Economic Observer reporter Wang Yajie

Ma Hongyuan, a small business owner in Zhongwei, Ningxia, has recently made a decision to give up applying for a national industry list - the recommended list of "National Green Computing Facilities for 2026".

On September 21st, the Ministry of Industry and Information Technology, the National Development and Reform Commission, the State Administration of Financial Supervision and Administration, the State Administration of Civil Affairs, and the National Energy Administration jointly issued the "Notice on Organizing the Recommendation of National Green Computing Facilities for 2026" (hereinafter referred to as the "Notice"), which clearly requires all relevant units to submit relevant recommendation materials before November 16th.

Ma Hongyuan said that in layman's terms, this list assesses the green operation level of computing power data centers, and data centers that meet the standards will be awarded national certification qualifications. Although the list is not mandatory to declare, giving up the declaration means that the enterprise will bear a real industry "disadvantage". Government bidding projects, purchase orders of Internet giants, and special low interest loans of CALI will give priority to licensed CALI computer rooms.

What is the green operation level of the computing power room? The core depends on the electricity consumption structure, distinguishing between thermal power and clean energy such as wind power and photovoltaics. Taking the computer room of the computing power enterprise managed by Ma Hongyuan as an example, 80% of its electricity comes from local wind and solar resources in Zhongwei, and the door-to-door electricity price is only 0.36 yuan/kWh. From the perspective of green power endowment alone, it has a competitive advantage among computing power rooms in the country.

Even though he holds high-quality green power resources, Ma Hongyuan still chooses to give up the application. In addition to assessing the proportion of green electricity usage, this list application also sets relevant mandatory entry thresholds: for example, the standard rack size for general computing facilities is not less than 3000, and the rack shelving rate is not less than 60% (excluding public institutions, financial sector computing facilities, and smart computing facilities). Although the policy exempts the scale requirement of 3000 racks for intelligent computing facilities, meeting the shelving rate is still a rigid assessment indicator. Ma Hongyuan's main business is domestic AI (artificial intelligence) computing power services, with scattered AI entrepreneurial teams as the main customers. The business is highly volatile, and the cabinet load is fluctuating, making it difficult to maintain a stable shelving rate of over 60% in the long run, which has become the core obstacle to its application.

The policy level includes "carbon efficiency" in the scoring table of computing power assessment to measure the "green content" of the data center. Related companies are starting to 'move'. The Economic Observer interviewed companies such as computing power leasing companies, AI chip manufacturers, and computing power park operators, and learned that some companies are urgently supplementing their green power indicators and improving their cabinet shelving rates to apply, while some small and medium-sized enterprises have started to dispose of computing power equipment and plan to gradually exit the market.

Spelling 'green'

After giving up the application, Ma Hongyuan finalized the company's follow-up business strategy, connecting some computing power resources with third-party institutions for entrusted operation, revitalizing idle computing power for external output, and focusing the remaining computing power on connecting with large model enterprises and research institutes in the east, striving to win medium - and long-term framework cooperation orders.

Ma Hongyuan said, "Instead of struggling, it's better to first stabilize the utilization rate of your computing power

The Economic Observer contacted small and medium-sized computing power enterprises in Jiangsu, Gansu and other places and learned that some enterprises have attempted to gather the required operating volume through asset consolidation, but core interest issues such as equity division and income distribution are difficult to unify, leading to the suspension or miscarriage of cooperation intentions. Some companies also plan to host their own servers in the data centers of top tier operators, relying on their qualifications to participate in the application process. However, top tier operators have strict requirements for equipment quality, operation and maintenance standards, and business qualification audits for accessing assets.

Compared to ordinary small and medium-sized players, Zhang Shouye, the operator of a smart computing park in Inner Mongolia, has more advantageous application conditions. The proportion of green electricity in the park has remained stable at over 84% throughout the year, with a household electricity price of 0.35 yuan/kWh. The green electricity qualification is at an advantageous level among the computing power hubs in the western region.

In Zhang Shouye's view, the official declaration and assessment are based on normalized operational data, and the shelving rate built up by short-term orders is easily invalidated in the annual review. For this reason, he chose to only push the long-term stable operation of the data center section to participate in the application, and reserved unsaturated data centers to temporarily not participate in the evaluation, in order to avoid the risk of subsequent review.

The layout rhythm of leading enterprises in the industry is more advanced and efficient. A leading computing rental enterprise in Wuxi has currently landed 5 billion yuan of three-year long-term computing contract orders, achieving full load operation of high-end computing rental rate, and the order reserve has been scheduled until 2028. The enterprise plans to apply for the list of national level green computing power facilities this time, aiming to apply for zero carbon factory qualifications corresponding to core targeted qualifications, as well as special credit dividends from financial institutions for green computing power projects.

The person in charge of the enterprise, Zhang Lei, predicted that "for projects with green qualifications and equipment of the same scale, the comprehensive financing cost can be reduced by several tens of basis points

During the current application window, the competent departments of computing power hubs in various western regions are actively promoting preparatory work, and the logic of regional investment attraction has undergone a significant change. Green power endowment and green operation and maintenance capabilities have become the core bonus points for computing power investment attraction.

A person from the Western Investment Promotion System revealed to the Economic Observer that in the current negotiations for the landing of computing power projects, investors are primarily concerned with green indicators such as regional green power supply capacity, green power traceability compliance, and computing power collaborative supporting system.

From the perspective of the application situation of the Northwest Computing Hub, the speed of promotion of computing power entities above a certain scale is leading, and the three major operators and top intelligent computing parks have all accelerated the preparation of application materials. Some small and medium-sized computing power enterprises are adopting a wait-and-see approach due to their failure to meet targets and unclear prospects.

Green electricity on the scale

The core consideration for the differentiated choices of Ma Hongyuan, Zhang Shouye, and Zhang Lei lies in the updated evaluation logic of green computing power in this round of application.

A person who participated in the development of the "National Green Computing Facility Evaluation Index System" (which was released simultaneously with the "Notice" and serves as the evaluation criteria for this round of application, hereinafter referred to as the "Index System") explained to the Economic Observer reporter that the new version of the "Index System" has been optimized and upgraded based on the existing evaluation framework, covering multiple types of facilities such as intelligent computing, general computing, and supercomputing. It has newly included indicators such as unit computing power efficiency, carbon utilization efficiency, digital carbon management, product carbon footprint, etc., and has increased the score weight of carbon efficiency, water efficiency, green electricity proportion, and shelving rate. The "predecessor" of this indicator system was the "National Green Data Center" recommendation and evaluation system, as well as the supporting "Green Data Center Evaluation Indicator System," organized by the Ministry of Industry and Information Technology, the State Administration of Civil Affairs, the National Energy Administration, and other departments since 2020. At that time, the evaluation focused on dimensions such as project construction plans and design parameters, and was benchmarked by traditional computing data centers, covering 5 aspects and 17 indicators including efficient utilization of energy resources, green design and procurement. In the past two years, AI intelligent computing clusters have been put into large-scale production, and their business models, power consumption characteristics, operation and maintenance logic differ greatly from those of general computing data centers. The original evaluation system can no longer adapt to the current industry situation, which has given rise to the upgraded version of the "Indicator System" for computing power facilities and covering intelligent computing/general computing/supercomputing. Standard computing power facilities will be given priority access to government bidding, large factory procurement, and low interest loans.

This also means that green power quality has officially become an independent assessment indicator for computing power facilities.

The 2026 version of the indicators will add CUE (carbon utilization efficiency) and include the carbon emission factor of electricity in the accounting system. The higher the proportion of green electricity in the computer room, the lower the carbon emission intensity and the higher the carbon efficiency assessment score. There is a consensus in the industry that PUE (Electricity Utilization Efficiency, which refers to the ratio of total power consumption in data centers to power consumption of IT equipment) is responsible for energy conservation, while CUE is responsible for carbon reduction. These two indicators complement each other and together form the core system for green evaluation of computing power.

At the same time, this application clarifies that the list review will be conducted once a year, and if it does not meet the standards, it will be dynamically recalled. Specifically, five core operational data points, including electricity utilization efficiency, carbon utilization efficiency, water resource utilization efficiency, proportion of renewable energy electricity, and cabinet shelving rate, need to be reviewed annually. Failure to meet annual targets will result in dynamic removal from the list.

The above-mentioned individuals added that the core purpose of the dynamic entry and exit mechanism is to establish a normalized survival of the fittest system, forcing enterprises to shift their green operation and maintenance from "reporting for sudden rectification" to normalized operation standards.

At the practical implementation level, green electricity accounting has strict and detailed execution rules.

A representative from a green certificate accounting agency told Economic Observer that the green electricity and green certificates purchased by enterprises must complete the official verification process before they can be included in the proportion of renewable energy consumption. The three major links of transaction performance, platform verification, and annual data matching are indispensable. Some enterprises have only completed green certificate procurement and have not completed verification and filing, and their application materials have been directly returned for correction.

The person believes that the carbon footprint of future products may extend to the entire lifecycle of the supply chain. At the same time, the carbon footprint accounting chain of computing power equipment is complex, covering the entire life cycle of hardware raw material manufacturing, equipment transportation, data center operation, and scrap recycling, which puts extremely high demands on the data collection ability of enterprise supply chain.

Shuffle

After the official implementation and assessment of green electricity and carbon efficiency indicators, in Zhang Shouye's view, the core value of green computing qualification lies in breaking through the entry threshold of the high-end market, rather than directly raising the rent of cabinets.

According to the calculations of the green certificate accounting agency mentioned above, the compliance cost of green computing power has clear quantitative differences. The annual cost of purchasing only green certificates for a single cabinet would increase by thousands of yuan. Combined with soft compliance expenditures such as carbon inventory checks, carbon footprint accounting, and special inspections, small and medium-sized computing power enterprises would incur fixed compliance costs of hundreds of thousands of yuan annually.

The new green assessment threshold is reshaping the industrial pattern of computing power in the eastern and western regions of China. The western region, relying on abundant and low-priced scenic resources, has formed a natural advantage in green computing power.

A local investment director in Inner Mongolia told reporters that green power is the core chip for attracting investment in western computing power, but it is not the only chip. Investors will consider multiple factors such as network latency, logistics support, and talent reserves when selecting a location.

The eastern region, relying on technological and scheduling advantages, has embarked on a differentiated path of green computing power development.

Hu Rong, Chairman of Wuhu Big Data Construction Investment Operation Co., Ltd., introduced that the Wuhu computing power cluster mainly adopts market-oriented procurement of green electricity, relies on the provincial computing power coordination and scheduling platform, and builds a computing power collaborative service system. By predicting electricity prices and fluctuations in green electricity output, non urgent computing power tasks are staggered and scheduled to operate during periods of sufficient green electricity, maximizing the consumption of local wind and solar resources; When there is a shortage of local green power supply, cross regional scheduling of computing power tasks to green power rich areas in western China such as Xinjiang can achieve optimal allocation of computing power resources across the entire region.

Ma Hongyuan stated that the reshuffle of small and medium-sized computing power enterprises has quietly begun. Shrinking business scale, managing sublet assets, and completely exiting the track have become the three mainstream choices for small and medium-sized enterprises in the industry.

In Zhang Lei's view, the upgrading of green qualifications and compliance thresholds will accelerate the aggregation of orders and capital to top enterprises. Small and medium-sized intelligent computing operators that rely solely on scattered orders and lack scale qualifications will continue to compress their survival space.

A top downstream computing power purchaser told reporters that the green computing power list has become an important basis for customer segmentation. Unqualified suppliers can continue to undertake market-oriented commercial orders, but high value-added and high stability project resources such as government and state-owned enterprises will be fully tilted towards certified top enterprises.

Along with the upgrading of industry barriers, new compliance arbitrage risks have also emerged.

Hu Rong said, "In the past, the chaos in the industry was mainly focused on illegally obtaining policy subsidies. In the future, the risk of companies engaging in illegal operations will shift to falsifying green qualifications." Some companies, in order to meet standards, have hoarded green certificates and inflated the rate of cabinet shelving. In this regard, she suggested that the listing rate should be cross checked with electricity meters and network traffic data, and green certificates must be bound to real computing power business orders to prevent violations of idle flashing indicators.

The new evaluation system also highlights to some extent the shortcomings of domestic AI computing power. Chen Xinxin, the business manager of a domestic AI chip manufacturer, told reporters that under the same scale computing power cluster, there is still a certain gap between the actual effective computing power of domestic chips and overseas chips. At present, state-owned projects can moderately accommodate energy efficiency gaps due to their independent and controllable needs. But market-oriented commercial customers have higher requirements for computing power efficiency and carbon emissions indicators, and will be more cautious and conservative in equipment selection.

With the comprehensive implementation of this new green assessment system, the resource allocation, competitive landscape, and development logic of China's computing industry are undergoing a round of reshuffle.

(At the request of the interviewee, Zhang Lei uses a pseudonym)