Are all the terawatts needed? Regulators are holding back new battery manufacturers. Western news reports suggest there is major production overcapacity in lithium batteries for electric producers. Jiemian News (界面新闻) has a more nuanced take.
Sodium-ion & solid-state are encouraged by a tax exemption. From Claude
The freeze is an entry barrier, not a production cut
On 6 September 2026, Cailianshe (财联社) reported that Chinese regulators had begun a comprehensive inventory of existing and planned battery manufacturing capacity, with storage cells (储能电芯) a priority area. Projects still at the planning stage and not formally broken ground would have approvals and advancement suspended. Filed projects (已备案) under construction were unaffected. The measure was described as subject to dynamic adjustment rather than a blanket halt (并非一刀切).
English-language pickup framed this as China curbing overexpansion. The Chinese reporting supports a narrower and more consequential reading.
Jiemian News (界面新闻) confirmed the substance with several battery companies. A senior executive at a domestic battery firm, given the pseudonym Huang Yi (黄毅), said controls target the battery segment specifically, new capacity is restricted, and departmental-level approvals have basically stopped. But an energy storage company source added the qualifier that reframes everything: most lithium companies’ construction capacity started in the past two years and is now near completion, so those projects are untouched — while new players will likely find it hard to enter the market going forward. A second-tier manufacturer separately confirmed its own new capacity had cleared approval early, limiting the impact on the company.
The freeze therefore does not reduce output. It closes the door behind the firms already inside. For incumbents this is a moat. For the industry’s competitive structure it may be the most consequential Chinese battery policy of the year, and it is not what the headlines described.
A second transmission channel runs through capital rather than permits. Mo Ke (墨柯) of RealLi Research (真锂研究) has argued that if approvals keep tightening, storage-focused battery makers currently in a critical financing window face heavier capital pressure and may adjust production rhythms, while diversified leaders are relatively insulated. The policy bites at the financing stage before it bites at the permitting stage.
The document trail runs through a county government
No ministry has confirmed any of this. Jiemian sent an interview letter to MIIT on 8 September and had received no reply by publication. The company source interviewed on 7 September opened by noting rumours had circulated since early in the year.
The only known written artifact is local. In June, a public document on the website of Zhongjiang County government (中江县, Deyang, Sichuan) stated that MIIT would lead a nationwide comprehensive survey of lithium battery capacity and establish an early-warning mechanism; until that mechanism was formally established, all regions must suspend procedures for power and storage lithium battery projects. Consumer battery and lithium battery materials projects were exempt. Districts and counties were instructed to survey capacity in their jurisdictions immediately, to prepare for the national count and avoid information asymmetry affecting subsequent work.
A county-level restatement of national policy, three months before the story broke nationally. Retrieve it from the county site before citing it, and before the site removes it.
| Date | Action |
|---|---|
| 28 Nov 2025 | MIIT convenes battery manufacturers; Minister Li Lecheng (李乐成) calls for governing irrational competition by law and strengthening capacity monitoring, early warning and regulation |
| 7 Jan 2026 | MIIT, NDRC, SAMR and NEA hold joint symposium on competition order in the battery industry |
| ~Jun 2026 | Zhongjiang County notice; national capacity survey flagged |
| 30 Jul 2026 | Launch meeting for group standard on power and storage battery cost accounting |
| 3 Aug 2026 | 15th Five-Year Plan for new power system construction issued (发改能源〔2026〕942号) |
| 1 Sep 2026 | 2% consumption tax on lithium batteries takes effect |
| 6 Sep 2026 | Cailianshe capacity inventory report |
| 10 Sep 2026 | Caixin adds year-end review and early-warning mechanism detail |
How much capacity is actually being built
Three datasets, three scopes, no clean reconciliation.
CESA Energy Storage Application Branch, first seven months of 2026: 57 companies expanding lithium capacity; the top ten disclosed 1,787 GWh of designed capacity, or 68.5% of the total. That implies roughly 2,600 GWh of disclosed design capacity in seven months, covering power and storage lithium combined.
GGII, H1 2026: nearly 20 storage battery projects launched, planned capacity above 500 GWh, total investment above RMB 47 billion, up more than 110% year-on-year — the fastest-growing new-capacity sub-sector in lithium.
Envision SVP Tian Qingjun (田庆军): industry planned storage cell expansion above 800 GWh this year, built capacity of roughly 1.2–1.5 TWh by year-end, total planned capacity approaching 2 TWh, far exceeding real global demand in his assessment.
Against Tian, Zhang Jinhui (张金惠) of Xinluo Information (鑫椤资讯) argues planning and building are different things; strip out the exaggeration and storage cells remain in shortage. Cell lines have run at full capacity since Q3 2025 and would stay full into next year even if orders stopped, because production is scheduled against project orders and long-term supply agreements — even framework agreements specify volumes for future periods. Zhang’s demand case: global storage cell shipments of 1,200 GWh this year rising to 1,600 GWh next year, with a 20–30% CAGR to 2030, at which point storage and power battery volumes would be roughly comparable.
The reconciling datapoint is CNESA’s. Installed storage cell capacity nationwide stood at 809.5 GWh at end-June, excluding flexible lines switchable between EV and storage production. That is well short of Tian’s year-end projection and it is the figure closest to physical reality. CNESA also notes growth in new cell capacity additions had already slowed noticeably before any freeze.
Demand is consolidating, not collapsing
CNESA half-year data, released by chairman Chen Haisheng (陈海生) at the 11th Western Energy Storage Forum on 27 August:
- Cumulative new-type storage: 168.3 GW / 448.7 GWh, up 59% / 71% year-on-year
- Total electricity storage: 237.7 GW, up 41.7%, with pumped hydro falling below 30% of the total for the first time
- H1 newly commissioned new-type storage: 21.81 GW / 58.60 GWh, down 18% / 16%
- New project count down 51%, while the share of 100 MW-plus projects rose 8 points
- Overseas order signings by Chinese firms: 298 GWh, up 83%, with Europe still the largest market and the Middle East, India and Chile growing fastest
Fewer, larger projects. Chen’s framing is that the installation slowdown reflects a return to normal build pace rather than a trend decline in demand, with the industry moving from scale expansion to value reconstruction (价值重构).
Part of the H1 decline is policy digestion. Document 136 (136号文, February 2025) barred mandatory storage as a precondition for renewable project approval, grid connection or dispatch. More than ten provinces adjusted or exited mandatory pairing by end-2025, pushing the market toward independent storage backed by provincial capacity compensation mechanisms. Mandatory pairing has partly survived as an implicit rule: Shaanxi’s 2026 competitive allocation scheme weights preliminary work and grid-access capability at no less than 60% combined, and storage is a core variable in the grid-access score — trade analysis estimates a 10–15 point total scoring gap between projects with and without storage, which in competitive allocation is the difference between placing and not placing.
Demand policy runs the opposite direction
The 15th Five-Year Plan for new power system construction (发改能源〔2026〕942号, NDRC and NEA, 3 August 2026) sets roughly 160 GW of new new-type storage over the plan period, reaching 300 GW nationally by 2030; roughly 80 GW of new grid-side independent storage capable of peak supply assurance, with an equivalent peaking duration target of four hours, reaching 140 GW by 2030; and a 50% non-fossil generation share by 2030.
A state demand guarantee issued five weeks before the state froze supply-side entry.
The cost stack
The consumption tax (财政部/海关总署/税务总局 2026年第20号公告) restored a 2% levy on lithium primary and lithium-ion batteries from 1 September 2026, rising to 4% from 1 September 2027 — the first substantive change since lithium batteries entered the exempt list in February 2015. Structural details that matter:
- Storage systems, as assembled electrical equipment, are not taxable. The burden sits at the cell-to-pack manufacturing stage, not at the storage plant or the EV.
- Sodium-ion, solid-state and fuel cells are exempt from 1 September 2026 through 31 December 2028. Semi-solid-state cells are not exempt. All-vanadium flow follows the lithium schedule.
- Exports operate on levy-then-rebate, so export competitiveness is unaffected.
EVE Energy issued the first pass-through notice on 24 July, adding 2% to domestic supply prices from 1 September and applying it to unbilled orders and undelivered framework contracts. Whether it travels downstream is contested: SMM’s Li Yisha notes cell makers’ net margins sit near 5%, so absorbing 2% halves profitability; Mo Ke counters that project owners cannot absorb it either without returns turning negative.
This sits on top of a broader price reversal. Chinese storage cell prices rose through H1 2026 on domestic supply tightness, upstream material cost increases, and AI-driven competition for power semiconductors and components — dynamics independent of US trade policy. Keep that distinction in any pricing or supply-chain analysis.
AI data centers: the demand thesis, and its domestic gap
AI computing power is the demand story vendors use to justify the expansion regulators just froze. Jiemian names it as a “new rigid demand” (新刚需) alongside geopolitical energy disruption and the Five-Year Plan, noting the industry passed through an actual cell shortage.
Chinese trade reporting is more skeptical than English coverage. Huang Pingting’s ESIE 2026 report for 新能源产业家 asks directly whether AIDC storage is a genuine power backbone for AI or the industry’s “demand anxiety” (需求焦虑症) under saturated conditions elsewhere. Her finding: almost all announcements were products and solutions rather than delivered projects, with limited disclosure of delivered 100 MWh-class systems or stable procurement orders. The main actors domestically remain manufacturers; hyperscale buyers have not arrived at scale.
Her structural explanation is the valuable part. A storage industry source told her China has not produced rigid, scalable AIDC storage demand precisely because the Chinese grid is strong — stable, low transformer expansion pressure, adequate new transmission and distribution capacity for new data centers. China therefore defaults to green-power direct connection (绿电直连) plus integrated source-grid-load-storage rather than behind-the-meter storage, and domestic storage profitability is weak because market trading alone cannot secure stable returns. That, she argues, is the core bottleneck on domestic scale-up.
Her contrast cases are all foreign: Tesla Megapack above 40% share of North American hyperscale data center storage pairings in 2025; Fluence with over 30 GWh of AIDC storage deals in discussion, 80% initiated within roughly six months; Oracle’s 1.4 GW Michigan build for OpenAI paired with 1.383 GW of storage, close to 1:1.
Chinese policy for data center power is real but indirect: the data center green and low-carbon action plan (发改环资〔2024〕970号) makes renewable utilisation a key element of energy conservation review for new projects; the 东数西算 implementation opinions target above 80% green power at national hub node new data centers; the May 2025 green-power direct connection notice created point-to-point supply for single large users at up to 220 kV, with 84 projects approved covering 32.59 GW of renewable capacity as of early 2026. In May 2026 compute networks were folded into a national “six networks” infrastructure category, elevating compute-power coordination (算电协同) to national new-infrastructure status.
Corporate positioning is nonetheless heavy. CATL took a roughly RMB 4.1 billion stake in Zhongheng Electric (中恒电气) for data center HVDC, launched a 300+ Ah sodium cell rated above 15,000 cycles for AIDC applications, and won a 2 GW/4 GWh project at a Guizhou computing center worth over RMB 1.5 billion. Shuangdeng (双登股份) signed in April for a 12 GWh semi-solid-state AIDC cell plant at Xuyi, Jiangsu — semi-solid, note, does not receive the tax exemption. Sungrow released a solid-state transformer as an AIDC power hub in July. Envision and Tencent built a fully green-power-supplied data center at Chifeng, Inner Mongolia, with claimed integrated energy cost reductions above 40%.
What to watch
Both Jiemian sources expect approvals to resume selectively rather than uniformly: frontier technology projects approvable, ordinary capacity not; solid-state and sodium routes more easily approved, conventional liquid lithium expansion strictly controlled.
That maps onto the consumption tax exemptions exactly — the same technologies advantaged by both instruments, with semi-solid excluded from both. Two independent policy tools pointing the same direction is the strongest available evidence that this is industrial policy with a technology preference, not emergency braking.
Near-term tests: whether the year-end review Caixin describes produces a published national early-warning mechanism, and whether any ministry ever confirms the freeze in writing.
Verification status
Tier 1 — publishable without caveat. 发改能源〔2026〕942号 (NDRC site, primary document) and 2026年第20号公告 (MOF/GAC/STA, primary document); CNESA DataLink H1 2026 data as released by Chen Haisheng on 27 August and carried by Xinhua and Shanghai Securities News; CATL’s Zhongheng Electric stake and Shuangdeng’s Xuyi agreement (both exchange announcements); EVE Energy’s 24 July pass-through notice; 发改环资〔2024〕970号.
Tier 2 — attribute, and flag where load-bearing. The approval freeze itself. Every account traces to one Cailianshe report of 6 September sourced to unnamed supply-chain figures. Jiemian is the only outlet that reported independently, confirming with several companies — one senior source pseudonymous — while noting that other supply-chain firms had received no such notice. MIIT did not respond to a written interview request. This is the central claim of the piece and it is not independently confirmed by any government document. Say so in any published version.
Also Tier 2: the CESA and GGII expansion datasets (industry association databases, described as incomplete statistics by their own publishers); Tian Qingjun’s and Zhang Jinhui’s figures (media interviews, quoted secondhand); Huang Pingting’s ESIE report (bylined trade reporting with named companies and one anonymous source, but a WeChat-origin trade account rather than a newsroom).
The Zhongjiang County document is Tier 2 at one remove. Quoted in Jiemian, not retrieved in this pass. It is the only written artifact of the policy in existence and should be pulled directly before being characterised as a primary source.
Unresolved discrepancies.
Caixin’s exclusives are single-sourced behind a paywall. The year-end review timing and the claim that planned new capacity runs about 1.5× China’s total 2025 output could not be confirmed against a second source. The CESA figure — roughly 2,600 GWh of disclosed design capacity in seven months — makes the 1.5× claim arithmetically plausible where it first appeared irreconcilable, but scope differs (power plus storage lithium vs. batteries generally) and it is not confirmation.
Demand forecasts do not reconcile across sources. Zhang Jinhui’s 1,200 GWh of global storage cell shipments in 2026 sits against GGII’s 485 GWh for Chinese storage cell shipments in H1. Different scopes, plausibly compatible, but neither series was independently verified and they should not be presented side by side without reconciling the definitions.
The 10 GWh global AIDC storage shipment figure to end-May 2026 appears in multiple Chinese outlets attributed only to unnamed “industry institutions.” No named source was located. Omit or attribute vaguely.
Tian Qingjun’s capacity figures come from a media interview quoted across multiple outlets; the original interview was not located.
Language and attribution cautions. 鑫椤资讯 romanises as Xinluo, not Xinlun. Available English translations of the Jiemian piece are internally inconsistent on Zhang Jinhui’s pronouns and invert the photovoltaic comparison — avoid pronouns for Zhang and do not quote from translation.
Do not cite. IT之家, 新浪, 凤凰网 and 网易 versions are verbatim reposts of a single aggregation, not independent confirmation. A 大纪元 analysis dated 11 September recycles the same material with an editorial “China will repeat the solar collapse” frame. Zhihu and Toutiao AIDC pieces are stock promotion. Vendor product announcements distributed via wire services are Tier 3 marketing regardless of the wire’s name.