On May 20, 2026, the National Development and Reform Commission (NDRC) and the National Energy Administration (NEA) jointly issued the Notice on Matters Concerning the Orderly Development of Multi-User Direct Green Power Connections (NDRC Energy [2026] No. 688; hereinafter “Document No. 688”). Its most important change can be summed up in one line: direct green power connections have officially moved from a one-to-one model to a one-to-many model.
For the data center industry, this is more than a regulatory breakthrough. It rewrites the logic of how computing infrastructure can source and use electricity.
What Document No. 688 Changes for Data Centers
- A Clearer National Signal: From “Encouraged” to “Priority Support”
For the first time in a national-level policy document, Document No. 688 states that priority support should be given to direct green power connections for emerging and future industries, including computing facilities and green hydrogen, ammonia, and methanol. The stronger wording signals a substantive shift in resource allocation toward computing and other emerging industries.
The policy shift comes at a time when computing infrastructure sits at the intersection of two major priorities under the 15th Five-Year Plan: the digital economy and the new energy system. That makes data centers one of the most suitable types of load for absorbing renewable generation.
In other words, data centers are no longer treated simply as passive electricity consumers; they are now explicitly positioned among the priority beneficiaries of direct green power connections.
- A New Model: From One-to-One to One-to-Many
The central breakthrough in Document No. 688 is the introduction of a multi-user model. Under a multi-user direct green power connection, renewable generators — including wind, solar, and biomass facilities — do not supply users through the public grid. Instead, dedicated lines and substations deliver green electricity directly to multiple users.
This change has three major effects:
First, it lowers the entry barrier. Under the previous single-user model, dedicated transmission lines were expensive and the required energy-storage configuration created additional economic pressure. The new policy allows several users to share the same arrangement, aligning much more closely with aggregated loads in computing clusters and industrial parks. The industry generally views this as a key institutional step from one-to-one to one-to-many direct green power supply, opening the door to deployment at greater scale.
Second, it spreads risk. Under the old model, one wind farm might be tied to one factory; if that user withdrew, the entire project could be put at risk. In a multi-user structure, the departure of a single user does not bring the whole project to a halt.
Third, it broadens the range of viable use cases. A representative of Jinko Power has noted that cost sharing and aggregation mechanisms can extend the benefits of green electricity to a much wider group of small and medium-sized enterprises. In practical terms, the opportunity is expanding from a handful of large users to entire industrial-park ecosystems, creating more room for direct green power connections in energy-intensive settings such as industrial parks and data centers.
- Mechanism Design: Matching Supply to Load, Backed by Quantitative Targets
To ensure that direct green power projects actually consume the renewable electricity they are built around, Document No. 688 sets out strict quantitative requirements:
Projects should size renewable generation according to the principle of matching generation capacity to load. Annual self-generated and self-consumed electricity must account for at least 60% of total available generation and at least 30% of total electricity consumption, rising to no less than 35% before 2030.
Data centers have a particular advantage here: they combine large, highly reliable base loads with a degree of demand-side flexibility. That makes them natural anchor loads for a supply-follows-load model. It also highlights the fundamental difference between direct green power connections and green-certificate trading — moving from ex-post accounting to real-time matching of generation and consumption.
- Governance: A Single Lead Entity for Each Project
Coordination becomes the biggest challenge in a multi-user arrangement. Document No. 688 therefore requires each project to designate a single legally incorporated lead entity, in principle established jointly by the power-generation and load parties. This entity takes overall responsibility for project planning and construction, user coordination, day-to-day operations and maintenance, grid interfacing, and lifecycle management.
As a general rule, the lead entity is also responsible for investing in and building the project’s connection lines, substations, energy-storage facilities, and operating platform. The project and its internal resources are exempt from power-business licensing requirements.
Early Implementation: Xinjiang Moves First
Just two days after Document No. 688 was released, Xinjiang approved its first two multi-user direct green power projects serving computing loads on May 22, 2026.
The two projects — the Computing Innovation Demonstration Zone in Yiwu County, Hami, and the Cloud Computing Industrial Park in Karamay — both serve data-center loads. Together they include 160 MW of renewable generation capacity: 40 MW of wind and 120 MW of solar. Once operational, they are expected to consume 263 million kWh of green electricity annually, equivalent to saving 79,000 tonnes of standard coal and avoiding 213,000 tonnes of CO2 emissions.
The more striking change is in utilization. The self-consumption rate of renewable generation is expected to rise from the current 30–50% to more than 90%, with green electricity supplying over 60% of total power consumption in the demonstration zones. Zhang Xiaoyi, director of the Hami Development and Reform Commission, put the objective plainly: the project is intended to ensure that computing companies can both afford green electricity and actually access it.
This is not an isolated case. It is the first real test of how Document No. 688 moves from policy language into engineering practice.
But Document No. 688 Also Raises Hard Questions for the Industry
Question 1: Lower Electricity Prices May Be More Hope Than Reality
There has long been a simple expectation in the industry: direct green power could bypass the grid, reduce electricity costs for users, and raise returns for generators. The problem is that most projects cannot truly operate without the grid. Grid-connected projects — the choice for the vast majority of data centers — still have to pay transmission and distribution charges, system-operation fees, policy-related cross-subsidies, government funds, and surcharges. The capacity-based tariff mechanism established under Document No. 1192 adds another layer of pressure: the larger the declared grid-connection capacity, the higher the capacity charge; reduce that capacity, and outage risk rises. At low load factors, fixed costs are spread over fewer kilowatt-hours and the effective cost per kWh increases. High-load-factor data centers, meanwhile, must declare connection capacity close to their full load, so their per-kWh allocation is also difficult to reduce. Either way, the economics are uncomfortable — which means the promise of materially cheaper power is likely overstated for many projects.
Even in regions rich in renewable resources, the all-in cost of electricity under direct green power arrangements can remain noticeably higher than local industrial power tariffs.
Question 2: The Fundamental Tension Between “Absolute Reliability” and “Natural Variability”
Data centers require highly reliable power with virtually no room for compromise. AIDC load characteristics mean they cannot shift their electricity use as freely as some industrial facilities can, while wind and solar generation are inherently intermittent and variable. Document No. 688 does not eliminate this mismatch; in practice, much of the burden is shifted to energy storage. But storage is itself a major incremental cost. A 110 kV direct-connection line can already cost more than RMB 1 million per kilometer, before substations and storage systems are added.
Put simply, every one-cent saving on green electricity may require two cents of additional spending on storage to buy back reliability — and the industry still does not have a definitive answer for that gap.
Question 3: Grouping Users Lowers the Barrier — but Creates New Governance Problems
The defining feature of the one-to-many model is that users can participate as a group, but that also creates new organizational challenges. Document No. 688 requires a legally incorporated lead project entity. In practice, cooperation across multiple parties can run into structural problems around exit mechanisms and the distribution of benefits. One major renewable-energy province in western China, for example, requires participants to remain in a project for three years after it begins operation, while a cross-party exit must be initiated jointly. If one participant runs into financial difficulty but another refuses to agree, the exit process cannot move forward. Governance issues such as electricity allocation, sharing storage costs, and priority rules when green power is insufficient therefore need to be designed in advance.
The single-user model tests whether a project can be built. The multi-user model tests whether it can be governed.
Question 4: Provincial Rules Are Becoming a New “Invisible Threshold”
Document No. 688 leaves many important implementation details to provincial authorities, which means outcomes can differ sharply from one region to another. Some provincial rules already go further than the national policy. For example, the share of surplus electricity allowed to be fed into the grid has in some cases been tightened from “generally no more than 20%” to “in principle no more than 10%.” Energy-storage facilities may also be barred from participating independently in ancillary-services markets or from being leased externally for profit. The same national policy can therefore translate into very different project economics and operating room across provinces.
The regions that establish clear, workable implementation rules first will gain an advantage in attracting green, power-intensive industries.
Question 5: Technical Redesign Is Becoming the Only Real Way Through
Policy has opened the door, but the economics still do not always add up. That tension is pushing the industry toward fundamental changes in power architecture. TGOOD, for example, has introduced a “Compute-Power Island” in which wind and solar generation can connect directly to an 800 V DC bus, avoiding losses from multiple AC/DC conversion stages. The design combines direct-connected energy storage with dynamic grid-forming technology and is intended to reduce full-site delivery time from 12–18 months to about 150 days. This is more than an equipment upgrade; it is a redesign of how the entire power system is controlled. As large volumes of distributed, intermittent renewable generation connect to computing centers, traditional one-way radial grids will face far greater pressure in power-flow control and relay protection.
The first compute–power coordination pilot projects have already entered engineering implementation in Inner Mongolia and Ningxia. Whether the technology can genuinely bridge the economics gap will be one of the defining competitive questions for the industry’s next phase.
Document No. 688 establishes the policy ceiling: direct green power connections now have a formal institutional pathway. What will determine success, however, is who can first solve three practical variables — the economics of energy storage, multi-party governance, and workable provincial implementation rules. For the data center industry, direct green power is no longer a yes-or-no question of whether to participate; it is a question of how to make the model work. For more discussion, follow DITExpo 2026, taking place in Beijing on December 10–11.
