerate China’s transition to a more sustainable and efficient AI compute ecosystem.
However, challenges remain. Scaling this model requires sophisticated real-time monitoring, AI-driven load scheduling algorithms, and regulatory frameworks that allow data centers to participate in spot markets without compromising service quality. Additionally, balancing the needs of AI workloads, often latency-sensitive and resource-intensive, with the volatility of spot electricity prices will require advanced orchestration technologies.
Further developments in China’s domestic AI chip supply chain, such as the milestone shipment of over one million Loongson 1M CPUs, will also influence the efficiency and flexibility of data center operations. Combined with ongoing regulatory trends and new AI compute pricing strategies, China’s data centers are entering a new era of energy-smart operation.
As China continues to integrate its AI ambitions with energy innovation, the data center spot trading pilot in Guangdong could become a blueprint for a new generation of AI infrastructure, one where compute power adjusts fluidly to the rhythms of the electricity grid, making Chinese AI more cost-effective, resilient, and sustainable.