China's new storage plan is usually reported as a 300 GW target for 2030. The more revealing number is three hours. At 300 GW and an average duration of three hours, the plan implies about 900 GWh of storage energy capacity. Relative to end-2024, that is roughly five times the energy stock and a faster multiple than the roughly fourfold increase in power capacity. The shift matters because grids pay for energy moved across time and capacity available at peaks, not for megawatts on a datasheet.
The document is the 15th Five-Year New Energy Storage Implementation Plan, issued by the National Development and Reform Commission and the National Energy Administration on 13 November 2026, with a document date of 29 October. It sets 2030 at 300 GW, average duration around three hours, and moves the sector from a scale-up phase to a market phase. It has eight parts, 16 articles and six boxes.
The 300 GW figure is a floor, not a sprint. The internal split is more useful. Provincial self-use storage is to exceed 240 GW by 2030, including about 110 GW of new capacity during the 15th Five-Year Plan. Storage configured with new energy bases is 50-60 GW. Grid-side independent storage is about 140 GW, but that is a functional slice for peak supply, not an additional 140 GW on top of the total. The 240 GW plus the 50-60 GW roughly reconciles with 300 GW.
The starting point is already high. China had 167 GW of new-type storage at the end of September 2026, up from 136 GW at the end of 2025 and about 3 GW at the end of 2020. At the 2026 run rate, the stock would be about 177 GW by the end of 2026. The plan then implies a net addition of about 123 GW from 2026 to 2030, or roughly 31 GW a year. That is about half of the roughly 62 GW added in 2025 and about a quarter below the annualized pace of the first nine months of 2026, around 41 GW. The official interpretation calls 300 GW a prudent, scientifically calculated floor intended to prevent an investment rush; the supply-side text warns against blind expansion. Demand gets a floor, supply gets a ceiling.
The duration target is where the engineering signal sits. A two-hour lithium project built purely for power capacity has less marginal value in this structure. The plan favors two technical lines: middle-duration, high-safety systems and long-duration, low-cost systems. Lithium-ion continues to iterate; sodium-ion moves from demonstration to promotion; compressed-air and flow batteries move from pilot to commercial application; solid-state, hydrogen and thermal storage are listed as frontier reserves. Grid-forming storage gets a separate mention, with standards for planning, configuration and grid connection.
The application list is broader than arbitrage. It includes base transmission, grid peak capacity and user-side storage, plus green-power direct connection, zero-carbon parks, virtual power plants and vehicle-to-grid. It also names the eight national computing hubs and the Qinghai, Xinjiang and Heilongjiang computing-storage coordination areas. Data centers are now part of the storage demand story.
The most consequential change is in the rules. Dispatch is split by market state. In provinces where the spot market runs continuously, storage is dispatched according to market clearing results. In provinces where it does not, dispatch can still consider economics and system need. In an emergency, all storage can be directly controlled. On price, the plan establishes a capacity price mechanism for grid-side independent storage under list management. After the spot market runs continuously, a reliable capacity compensation mechanism is to be built in an orderly way under unified principles. Provincial details are left open.
List management is the gate. A project must be on the provincial list, selected against peak supply and system needs, not simply built and registered. Build-attached storage is not directly covered by the capacity price; its value outlet is the capacity leasing market. The plan gives the category a name and a route, but not a number. Compensation standards, list conditions and provincial timing are all blank.
A sample calculation from the source shows why that matters. Provincial capacity compensation in places such as Shandong, Xinjiang and Gansu has been roughly 100-330 yuan per kW per year. For a 100 MW/400 MWh independent project, a capacity payment of about 92 yuan per kW per year for three years would be about 30% of first-year revenue and then fall to zero. The source's all-in IRR is about 1.4%, with negative equity return. That is a specific project, not a universal result, but the arithmetic is a warning: a capacity mechanism can lower the variance of returns without raising the expected return.
The plan therefore does two things at once. It legitimizes a revenue category and reduces policy uncertainty. It does not set a price or guarantee a customer. The real tests are provincial: can the project get on the list, and can it earn a spread in the spot market? A two-hour pure arbitrage project in a low-spread province is structurally weak. A longer-duration project at a constrained node, with capacity value and grid services, has a better case. The 300 GW headline is the easy part. The three-hour average and the list are the actual policy.