Two Assessors Agree on the Level; They Disagree on Whether It Moved

LFP ESS cell spot converged at CNY 0.365/Wh across both tracked assessors this week, but the first derivative is not readable: the assessments blend transactions priced before and after the September 1 consumption tax. EV cells unchanged for a third consecutive week. Lithium carbonate rose on the weekly print and reversed on the daily.

Application Grid

EV cells. SMM assessed 100 Ah and 174 Ah prismatic LFP at $52.8/kWh and $53.1/kWh (September 4), zero change. NMC six-series and eight-series prismatic at $82.4/kWh and $95.0/kWh, also flat. All four series have held unchanged for at least three consecutive weekly readings, extending the pattern noted in Issue #11 when lithium was already moving and cell assessments were not.

BESS cells. InfoLink assessed 280 Ah and 314 Ah China LFP ESS cells at CNY 0.340–0.390/Wh, averaging CNY 0.365/Wh, +0.7% WoW (September 2). SMM assessed 314 Ah at CNY 0.365/Wh, flat (September 4). Level agreement, direction disagreement. Synthesis below.

Defense-relevant formats. No public clearing price located for UAV, UUV, or 6T cells. The DIBC cylindrical-cell solicitation remains closed, not awarded.

Pack-level. SMM's 18.3 kWh and 51.9 kWh prismatic LFP EV-pack series terminated after August 28 at $82.0/kWh and $85.9/kWh. No continuous public China EV-pack velocity reading exists for September. BNEF's December 2025 survey ($108/kWh global, $84/kWh China) remains annual calibration only.

BESS Cell Readings

InfoLink and SMM both landed at CNY 0.365/Wh for 314 Ah LFP ESS cells. That convergence deserves scrutiny rather than confidence. Issue #12 documented a standing CNY 5–7/kWh gap between the two, reflecting different transaction populations. The gap may have genuinely compressed. Both assessors may also have centred on a round figure in a week when identifying comparable transactions got harder.

The disagreement on direction is the more useful observation. InfoLink's +0.7% WoW is its second consecutive weekly increase in the ESS cell series, and the first to span the September 1 boundary when the 2% consumption tax on lithium-ion cells took effect. Assessed September 2, one day after, it necessarily mixes deals signed under both regimes. InfoLink's accompanying narrative notes manufacturers circulating notices seeking to pass the tax into new orders, but the published table reports a single tax-inclusive average across nearly 20 domestic cell manufacturers, without separating pre-tax deliveries, repriced undelivered orders, and newly signed post-tax contracts.

That +0.7% therefore cannot be decomposed into tax pass-through, shifting transaction mix as pre-tax deals roll off, and upstream cost pressure from lithium or copper foil. The implementation rules make the decomposition harder still: liability timing turns on whether the sale is on credit, installment, or prepayment terms, so two contracts signed the same week can carry different tax treatment depending on payment structure that neither assessor discloses.

SMM's flat reading is consistent with a more conservative methodology, a different transaction population, or simply the two-day lag (September 4 against September 2). I cannot distinguish among those from what either publishes. The defensible statement for the week: the market is approximately CNY 0.365/Wh, and weekly direction is not measurable through the transition.

The Domestic-Export Comparison Isn't Executable Yet

Eligible exports retain consumption-tax exemption or refund, which makes the domestic-versus-export realization on the same product the cleanest available decomposition. Domestic up roughly 2% with export flat isolates pass-through. Both moving together points upstream.

The test is identifiable but not yet executable. The most recent standalone Chinese LFP ESS cell FOB observation is InfoLink's July average, relayed by SolarQuarter on August 26: $50.3/kWh for 280 Ah, $50.0/kWh for 314 Ah. It predates the tax and carries no destination, order-size, or delivery-cohort detail. SMM publishes FOB assessments for exported 5 MWh battery containers, but that prices a complete system, not a cell. If InfoLink holds its monthly export-price cadence, a post-September FOB cell figure could land by early October. Until it does, the comparison is a framework waiting on data.

Lithium Reversed Mid-Week

InfoLink's September 2 weekly assessment averaged CNY 152,000/MT, +1.7% WoW. SMM's September 4 daily assessment averaged the same CNY 152,000/MT but was down CNY 4,000/MT, or −2.56%, from the prior day, consistent with SMM's expectation of September carbonate supply up roughly 11% MoM. A weekly rise followed by a daily pullback reads as a reversal forming mid-week rather than a trajectory change.

As in Issue #11, the modelled cost impulse from carbonate at these levels is smaller than the methodological gap between the cell assessments themselves. It matters for where cell quotes go in eight to twelve weeks, not for repricing the current tape.

Fastmarkets revised its CIF CJK battery-grade lithium-carbonate specification on September 1: higher minimum transaction size, revised payment and qualification terms. I located no current free Fastmarkets print to bridge the change, so the seaborne-versus-domestic spread this section normally tracks is interrupted until a post-methodology series accumulates enough prints to compare.

Copper Foil Is the Supply-Side Variable Worth Watching

SMM's September 4 assessment put the 6 μm lithium-battery copper-foil processing premium at CNY 23,000/MT, +4.55% WoW. That is the processing fee alone, not full foil cost including the copper input. SMM's September 3 ESS review attributes part of September's cell-production schedule deceleration to foil tightness, with leading foil producers directing new capacity toward higher-margin electronic-grade product. China ESS-cell output growth is forecast to slow from +5.6% MoM in August to +4.09% MoM in September, against foil operating rates already above 92% in July.

This is the one input this week that could support firm ESS cell pricing independent of the tax. Its separate contribution cannot be quantified from what SMM publishes, and I am not going to attach a number to it. But it is the variable I would track through October, because unlike the tax it does not wash out of the series once the cohorts clear.

US Procurement Signals

Illinois approved 600 MW against a 1,038 MW target on September 1, split 520 MW MISO and 80 MW PJM, at an average $110.52/MWh strike under 20-year Indexed Storage Credit contracts. With 87% of awarded capacity in MISO, the split most plausibly reflects interconnection queue position and revenue-stack differences between the two markets rather than any variation in hardware cost. The $110.52/MWh is a revenue-settlement reference against which modelled arbitrage and capacity income are netted. It should not be carried into a hardware model.

Fluence's June 30 10-Q disclosed higher estimated costs on certain projects, attributed partly to battery-price increases, with minimum purchase commitments rising to $2.913 billion from $2.640 billion at March 31. Filing detail sits in Issue #11's sidebar. What it establishes for this section: cost pressure from cell procurement is showing up in US integrator financials at the same time Chinese domestic cell assessments sit near recent lows.

Korean ASP

Q2 disclosures from LG Energy Solution, Samsung SDI, and SK On provide neither a standalone cell ASP nor sufficient application-level volume to derive a reproducible $/kWh. No update this week.

What Broke in the Measurement Infrastructure

Three of the series this section reads changed around September 1. SMM terminated its two tracked pack benchmarks after August 28, leaving no continuous public China EV-pack reading. Fastmarkets revised the CIF CJK specification, interrupting the seaborne spread. And the consumption tax put a mixed-cohort problem into every domestic cell assessment that will take four to six weeks to clear.

For anyone marking a model this week, the levels hold up and the weekly rates of change do not. Wait for the October prints before reading trajectory off any of these series.