On October 21, the Ministry of Industry and Information Technology (MIIT) issued the "Implementation Plan for Accelerating the Construction of a Dual Carbon Control Work System in the Industrial and Information Technology Sector (2026–2030)". The plan explicitly proposes that by 2030, approximately 500 national-level zero-carbon factories and zero-carbon computing facilities will be built, around 400 carbon footprint accounting rules and standards for key industrial products will be formulated, and CO2 emissions per unit of added value for industries above designated size will decrease by more than 17% compared to 2025.
For the first time, computing power facilities are listed as an independent category alongside zero-carbon factories, zero-carbon parks, and low-carbon/zero-carbon supply chains, entering the sequence of national-level zero-carbon benchmarks.
01、Three "Hard Constraints" Reshape Assessment Rules
The substantive change in this policy lies not in "another green certification," but in the comprehensive tightening of assessment indicators, entry mechanisms, and regulatory methods.
1) Change in Indicators: PUE is No Longer Universal; Green Electricity Must Be "Physically Directly Supplied"
In the past, data centers competed primarily on PUE. However, a data center with extremely low PUE still generates high carbon emissions if it relies entirely on coal-fired power. Zero-carbon computing facilities set three core indicators: carbon emissions per unit of energy consumption, proportion of non-fossil energy consumption, and proportion of physically certified non-fossil electricity consumption.
The specific values are divided into two tiers: Carbon emissions per unit of energy consumption: basic requirement ≤1.8 tons CO2/ton standard coal, target requirement ≤0.2 tons CO2/ton standard coal; Proportion of non-fossil energy consumption: basic requirement ≥30%, target requirement ≥95%; Proportion of physically certified non-fossil electricity consumption: no basic requirement for computing facilities, target requirement ≥35%, applicable to applicant units with annual electricity consumption exceeding 5 million kWh.
The distinction between "basic requirements" and "target requirements" is critical: the former is the application threshold, while the latter is the completion line that must be reached before 2030. The notice clearly states that applicant units should meet the basic requirements of the core indicators and commit to reaching the target requirements during the construction period.
The third indicator carries the most weight. According to the accompanying indicator explanations, only non-fossil electricity that can be physically traced—such as direct green power connection, self-generation and self-use, and microgrids—is counted in this item. Simply purchasing Green Energy Certificates (GECs) can count towards the proportion of non-fossil energy consumption but does not count towards the proportion of physically certified non-fossil electricity consumption. Therefore, companies relying solely on GECs may pass the application threshold but cannot meet the completion line before 2030.
2) Change in Thresholds: Multi-Dimensional Systematic Entry Screening
Zero-carbon certification is not open to all enterprises. In addition to requiring selection into the National Green Computing Facility List and a scale of no less than 3,000 standard racks (this requirement does not apply to AI computing centers), applicant entities must also possess independent legal person status, clear property rights, defined and complete physical boundaries, independent power distribution and cooling systems, and achieve an energy efficiency level of Grade 2 or higher according to the "Energy Efficiency Limits and Grades for Data Centers" (GB 40879—2021).
This means that zero-carbon computing facilities are not assessed on a single carbon emission metric but involve systematic screening of physical boundaries, asset ownership, infrastructure independence, and baseline energy efficiency.
3) Change in Regulation: From "One-Time Evaluation" to "Full-Process Dynamic Management"
This notice is positioned to emphasize "technology, construction, and actual results, rather than one-time evaluation and certification." Units included in the national zero-carbon factory construction list must conduct self-assessments against the annual goals and key tasks determined in the construction plan and submit regular reports. In other words, getting on the list is not the end point; there are subsequent annual assessments, regular reporting, and spot checks.
Meanwhile, the October plan lists energy conservation supervision enforcement, diagnostic services, technology promotion, and standard support among measures "oriented toward key industries," and explicitly states the need to study and establish a carbon emission accounting mechanism for the industrial sector. For data centers, this implies a shift in management actions from "reporting PUE" to gradually "reporting carbon emissions."
02、What Actual Benefits Can Enterprises Realize After Investing Resources to Meet Standards?
It must be clarified that the realization of incentives depends on completion acceptance and formal certification before 2030, not merely entering the cultivation list.
Funding: Priority application for special funds for energy saving and carbon reduction, with local subsidies up to approximately CNY 2 million; central budgetary subsidies for low-carbon/zero-carbon demonstration projects can reach up to 20% of total investment.
Energy Use: After direct green power connection, the cost per kWh can drop by about half, and emission reductions can be developed as Chinese Certified Emission Reductions (CCERs) for monetization in the market.
Financing: Green loans and carbon footprint-linked loans can be implemented, with interest rates decreasing as PUE drops or emission reduction targets are met.
Orders: Procurement by central state-owned enterprises already requires recognition of green factory and zero-carbon park certifications; green qualifications are shifting from bonus points to entry conditions.
Screenshot of the "Guidelines for Key Industrial Technology Innovation Project Applications under Jiangsu Province's Manufacturing Powerhouse Construction Special Fund for 2026"
03、Restructuring of Location Selection, Costs, and Landscape
After the policy implementation, the operational logic of the computing industry will undergo substantive changes across three dimensions.
1) Location Logic: Shifting from "Electricity Price Priority" to "Availability of Physical Green Electricity Priority"
When physical green electricity becomes a core assessment indicator, enterprises must answer when selecting locations: Can the area provide stable direct green power supply? This explains the obvious divergence in energy consumption policies among recent hub provinces.
For example, Inner Mongolia Autonomous Region, in its "Several Opinions on Supporting the Development of Green Computing Industry in the Helan Cluster," set parallel requirements for new data centers: PUE ≤1.2 and green electricity ratio ≥80%. Qingyang, Gansu, achieved over 80% green electricity ratio in parks through "green power aggregation" while reducing the delivered electricity price to CNY 0.398/kWh. Under the new "physically traceable" regulations, physical green electricity provided by local nearby consumption in Zhongwei or source-grid-load-storage integration in Qingyang is the true asset that translates into certification scores. Hubs lacking local physical green electricity sources will face significantly increased difficulty in certifying their computing facilities.
Table: Review of Energy Consumption Policies for Data Centers in Eight Major Hubs
Data Source: MIR DATABANK compiled based on public information
2) Cost Re-evaluation: Converting Short-Term Compliance Investment into Long-Term Energy Advantages
In the short term, enterprises need to pay additional capital expenditures for direct green power connections, energy storage configuration, and carbon accounting systems. However, in the long run, locking in direct green power supply in areas with good resource endowments can indeed result in lower per-kWh costs than in eastern regions. The delivered electricity price of CNY 0.398/kWh in Qingyang is roughly half of the industrial electricity price in the east, serving as the most direct example. However, this figure comes from a combination of "green power aggregation + direct supply + local concessions" and is not a universal rule. To judge whether a project can be replicated, four accounts must be calculated simultaneously: the electricity price of green power itself, transmission/distribution prices and capacity charges, configuration costs for energy storage and backup power, and rack utilization and equipment usage rates. Comparing only electricity prices leads to overly optimistic conclusions.
3) Landscape Differentiation: Small and Medium Computing Enterprises Face "Implicit Elimination"
Applications must cross two thresholds simultaneously: reaching a scale of 3,000 standard racks and being selected into the National Green Computing Facility List. This means policy benefits are comprehensively concentrated among leading players. For small and medium-sized data centers, the issue is no longer "whether they can be rated as zero-carbon benchmarks," but a survival test of passing energy conservation inspections and accessing physical green electricity. Industry reshuffling will accelerate in the form of compliance thresholds.
However, the other side of "elimination" is switching tracks. The way out for small and medium computing enterprises is no longer on the path of "holding assets themselves and applying for benchmarks themselves," but in converting heavy assets into operational capabilities.
Three paths are being validated:
First, entrusted operation and maintenance. Assets belong to state-owned enterprises or leading players, while small teams output scheduling, O&M, and customer interface capabilities. Xinglian Technology connects 15 national-level supercomputers and 48 market-oriented AI computing centers using this "computing network" model, allowing partners to commercialize operations without building their own sales and O&M systems;
Second, turning O&M into a product. Deepening capabilities in energy storage scheduling, cold plate liquid cooling retrofitting, and airflow organization optimization allows customers' saved electricity bills to become the basis for charging fees. This path relies not on low prices, but on field capabilities that others cannot learn quickly;
Third, leveraging policy tools to enter lightly. MIIT has explicitly encouraged "dedicated computing pools for SMEs" and "computing banks." Many regions subsidize computing costs with "computing vouchers." Small players do not need to build their own server rooms but can flexibly access resources and match orders by "GPU-hours" or "core-hours."
Conclusion
This policy upgrades the low-carbon proposition of the computing industry from an "efficiency race" centered on equipment energy efficiency to a "structural race" centered on carbon emission intensity and physical green electricity supply capability. For computing enterprises, the real watershed is not whether they can lower PUE by another 0.01, but whether they can establish a verifiable, certifiable, and sustainable physical green electricity supply system in their region. Whoever completes this switch first will gain the initiative in the next stage of entry and cost competition.