Let the weaker storage factories close

In 2022 we could not buy enough lithium iron phosphate cells. Suppliers rationed their output, prices for storage cells rose above RMB 0.9 per watt-hour at their peak, and our integration plant at Zibo waited for deliveries. Two years later we could buy cells for about a third of that price, and suppliers were telephoning us. Domestic tender prices for complete two-hour storage systems fell from about RMB 1.2 per watt-hour early in 2023 to below RMB 0.55 by late 2024, and some bids went lower still.

Lower prices have been good for storage owners, including us, and for the energy transition. They have not been good for the factories that make the equipment, many of which now sell below the full cost of what they produce. I want to argue something unpopular in my industry: the price war will end sooner, and with less damage, if the weaker factories are allowed to close.

How a shortage became a glut

The glut was built rationally, one decision at a time. Demand for electric vehicles and storage grew faster than almost anyone forecast between 2020 and 2022. Cell makers announced expansions. Local governments, eager for manufacturing investment, offered land, cheap power, equipment subsidies and introductions to bank finance. Newcomers with no history in batteries, from appliance makers to property developers in search of a new business, built plants. Each plant made sense to its owner at the prices of its year. Together they created capacity far beyond what the world could use in the near term.

Integrators like us added to it. In 2023 our board approved a third assembly line at Zibo, taking capacity to 12 gigawatt-hours a year. I supported that decision, and I was wrong. The line was finished in 2024, the year domestic prices collapsed. Our plant ran at 46 per cent of capacity that year and 56 per cent in 2025. I mention it because an argument about overcapacity ought to begin with the arguer's own share of it.

The economics of an overbuilt industry are cruel in a particular way. A factory that covers its cash costs can keep running even if it never earns back its capital, so prices fall towards the cash cost of the weakest producer still willing to sell. Strong producers then lose money alongside weak ones. Quality is the first casualty, because the cheapest way to cut cost is to test less, sample less and tolerate a little more variation between cells. In a battery that variation is not cosmetic. It is the difference between a module that ages evenly and one that overheats.

Support that keeps losses alive

Overcapacity corrects itself when weak producers stop producing. In China that correction has been slow, because many plants carry obligations beyond their owners. A local government that brought a factory to its district wants the jobs and tax revenue it was promised. A local bank that lent against the plant prefers rolling over the loan to recognising the loss. Suppliers are owed money. So the plant keeps running at a loss, sells at whatever price fills its order book and pulls the market down for everyone.

Central authorities have recognised the problem. Since 2024 they have spoken repeatedly against disorderly low-price competition, and the industry conditions for battery manufacturing have raised expectations on technology and quality for new capacity. I support both. But exhortation does not close a factory, and a standard written for new plants does little about an existing plant that sells at a loss to stay open.

What would help is less dramatic. Local governments could stop extending support to plants whose losses have run for several years, and allow their assets to pass to stronger operators or to close. Lenders could price the risk of a factory's survival honestly instead of waiting for a recovery that depends on everyone else's restraint. And industry associations could publish more about the real cost of producing safely, so that a bid below that cost is recognisable for what it is.

I understand the human cost. When we did not renew the agreements of 310 contract production workers at Zibo in 2024, I met some of them, and none was comforted by an argument about market structure. Closures are hard on towns that depend on a single plant. But prolonging losses does not save those jobs. It spreads the damage across the industry and delays the moment when the survivors can invest again in safety and research.

What buyers should pay for

Buyers decide how long a price war lasts. Most large storage purchases in China are made by state-owned generators and grid-related companies through public tenders, and many tenders still weight the lowest price heavily. A supplier selling below cost can win such a tender, and does, because the scoring rewards it. The buyer then owns a station for fifteen years or more whose cells came from a factory that may not exist to honour its warranty in year eight.

In 2024 we stopped bidding below a floor of RMB 0.58 per watt-hour for a standard two-hour system: our full cost, a reserve for fifteen years of warranty and service, and a thin margin. We lost a 1.6 gigawatt-hour framework tender from a central state-owned generator that cleared at about RMB 0.47. We lost others. Our share of the domestic tenders we tracked fell from about five per cent to under three. I do not claim our floor was perfect. I do claim that a bid below the full cost of a warranty transfers risk to the buyer, whether or not the buyer notices.

Buyers can notice. They can score bids on lifetime cost, including expected degradation, availability and the cost of replacing a supplier that fails. They can ask suppliers to show the financial strength behind a fifteen-year warranty, as a lender would. They can require cell traceability, incoming inspection and teardown sampling, and they can visit the factory before they sign. Several buyers have moved this way since 2025, and their tenders already look different: fewer bidders, higher prices and more attention to safety records.

A smaller industry with fewer, stronger factories would still be large by any standard, and prices would not return to their 2022 level. They would settle where a careful producer can earn a modest return. That is the price at which storage stays safe for the twenty years a station must run. The alternative is a long period in which everyone loses money and the losses surface, eventually, in the stations themselves.