LFP ESS cells at RMB 0.360/Wh (InfoLink, August 5), flat week-on-week after a 2.7% decline the prior week. SMM's August 7 assessment: 314 Ah LFP ESS at $0.0479/Wh, down fractionally; 280 Ah flat at $0.0480/Wh. Battery-grade lithium carbonate fell 2.8% to RMB 140,000/MT on the same InfoLink reading. (InfoLink and SMM assess different product specifications under different methodologies, so the gap between their readings is a measurement artifact, not a data conflict. Both point the same direction: cell price decline decelerating toward zero.)
Cell price decline has decelerated to approximately zero while the primary input cost keeps falling. The carbonate trajectory is still moving and the cell trajectory has stalled, which means producers are absorbing the margin improvement rather than passing it through. Two weeks ago, Issue #8 named breadth as the test for whether the late-July RMB 0.010/Wh concession represented a clearing-price shift or selective pricing against volume commitments. It has not broadened. The 280 Ah price is unchanged; the 314 Ah decline is marginal.
Producers holding price while input costs fall is not what aggregate overcapacity is supposed to produce, and the pattern has now persisted long enough to require a structural update to the model set out in Issue #3's "Four Floors, Not One" and Issue #7's "Three Filters, No Exits."
What sets the price a given buyer faces has migrated from aggregate overcapacity toward format-specific qualification, buyer-imposed eligibility gates, and a policy-clock sequence running September through January.
Nameplate overcapacity persists. The volume accessible to a specific buyer at a specific format, qualification standard, and compliance status is a different and considerably tighter number.
Overcapacity: nameplate persists, qualified output thins
Status: Stable in aggregate, tightening in qualified segments. CATL's reported utilization of 94.86% (H1 2026, per its interim earnings) coexists with Tier-3 producers widely estimated at 30–40% (CRU Group structural analysis; the underlying provincial data is not independently verifiable from English-language sources). Nameplate capacity continues to grow. No Tier-3 exit — bankruptcy, permanent shutdown, formal capacity decommissioning — has been publicly documented through August 8. The Tier-2 cohort examined below is itself the evidence: Gotion, Great Power, and Farasis are all still operating and still shipping on economics that range from thin to negative. Provincial employment mandates and local incentive structures sustain that. CRU identifies provincial economic-development logic as the mechanism, with battery plants serving as industrial anchors in regions like Guangxi that are seeking manufacturing diversification, though the fiscal detail behind it stays opaque in English-language sources.
The format transition underway in grid-scale ESS (covered in detail in a companion feature; referenced here at system level only) creates a qualification gate that aggregate capacity figures obscure. Capacity certified for one cell format is not fungible with capacity certified for another until lines are converted, products qualified, and reference lists updated. So nameplate overcapacity and format-specific tightness coexist — a combination the static overcapacity narrative has no way to represent.
Timeline: Format-driven thinning is rolling, not dated. The overcapacity narrative stays directionally correct at the nameplate level for at least four quarters. Its usefulness as a price-prediction tool degrades quarter by quarter as the share of procurement governed by format and qualification gates grows.
Margin bifurcation: four positions
Status: Widening. Issue #7's margin analysis established the gap between CATL's segment margins and the Tier-2 field. H1 2026 profit alerts from Gotion, Great Power, and Farasis fill that in, and they resolve into four distinguishable positions rather than a spectrum.
CATL is margin-secure, cash-rich, and running at 94.86% utilization. BYD's battery economics are not separable from its group filing — Q1 group profit fell 55% YoY — but its 18.49% share of June domestic installations and its vertical integration put it alongside CATL as a price-setter rather than a price-taker. Neither has any reason to cut.
Great Power's H1 alert is the cleanest turnaround in the cohort: RMB 800–866M in attributable profit against a prior-year loss of RMB 88M, with non-recurring items contributing only RMB 8–10M. The company attributed it to increased orders and revenue growth but disclosed neither revenue nor operating cash flow. Against a 2025 base of RMB 11.9B in revenue and RMB 206M in profit, that is a real step up, if the full interim confirms it. This is stabilization at a level that permits continued operation and not much else.
Gotion forecast H1 2026 attributable net profit of RMB 1.20–1.55B, up 227–323% year-on-year. The headline misleads. Non-recurring gains, principally investment income and fair-value changes on equity holdings, account for RMB 1.1–1.4B of that. Core operating profit comes to RMB 85–120M on a business that generated RMB 45B in 2025 revenue. The improvement is real and it is also thin enough that one quarter of equity-market weakness would erase it from the headline. The operating business is fragile and the portfolio income is covering for it.
Farasis reported Q1 revenue down 28% year-on-year, a net loss of RMB 140M, and operating cash flow of negative RMB 397M against negative RMB 126M a year earlier. The company described a "market-development period" involving new-customer cultivation. Cash is leaving faster than it was.
For price, the relevant point is that the two ends of this distribution are both content at current levels for entirely unrelated reasons. The leaders have margin room and no competitive need to cut. The distressed producers keep spot supply flowing because accepting orders below variable cost is the rational choice when the alternative is shutdown against provincial expectations. The middle determines whether the floor eventually moves. If Great Power's stabilization proves durable, it validates the current price as operationally sustainable for an efficient Tier-2 producer. If Gotion's non-operating income dries up and Farasis's burn accelerates, the distressed tail shortens — fewer producers willing to sell at any price — and the floor could rise rather than fall.
Timeline: Full H1 interims through August and September will give matched revenue, margin, and cash-flow data. Direction is visible now. Magnitude is not.
Input cost pass-through: absorbed
Status: Decoupled. Lithium carbonate at RMB 140,000/MT is down 2.8% week-on-week while cell prices are flat. The connection between them runs through cathode procurement: cell producers buy LFP or NMC cathode on contracts that reflect carbonate prices with a lag of roughly four to six weeks, mediated by cathode producers' own inventory cycles. When carbonate falls, cathode cost follows on that lag and cell producers' input costs decline. Whether the reduction reaches cell buyers depends on competitive pressure and margin position, and right now it is not reaching them.
That is consistent with a market recovering from the compression of 2024–2025. Producers who spent two years watching margin disappear are rebuilding it as input costs fall rather than competing the savings away. The absorption holds as long as demand, ESS procurement in particular, stays strong enough that nobody needs to cut price to fill lines.
Pass-through restarts if lithium carbonate sits below RMB 120,000/MT for more than a quarter. At that level Tier-2 variable costs compress enough that even margin-rebuilding producers face pressure to share the savings. Above it, absorption continues.
Timeline: Carbonate is the leading indicator. At the current rate of decline, RMB 120,000/MT is approximately two quarters away, but carbonate is volatile enough that the estimate carries wide uncertainty.
Regulatory enforcement: sampling, not yet consequences
Status: Advancing slowly. GB 38031-2025 is effective. MIIT's May inspection notice authorized sampling of power-battery packs with a one-month testing window. A July 17 follow-up described supplier self-review and sampling activity. GAC Aion and XPeng vehicles were sampled around July 24, so the testing window had not elapsed by August 8.
Nothing identified through August 8 — MIIT notice, company filing, trade-press report — discloses a battery product failing GB 38031-2025, a supplier dropped from an OEM qualification list, or a production line suspended. The MIIT automotive publication index carries no results bulletin for sampled batteries.
The standard's price effect depends on which channel activates. Enforcement that produces failures removes affected products from the MIIT vehicle catalogue, reducing qualified supply and supporting pricing for compliant producers. Enforcement that stays at sampling-without-consequences leaves the standard functioning as a qualification cost: testing, documentation, possible line modifications. Only the cost channel is currently observable.
Timeline: First results could surface in late August given the one-month window from late-July sampling. Whether they will be published proactively or only surface through catalogue changes is unknown.
ESS absorption: volume real, economics unproven
Status: Strengthening on volume, contingent on revenue. CABIA reported H1 ESS battery sales of 318.1 GWh (via CnEVPost); SMM reported 437.8 GWh of H1 ESS cell production. The figures cover different populations and measure different flows — sales against production — but both confirm ESS as a substantial absorption channel. The CEC's H1 operating data shows grid-side independent storage averaging 1,129 operating hours in H1, up 169 hours year-on-year. Q1 CEC data put daily equivalent cycles at 0.67, with independent storage at 3.28 daily utilization hours against 2.25 for renewable-co-located.
ESS demand absorbs cell production that would otherwise compete inside the EV supply chain, which supports pricing in both segments. The durability question is whether storage operators earn revenue sufficient to keep procuring independent of installation mandates. The January 2026 national capacity-price mechanism directed provinces to establish capacity payments for grid-side independent storage. Provincial implementation has reached the formula stage: Liaoning's May notice ties compensation to measurable operating performance, and a Qinghai project list identified one 100 MW station with only 22.4 MW of recognized reliable capacity at RMB 185/kW-year. No public record located through August 8 documents an operator actually receiving cash under the new framework.
That Qinghai derating is the detail worth carrying forward. Recognizing 22.4 MW against 100 MW nameplate means the mechanism is paying for performance rather than installation, which is good for durability and bad for the size of the number. Revenue available to operators will be materially smaller than nameplate-based estimates imply.
Timeline: Provincial disbursement evidence is the signal. H2 2026 provincial budget and audit disclosures may surface the first cash-receipt data.
The September–November–January sequence
Three policy events land within five months, the first of them 24 days out.
September 1: 2% consumption tax on lithium-ion batteries. EVE Energy's July 24 customer notice, reported by Futures Daily, applied the 2% pass-through to undelivered framework orders. That is the first documented producer attempt to push the tax forward. Whether buyers accepted it is undisclosed. For exports, EVE described a "levy first, refund later" treatment, so the tax hits working capital even where it is ultimately rebated.
November 10: MOFCOM export-control suspension expires. The suspension covers battery manufacturing equipment and technology — winding machines, lamination machines, liquid injection machines — plus cells at or above 300 Wh/kg, a threshold well above typical ESS cell energy densities. The cell-price effect is indirect. If the suspension lapses without renewal, Chinese equipment exports to overseas cell plants face licensing requirements, which slows non-Chinese capacity buildout and extends the window during which Chinese cell exports face less locally manufactured competition.
January 1, 2027: VAT export rebate eliminated. The current 6% rebate drops to zero. For exported cells, that 6-percentage-point elimination is the durable swing. The September consumption tax is a working-capital cost, levied at the factory and rebated after export declaration, not an equivalent permanent burden. For domestic sales, both the 2% tax and the lost rebate apply, which widens the gap between domestic and export economics through Q4 before January reverses the export advantage.
Between September and December, export economics temporarily improve: the consumption tax is levied but rebatable on export, and the 6% VAT rebate still applies. January is where the arithmetic turns against exports. Producers and traders with inventory flexibility will concentrate export shipments into Q4 — straightforward tax arbitrage. If Q4 monthly export volumes materially exceed the H1 average of approximately 30 GWh per month (181.3 GWh over six months, per CABIA), pull-forward is the leading explanation and should not be read as demand.
Where the forces oppose
Qualification thinning against nameplate persistence. Format transition and regulatory enforcement thin qualified supply; provincial incentives maintain nameplate. A buyer indifferent to format and unconstrained by compliance requirements faces the full weight of aggregate overcapacity. A buyer specifying a current-generation format with GB 38031 compliance and export-eligible origin faces a meaningfully smaller supply set, and pays accordingly. Qualification thinning is the stronger force for a growing share of procurement, because the population of format- and compliance-indifferent buyers is shrinking as ESS tenders tighten specifications and OEM requirements stiffen under GB 38031.
ESS volume against ESS economics. Installation-driven demand supports cell pricing now; revenue-driven demand would support it durably. The distance between them is the capacity-payment implementation gap, where formulas exist and cash has not been documented moving. If H2 disbursement evidence surfaces, ESS absorption upgrades from contingent to structural. Volume is winning at present: procurement continues regardless of operator economics, sustained by installation mandates and grid-planning targets. That makes the support policy-dependent, and policy dependencies are the ones that reverse without warning.
Export margin incentive against export policy tightening. CATL's overseas gross margin of 29.97% (H1 2026 interim) gives it strong reason to grow export volume, and the January rebate elimination cuts against that. For CATL the overseas margin is wide enough to absorb the loss, so the incentive side wins. For Tier-2 exporters carrying thinner margins, January may close the export channel entirely, concentrating export volume further among the leaders.
Durability assessment
The LFP ESS cell floor at RMB 0.360/Wh is durable for the next two quarters. But "the floor" is now an incomplete description of what a buyer actually faces, because the accessible price varies by format specification, qualification status, compliance requirement, and delivery timing relative to the September–January sequence. There are several floors, and which one applies depends on what the buyer is asking for.
For a domestic Chinese buyer specifying current-generation formats with standard qualification, the floor is firm. CATL is under no margin pressure to cut. Tier-2 producers are stabilizing (Great Power) or surviving on non-operating income (Gotion) rather than liquidating inventory into the market.
For an export buyer, the effective floor rises through H2: working-capital cost from the September consumption tax, then a permanent 6-percentage-point burden in January. The Q4 window, after the tax but before the rebate elimination, will generate export pricing that looks unusually competitive and should not be read as a durable level.
What breaks it: lithium carbonate below RMB 120,000/MT sustained for more than a quarter, compressing Tier-2 variable costs enough to restart aggressive competition. A GB 38031 enforcement action excluding a major Tier-2 producer's product line would also move the floor, though upward — removing supply. And ESS installation mandates weakening before capacity payments establish an independent revenue case would redirect cell supply into the EV market. The nearest catalysts are the first GB 38031 test results, expected late August or September; the September 1 consumption tax; and provincial capacity-payment disbursement evidence, if H2 reporting surfaces any.
- Jianxiawo mine output gap: CATL secured a work-safety permit for the Jianxiawo lithium mine on June 29, but InfoLink reported on August 5 that actual post-restart output has fallen short of expectations, keeping its near-term carbonate range at RMB 130,000–150,000/MT rather than treating the permit as restored supply.
- CEC unplanned outage rate: The CEC's H1 operating report recorded 699 unplanned outages averaging 47.33 hours each, with PCS-related events at 30.62% of incidents — a reliability signal that feeds directly into whether capacity-payment eligibility holds for the installed fleet.
- REPT Battero's interim: REPT's preliminary H1 profit alert showed RMB 700–850M in net profit versus a RMB 63M loss a year earlier, with full results due by August 31 — a fifth data point for the Tier-2 margin map once audited figures arrive.
- Aion recall resolution: GAC Aion acknowledged field failures in CALB 177 Ah-equipped vehicles and began the recall-application process, but no formal filing, supplier rerouting, or connection to GB 38031 testing has surfaced — the first case where a named cell product, a named OEM, and a regulatory process intersect.