InfoLink's public China ESS cell series has printed RMB 0.360/Wh for 314 Ah prismatic LFP since early February. It printed RMB 0.360/Wh again on August 12. Six months flat — approximately $50/kWh at ~7.2 RMB/USD. In the same week, InfoLink's own contributed analysis stated that 314 Ah cell prices had risen more than 15% over six months amid nearly year-long supply tightness.
Both carry InfoLink's name. They describe different objects. The public series is a VAT-inclusive assessment across nearly 20 manufacturers. The 15% figure specifies no starting price, no geography, no VAT basis, no delivery terms, no supplier population.
The public series says the rate of Chinese ESS cell price decline went to zero six months ago and has not resumed. That trajectory change matters more than the level. But the level it reports applies to one segment of the market, and other segments have moved independently.
The Chinese cell market presents a price surface, discontinuous across pools separated by qualification barriers, format requirements, certification gates, tax basis, and delivery terms. What you can buy at what cost depends on who you are, what you have qualified, and which pool you can reach. The spot assessment everyone tracks covers the broadest and least restricted of those pools.
Two axes organize what follows. Who holds surplus production and what finances their ability to carry it. And which pools that surplus can access, with what preventing it from flowing between them.
Who carries the surplus
Chinese battery production exceeded sales by 122.3 GWh in January–July 2026 — 10.5% of reported sales, per CABIA. The comparable January–July 2025 figure was 44.9 GWh, or 5.7%. July 2026 alone: 32.8 GWh. July 2025: 6.6 GWh. The gap has roughly tripled year-on-year in the monthly series while the year-to-date ratio has not quite doubled, which means the divergence has been widening through the year rather than sitting at a higher constant level.
| Period | Production–Sales Gap | % of Reported Sales |
|---|---|---|
| Jan–Jul 2025 | 44.9 GWh | 5.7% |
| Jan–Jul 2026 | 122.3 GWh | 10.5% |
| July 2025 | 6.6 GWh | — |
| July 2026 | 32.8 GWh | — |
The available comparison does not support a seasonal explanation.
One caveat on the data: CABIA does not publish a physical-inventory identity reconciling production, sales, exports, and domestic installations. The 122.3 GWh is reported production minus reported sales. Captive transfers, work-in-progress, reporting timing, and goods committed to later delivery all sit between the two series. Call it unabsorbed reported production. The scale and the direction still tell you something concrete about who is producing ahead of demand and what is paying for it.
Carrying capacity stratifies by financing mechanism.
Cash carry. CATL's RMB 372B cash position (H1 2026) and 94.86% utilization rate demonstrate production intensity, but utilization alone does not establish sell-through. Net finished goods more than doubled from RMB 22.6B to RMB 47.6B over H1 2026, while contract liabilities declined from RMB 49.2B to RMB 36.5B. The filing does not split finished goods by cells, systems, geography, or committed order, so it cannot distinguish planned buildup from unsold stock. What is clear: CATL's carry capacity is enormous, and it is now exercising some of it. The supplier-credit structure — RMB 355.5B in combined trade and notes payables, 96.6% in bank acceptance bills — means input costs are financed by suppliers on approximately 250–310-day settlement cycles. Treat CATL here as one node in the system; the balance-sheet case study is in Issue #10's feature.
Provincial credit and political support. Farasis's 2025 annual filing shows Guangzhou Industrial Investment Holdings and aligned parties holding 16.155% of effective voting rights with board control, making the Guangzhou municipal government the ultimate controller. The filing records RMB 108.55M in government grants recognized in 2025 P&L and RMB 391.70M in deferred government grants at year-end. Farasis reported a RMB 140M net loss and negative RMB 397M operating cash flow in Q1 2026 while its Ganzhou and Guangzhou bases continued ramping.
That evidence establishes continued operation during losses alongside municipal ownership and disclosed grants. It does not establish an employment mandate, and it does not prove that government support caused any specific production decision. Provincial incentive data is structurally opaque in English-language sources; CRU Group's periodic analytical work remains the best available read on these dynamics, and it is proprietary research, not verifiable filing data. What can be said with the filings alone: the carry duration for a municipally controlled producer holding asset-linked deferred grants exceeds what its standalone cash flow would permit.
Buyer deposits and export float. Downstream OEMs and project developers who have qualified a supplier and placed deposits finance a share of production ahead of delivery. Cells produced against a deposit are financially supported even where physical delivery is deferred. The scale is not publicly separable from aggregate balance-sheet items for most producers. Cells destined for overseas projects add a further 4–8 weeks of shipping, customs clearance, and destination-country certification before revenue recognition, with the producer carrying working capital through transit.
So the 122.3 GWh is financed by some mix of CATL-scale reserves, provincial support for loss-making producers, supplier credit on long settlement cycles, buyer prepayments, and export transit float. This publication has tracked the overcapacity structure since Issue #7. The question throughout has been how long each mechanism sustains production ahead of absorption, and what happens when individual mechanisms expire.
Two expired in 2026. Beijing Guoneng Battery's creditors rejected continued operations in February, and the company entered consolidated bankruptcy liquidation in April. Soundon New Energy's restructuring failed and the Xiangtan court declared bankruptcy on August 14. Separately, Xiongtao suspended a planned 5 GWh expansion in April on conservative downstream order forecasts — capacity withheld, not lines closed. These are the first documented 2026 exits, updating Issue #9's finding that no Tier-3 exit had been publicly confirmed through early August.
Neither court record quantifies how much currently qualified GWh left any specific supply pool, and both companies were already in court processes before their terminal events. Carry can expire; that much is now documented. Whether 2026 consolidation has materially reduced available Chinese cell supply is a different claim, and the evidence does not yet support it.
Which price pools the surplus can access
Surplus does not flow into a single market. It enters distinct pools — or fails to enter them — separated by barriers that are mostly about qualification, format, and certification rather than price.
Spot / unqualified. The broadest pool. InfoLink's public China series, covering nearly 20 manufacturers on a VAT-inclusive basis, sits here: RMB 0.330–0.390/Wh, averaging RMB 0.360/Wh, for both 280 Ah and 314 Ah cells as of mid-August. Any producer with functional lines and a willing buyer can transact here. Tier-3 producers running at 30–40% utilization place marginal volume in this pool, and this is where the floor is visible. That floor reflects the variable cost of the least efficient producer still receiving carry support — which, given provincial financing, can sit below what standalone economics would permit.
Domestic EV-qualified. Cells sold to Chinese OEMs for vehicle integration require supplier qualification involving months of testing, sample validation, and production-line audits. Once qualified, switching costs are high; an OEM will not re-qualify a supplier to save RMB 0.01/Wh. CATL and BYD's domestic concentration operates here. Pricing is structurally stickier than spot because the qualification barrier limits competitive entry. Surplus from an unqualified Tier-3 producer cannot reach this pool at any price it offers.
Global ESS-bankable. This is where the divergence from spot pricing plausibly lives. Utility-scale BESS projects financed by institutional lenders require cells from suppliers that have passed independent engineering review, fire-behavior testing (UL 9540A in the US, equivalent regimes elsewhere), deployment-history verification, and financial and warranty diligence. A documented DNV technical bankability evaluation took six months; PNNL's Grid Storage Launchpad notes that custom independent testing alone can run up to six months, separate from subsequent lender and insurer review.
Chinese domestic utility tenders impose their own hard gates. SPIC's 2026 procurement required at least 10 GWh of cumulative cell deliveries over three years and at least one 400 MWh single-project order. NARI's June framework required at least 4 GWh of cumulative 314 Ah supply since January 2023 and at least two contracts of 500 MWh or more. NARI also referenced 587 Ah lots, with a lower documentary bar — interim rather than final GB/T 36276-2023 test reports were accepted for that format.
These gates are format-specific. A producer qualified for 280 Ah does not automatically qualify for 314 Ah. That is the mechanism making the 314 Ah tightness claim coherent alongside flat aggregate spot: the set of producers who can deliver 314 Ah cells into a bankable project with the required deployment history, test reports, and financial standing is a small subset of the set that can produce LFP cells at any specification. As Issue #9's feature documented, Chinese utility tenders are splitting ESS cell supply by format faster than producers can follow.
Export-compliant. Exported cells face a distinct regulatory and tax basis. From September 1, a 2% consumption tax applies to lithium-ion cells and packs produced in China. For eligible direct exports by the manufacturer, the statutory mechanism is an exemption — zero net consumption-tax burden. For trading companies exporting purchased cells, the mechanism is a refund of upstream consumption tax already paid, which carries working-capital and administrative cost between payment and recovery. September 1 therefore adds friction to intermediated exports specifically, which matters because many Western procurement relationships run through trading intermediaries rather than direct manufacturer contracts.
The VAT export rebate continues its phaseout — from 9% to 6% as of April 1, 2026, falling to zero on January 1, 2027. Layered on top: MOFCOM's export-control regime, including a 300 Wh/kg energy-density licensing threshold under Announcement 58, currently suspended through November 10, 2026, and a permanent graphite licensing requirement that remains active — plus destination-specific tariffs. If the Announcement 58 suspension lapses without renewal, high-energy-density cell exports pick up a licensing step that adds lead time and approval uncertainty. Each layer narrows the population of producers and transactions that can access this pool at a given delivered cost.
How carry and pool access interact
Surplus pushes prices down where it can flow freely. Qualification barriers hold prices stable where entry is restricted. Both forces act on the same aggregate capacity at the same time, which produces a price surface that only looks contradictory if you read it as a single market.
Look at who is contributing to the 122.3 GWh gap. CATL at 94.86% utilization is producing at scale, though its finished-goods buildup means not all of that has cleared. BYD's captive integration absorbs its own output. The residual is weighted toward Tier-2 and Tier-3 producers — precisely the producers whose carry is financed by provincial support, supplier credit, and spot-market absorption of marginal volume. That cohort overwhelmingly lacks the deployment history, format-specific test reports, and financial standing to enter the bankable ESS or EV-qualified pools. Their surplus stays in or near spot because spot is the only pool whose barriers they clear. The bankable ESS pool can be genuinely tight at the same time, because aggregate capacity was never the binding constraint there. The qualification filter is.
Flat spot at RMB 0.360/Wh and a claimed 15% rise in 314 Ah pricing measure different pools. Surplus concentrates where qualification barriers are lowest, so a qualified pool can tighten while aggregate capacity remains in massive surplus.
The public China spot series aggregates across qualification tiers. The 15% figure, whatever its precise basis — InfoLink has not disclosed starting price, geography, or VAT treatment — most likely reflects a narrower population of qualified transactions: possibly global, possibly contract-based, possibly carrying delivery-window premiums for bankable supply. The public data does not permit a clean decomposition. When qualification barriers block flow between pools, the qualified pool can rise while the unqualified pool sits flat — and both readings of the data are correct within their respective domains.
Format transitions amplify the effect. The shift from 280 Ah to 314 Ah in ESS tenders creates a temporary qualification bottleneck, and deployment-history requirements (4 GWh cumulative, multiple 500 MWh contracts) cannot be accelerated by capital. During the transition, qualified 314 Ah supply is scarce relative to demand even with aggregate LFP capacity in massive surplus. Issue #9 identified the same overcapacity producing fewer eligible sellers. If 587 Ah gains traction beyond NARI's early tender language, a second bottleneck layers onto the first.
A secondary tension runs through the export-compliant pool. CATL's internationalization depends on expanding access to overseas markets, where its H1 2026 overseas gross margin of 29.97% runs above its 21.16% domestic margin. The regulatory trajectory is narrowing that access. September 1 adds friction for intermediated exports. January 1, 2027 removes the last export tax advantage. MOFCOM's licensing regime constrains high-energy-density shipments if the November suspension lapses. For CATL, with the scale and direct-export infrastructure to absorb those costs, the narrowing is manageable and may be competitively useful — it raises the bar for smaller exporters who rely on trading intermediaries and cannot fund the consumption-tax refund cycle. For Tier-2 producers depending on export volume to supplement weak domestic utilization, each layer of friction cuts the margin available in the export pool, which shortens the carry duration that export revenue can finance.
Durability assessment
The plateau at RMB 0.360/Wh in spot is durable through the next six months absent a lithium carbonate spike that forces input-cost pass-through. Provincial carry support shows no sign of systematic withdrawal; the two documented exits were already in court processes, and Farasis demonstrates continued municipal support through losses. September 1 does not burden direct exports and adds only modest friction domestically. The mechanisms holding the spot floor in place have not run out.
In the bankable ESS pool, pressure runs the other way — upward, or at minimum resistant to the downward pressure visible in spot. Qualification barriers are tightening as format transitions accelerate and deployment-history requirements compound. The pool of producers able to deliver 314 Ah into a bankable utility-scale project with the required track record is growing more slowly than demand for that specific qualified product. Domestic EV-qualified sits between the two: sticky pricing protected by switching costs, concentration still rising around CATL and BYD.
Three triggers would break this structure. First, a regulatory enforcement event forcing Tier-3 shutdowns at scale. GB 38031-2025, the thermal runaway standard effective July 1, 2026, has prompted MIIT inspections and battery sampling, but no enforcement action against a named producer appears in reviewed sources — plausible as a consolidation driver, unconfirmed as a near-term one. Second, a sustained lithium carbonate rebound lasting long enough to exhaust inventory hedges and force real-time procurement at elevated cost. Spot has moved from RMB 140,000/MT to RMB 149,000/MT in August. The direct cell-cost impulse from that move is small — roughly RMB 4.6/kWh on available lithium-intensity estimates, about 1.3% of a RMB 360/kWh cell — but a trajectory toward RMB 180,000/MT or above would compress margins already thin at the tail. Third, a demand shock large enough to absorb the production-sales gap and tighten spot, which requires Chinese EV and ESS installation growth to re-accelerate well beyond current rates.
Until one of those fires, the surface holds its shape: flat where surplus can reach, tightening where qualification restricts entry. The spot floor is real, and it is also the pool carrying the least information relevant to a qualified procurement decision. Anyone pricing a bankable 314 Ah supply agreement off the RMB 0.360/Wh print is reading the wrong instrument.
- CATL finished-goods composition: The H1 2026 filing shows net finished goods more than doubling to RMB 47.6B while goods dispatched barely moved, but the filing does not split by cells, systems, geography, or committed order — BYD's H1 results, scheduled for August 28, may offer a comparative read on whether inventory buildup is leader-wide or CATL-specific.
- GB 38031 private enforcement: MIIT has inspected and sampled vehicles and traction batteries at GAC Aion and Zhaoqing XPeng since July 1, but disclosed no test results, failed suppliers, or penalties — watch for changes in new-model supplier nominations and OEM tender qualification language as proxies for private enforcement ahead of public outcomes.
- ESS concentration reversal risk: InfoLink reports global ESS-cell CR10 declining from 91.2% in H1 2025 to 82.3% in H1 2026, driven by qualified demand moving to mid-tier suppliers, but the durability of those share gains depends on whether the qualification walls hold when 314 Ah supply loosens.
- September tax-basis break: The 2% consumption tax effective September 1 creates a component-boundary ambiguity for integrated ESS exports that official guidance has not resolved — post-September quote comparisons will need explicit tax inclusion, customs classification, and declaration-date matching to remain usable.