Four Departments Have Issued a Significant Document: Ai and Energy Mutually Enhance Each Other. What New Sectors Will The Energy and Power Industry Embrace?

On May 8th, the National Energy Administration and four other departments jointly issued the "Action Plan for Promoting the Two-way Empowerment of Artificial Intelligence and Energy", outlining 29 key tasks. This article dissects the four structural opportunities in the energy and power industry from an industrial perspective.

On May 8, 2026, the National Energy Administration, in conjunction with the National Development and Reform Commission, the Ministry of Industry and Information Technology, and the National Data Bureau, issued the "Action Plan for Promoting the Two-way Empowerment of Artificial Intelligence and Energy" (Document No. 34 of 2026 from the National Energy Administration), deploying 29 key tasks. The document states that by 2027, a supporting energy guarantee system for the innovation of artificial intelligence will be initially established, and by 2030, the supply guarantee capacity of clean energy for artificial intelligence computing facilities will reach the world-leading level.

This is China's first special policy document targeting the integration of AI and energy development. After "computing and electricity coordination" was included in the new infrastructure project in the 2026 government work report, the four departments completed the institutional closed-loop from concept to implementation with an action plan.

The document unfolds throughout with a single main thread: energy supports the development of artificial intelligence, and artificial intelligence empowers energy transformation. The two-way empowerment is not a simple overlay of the two industries, but a systematic reorganization of the "energy, computing power, scenarios, data, and models" five elements.

For the energy and power industry, the value of this document lies not in the macro narrative, but in the industrial opportunities implied in the 29 tasks. Each task direction corresponds to a clear incremental demand.

I. Green Electricity Direct Connection: From Pilot Exploration to Large-Scale Deployment

The document states that it is exploring the direct connection of nuclear power, hydrogen energy and other energy sources to supply power to computing facilities. It encourages computing facilities to configure grid-connected energy storage and improve policies for green electricity direct connection for computing facilities. It also studies the use of price policies as an incentive.

"Green electricity direct connection" is not a new concept. In December 2025, Guangdong Energy Group operated a 300,000-kilowatt photovoltaic computing-electricity integration project in Karamay, Xinjiang, with accompanying energy storage facilities, achieving a reduction in comprehensive energy costs by over 40%. The green electricity direct supply rate of multiple "point-to-point" green electricity supply projects to computing centers across the country has generally exceeded 80%. However, these projects were mostly local pilot projects and lacked national policy endorsement.

The breakthrough of this document lies in three aspects.

Firstly, it is the first time that nuclear power and hydrogen energy have been included in the official policy context of energy supply for computing power. This means that the energy supply for computing facilities has expanded from being mainly based on "wind and solar power" to multiple energy forms, providing differentiated paths for regions with different resource endowments.

Secondly, it is clearly stated that "based on the type of computing tasks, computing facilities should be managed in a classified manner". This means that not all data centers need to be directly connected to green electricity, but computing facilities with flexible regulation capabilities will receive policy preferences. The precision of the policy reduces the compliance costs for enterprises.

Thirdly, it is proposed to "explore the collaborative construction of 1 million-kilowatt artificial intelligence computing facilities and their supporting energy systems". This is a quantitative leap - the previous data center support was at the level of tens of thousands of kilowatts, and a 1 million-kilowatt level means that the computing facilities themselves are becoming the core load nodes of the energy system.

For the energy and power industry, at least four development paths have emerged here.

New scenarios for the development of new energy power stations. When "the collaborative construction of 1 million-kilowatt computing facilities and their supporting energy systems" is written into the document, the development logic of the western new energy bases changes - from "generating electricity and sending it out" to "generating electricity and being considered". The economic model of the source-network-load-storage integrated project will be rewritten.

Grid-forming energy storage has transformed from a technical direction to a necessary equipment. The document clearly states "encouraging computing facilities to configure grid-forming energy storage to enhance power supply stability and the active support capability for the power system". Guangdong's "15th Five-Year Plan" has already listed grid-forming energy storage as a key direction for the addition of 5 GW, and with the national endorsement of this document, the industrialization process of grid-forming energy storage will significantly accelerate.

Engineering increment of micro-grid EPC. Direct connection of green electricity is not as simple as just pulling a dedicated line; it involves the complete engineering system of power supply side, load side, power quality governance, energy storage configuration, and dispatching system. EPC enterprises with comprehensive energy system integration capabilities will directly benefit.

Upgrade demand for the power trading system. The document mentions "supporting computing facilities to increase the proportion of green electricity through participation in green certificate and green electricity transactions", which means that the power trading platform needs to have new capabilities to handle the characteristics of computing load - time-of-use pricing, green certificate traceability, and carbon accounting linkage.

Jun's perspective: The direct connection of green electricity is transforming data centers from mere "electricity consumers" to "energy system co-builders". For power and electrical enterprises, this means that the downstream demand structure is undergoing a fundamental change - the customer list no longer consists only of power grid companies and power plants, but also includes computing power operators and AI enterprises. Whoever possesses the combined capability of "green electricity resources + dedicated line connection + energy storage configuration" will secure the energy entry point in the computing power era.

II. Computing-Grid Synergy: Computing Load Becomes a “New Variable” in the Power Grid

The document proposes using electricity market price signals to guide computing facilities in optimizing energy management and cross-network, cross-regional dispatch, and encourages computing facilities to participate in grid operations as flexible, adjustable resources on the demand side.

This passage is extremely dense with information. When broken down, it sends three key signals.

The first signal is that computing power is set to participate in the electricity market. In the past, data centers were “major users” of the power grid—consuming large amounts of electricity with stable loads, yet they were largely passive recipients of electricity prices. Now, the document explicitly states that computing power scheduling should be “guided by electricity market price signals,” which means that the electricity consumption patterns of computing facilities will shift from rigid to flexible. During peak solar generation hours, training tasks can be scaled up, while during evening peak hours, non-urgent computing tasks can be postponed or shifted to western regions. Computing load has thus become a flexible regulation resource for the power grid.

The second signal is that the scope of virtual power plants is expanding. The essence of a virtual power plant is to aggregate dispersed, adjustable resources to participate in grid dispatch. In the past, it primarily aggregated interruptible loads from industry and commerce, distributed energy storage, and electric vehicle charging stations. Now, computing loads have been formally incorporated into the category of “flexible, adjustable resources.” For a 100-megawatt data center, if 30% of its load is adjustable, its annual revenue from participating in demand response could reach tens of millions of yuan. When hundreds or even thousands of data centers are integrated into the virtual power plant system, the market size will reach tens of billions of yuan.

The third signal is the deep integration of the “East Data, West Computing” initiative and the “West-to-East Power Transmission” program. The computing-power synergy intelligent dispatch platform, independently developed by State Grid Information & Communication, has been deployed in 12 provinces, enabling millisecond-level dynamic matching between computing load and power supply. In projects such as Zhongwei, Ningxia, green electricity accounts for over 60% of the mix, with electricity prices at just 45% of those in eastern regions. A viable business model is already taking shape in this area.

For the energy and power sector, the synergy between computing and electricity is giving rise to a new set of opportunities. Virtual power plant system integrators must understand both electricity market trading rules and the characteristics of computing loads; power spot trading software providers need to design new trading algorithms tailored to computing loads; load aggregators specialize in aggregating computing facilities to participate in ancillary service markets; and power quality management equipment suppliers benefit from the extreme sensitivity of computing facilities to issues such as voltage sags and harmonics.

He Jun's perspective: The direct connection of green electricity is transforming data centers from mere "electricity consumers" to "energy system co-builders". For power and electrical enterprises, this means that the downstream demand structure is undergoing a fundamental change - the customer list no longer consists only of power grid companies and power plants, but also includes computing power operators and AI enterprises. Whoever possesses the combined capability of "green electricity resources + dedicated line connection + energy storage configuration" will secure the energy entry point in the computing power era.

III. Scene Opening: From Data Isolates to High-Value Scenarios

The document proposes the establishment of a mechanism for selecting high-value scenarios and publishing their lists, as well as the construction of an open and shared platform for energy-related scenarios. It also envisions the establishment of a full-life cycle closed-loop management mechanism covering scenario release, research and development, testing and verification, engineering implementation, and effectiveness assessment.

This is the most innovative part of the document in terms of institutional innovation.

It addresses a long-standing structural contradiction: Energy enterprises have the most abundant application scenarios and massive data, but lack AI talents and engineering capabilities; AI enterprises have the most advanced algorithms, but lack real scenarios and high-quality data.

The "High Value Scenarios in the Energy Field" section in the document lists specific directions for five sub-sectors. Clean energy for reliable and flexible supply, including wind and solar power prediction, smart operation of power stations, intelligent perception of hydropower projects, and abnormal identification of nuclear power operations. Grid safety and stable operation, including intelligent assessment of grid planning, equipment defect diagnosis, smart diagnosis of distribution networks, and decision-making in power markets. Intelligent and efficient development of coal, including unmanned operation of mining faces, predictive maintenance of equipment, and safety intelligent warnings. Efficient exploration and development of oil and gas, including intelligent exploration modeling, drilling optimization, and digital twin basins. Diversified integration of energy new forms, including intelligent operation of energy and electricity coordination, coordinated scheduling of virtual power plants, and safety warnings for new energy storage.

Almost every sub-sector corresponds to a clear industrial opportunity. For power and electrical equipment enterprises, this means that the definition of products is being rewritten: transformers are no longer just a lump of iron, but intelligent nodes with sensors, edge computing modules, and fault diagnosis algorithms; switch cabinets are no longer just for on-off control, but data entry points embedded with load forecasting and energy efficiency optimization capabilities.

He Jun's perspective: The essence of scenario opening is to transform the "hidden knowledge" accumulated by the energy industry over several decades into replicable AI products. Those enterprises that have already made preparations for the intelligentization of equipment will gain a leading position in this wave of scenario opening. While those enterprises that are still stuck in the mindset of "selling hardware" will be rapidly marginalized. This is not a gradual substitution process, but a rapid redefinition.

IV. Professional Large Models: More than Five Models. Who will do it?

The document states that it will focus on areas such as power grids, power generation, coal, oil and gas, and comprehensive energy, and promote the in-depth application of more than five professional large-scale models in the industry. It will also accelerate the adaptation and optimization of self-developed intelligent computing chips and domestic deep learning frameworks.

The quantitative target of "more than five" is worthy of consideration.

Based on the current publicly available information, Huawei's Pan Gu power industry large model, Baidu's power industry large model, iFLYTEK's energy industry solution, and the dispatching large model independently developed by power grid companies are all in the process of being established. However, "more than five" is not the endpoint; it is merely a threshold.

The core judgment here is: The competitive barriers of industry large models have never been the algorithms themselves, but rather high-quality data and the engineering implementation capabilities. Algorithms can be open-sourced, models can be fine-tuned, but the ten-year sensor data, equipment failure cases, and dispatching experience accumulated in power grid operations - these are irreplicable.

The document also emphasizes "promoting the convergence and integration of industry data into professional large models" and "accelerating the adaptation and optimization of self-developed intelligent computing chips and domestic deep learning frameworks". Hardware autonomy and data convergence are the true moats of energy AI.

For the energy and power industry, the beneficiary directions include: energy data annotation and governance service providers, as large models require high-quality data, and the cleaning, annotation, and standardization of energy data is a huge professional service market; domestic intelligent computing chip adaptation service providers, as the adaptation and migration of Huawei Ascend, HuaMin, etc. domestic chips in energy scenarios require a large amount of engineering services; power industry AI application integrators, who embed the capabilities of large models into specific business processes.

He Jun's perspective: The competition in the energy big model field is ostensibly an algorithm battle, but fundamentally it is a data battle. Energy and power enterprises hold the most crucial industry data - equipment operation data, fault records, load curves. The value of these data was severely underestimated in the past. When large models need these data for training, the enterprises that possess the data gain the pricing power. The question is, have you already realized the value of the data in your hands?

V. Cold Thinking: The Three Major Challenges Beyond Opportunities

While sorting out the track, we also need to maintain the necessary prudence. The benefits of two-way empowerment do not come without constraints.

First, the economic feasibility still needs to be verified. Configuring energy storage for computing facilities, participating in the power market, and building direct green electricity connections all mean an increase in upfront investment. Currently, the operators of computing centers generally focus on PUE and operating costs, and there is still a need for clearer expectations regarding the revenue model brought by participating in grid regulation. The document mentions "price policy incentives" and "green electricity trading contracts" among other directions, but specific aspects such as time-of-use electricity pricing mechanisms, auxiliary service compensation standards, and green certificate revenue distribution still require clarification through local regulations and pilot practices.

Second, the balance between security and openness needs to be finely grasped. The energy industry belongs to critical information infrastructure. If the operational data of power grids, oil and gas pipelines, etc. is leaked or abused, it may bring risks that cannot be ignored. The document emphasizes "data classification and grading" and "privacy computing" among other technical means, but in actual operation, which scenarios can be opened up and which data can be shared still need to reach an industry consensus and operational standards. Moving too fast may bring security risks, while moving too slowly is not conducive to the release of data value.

Third, there is a shortage of multi-skilled talents. Multi-skilled talents who are familiar with power system operation and also understand artificial intelligence algorithms are currently relatively scarce in the market. The document proposes measures such as "industry-academia integration discipline clusters" and "open source communities", which are in the right direction, but the training cycle for talents is long, and it may still be a constraint in the short term.

He Jun's viewpoint: The period of policy benefits usually lasts only two to three years. The goal of "initial construction" in 2027 means that the next two years will be a crucial window period for various regions to intensively introduce supporting regulations, launch pilot projects, and allocate resources. During this window period, those who act first will enjoy resource advantages and the advantage of being the pioneer; after the window period, the overall situation will basically be formed, and those who follow will only be able to catch up. For energy and power enterprises, now is not the time to wait and observe, but the time to secure a favorable position.

Conclusion

This document was released exactly ten days ago. Its popularity has not yet waned, but not many people truly understand it.

Most people see the concept of "AI and energy", but we see that the power and electrical industry is being redefined - from the traditional linear chain of "generation, transmission, distribution" to a networked ecosystem of "energy, computing power, data, models".

In this new ecosystem, the strategic value of electricity is being revalued. For China's energy and power enterprises, two-way empowerment is not only an opportunity for order growth, but also a window period for transitioning from traditional manufacturing to a digital infrastructure service provider.

After all, in the AI era, those who control computing power win the competition, while those who control electricity control computing power.

Note: The policy document is sourced from the official website of the National Energy Administration, and the industry data is from public reports and industry research reports. Some predictive data is for reference only.