The Pass-Through Trap

Lithium carbonate posted its sharpest weekly decline in months; LFP ESS cells registered zero pass-through across both tracked sources, and SMM's 280 Ah format ticked up.

Application Grid

EV cells: No update. BNEF's December 2025 pack benchmark ($99/kWh BEV, $70/kWh stationary storage) remains the last public calibration. Samsung SDI's Q2 2026 earnings call is scheduled July 30; no preliminary release has been published. May yield an updated cell-level ASP.

BESS cells (LFP, China EXW):

USD approximations at ~7.25 RMB/USD. No NMC ESS cell spot exists on either source.

Direction: flat to marginally up. Second derivative: zero on 314 Ah; first positive inflection on 280 Ah.

Defense-relevant formats: No public price data. SMM July 24 lists 50 Ah prismatic LFP at RMB 0.418–0.447/Wh, avg 0.4325/Wh, flat WoW. Closest observable commercial proxy; defense procurement specs and volumes diverge materially.

Pack-level: No update. Last public calibration remains BNEF December 2025.

The Synthesis

The mechanism behind three weeks of disconnection

InfoLink's July 22 assessment places battery-grade lithium carbonate at RMB 141,000/MT average, down 7.2% WoW. SMM's July 24 reading is considerably more conservative: RMB 140,000–151,000/MT, average 145,500/MT, down roughly RMB 1,000 (~0.7% WoW).

The level gap matters less than the velocity gap. InfoLink sees lithium falling ten times faster than SMM does. Assessment timing is the most likely explanation: InfoLink's Tuesday snapshot may have caught a sharper intra-week move that SMM's Thursday range, spanning a wider bid-ask, smoothed out. I cannot confirm this from public methodology disclosures. The divergence is large enough that a reader citing InfoLink's -7.2% and a colleague citing SMM's -0.7% would reach incompatible conclusions about whether lithium is in freefall or merely drifting. Both agree on direction. They disagree materially on speed. For the cell-price question that follows, the divergence matters less than it appears: whether lithium fell 0.7% or 7.2%, cells didn't move.

Three consecutive weeks of that disconnection. In Issue #5, cells softened 0.7% WoW against a lithium decline. In Issue #6, lithium fell approximately 5% and systems held flat. Now lithium's weekly decline has accelerated to its steepest reading of the year per InfoLink, and cells have stopped even the marginal softening observed two weeks ago. SMM's 280 Ah format actually ticked up. The sequence runs: softening, then flat, then flat-to-up, all while the upstream input accelerated downward. Pass-through from lithium carbonate to LFP cells is not lagged. It is structurally impaired.

The 280 Ah uptick requires calibration. RMB 0.0025/Wh on a 0.365/Wh midpoint is 0.7%, within normal assessment variance for a single week. I treat it as directionally significant because it is the first non-negative reading across any tracked ESS cell format in this series, and because it occurred against accelerating lithium declines. The magnitude is within noise. The direction, given the context, is not.

The Fastmarkets CIF CJK lithium carbonate assessment remains behind a paywall. A July 2 pricing notice confirmed the assessment is active; methodology changes take effect September 1. The domestic-to-seaborne spread cannot be calculated from public data. Four consecutive issues without resolution on this gap.

System-level source divergence: 14%, persistent, structural

InfoLink's DC-side system quote (RMB 0.50/Wh) and SMM's (RMB 0.4325/Wh for a 5 MWh cabin) are too far apart for rounding. Almost certainly an object-definition difference: InfoLink's 2-hour liquid-cooled containerized system and SMM's 5 MWh DC-side prefabricated cabin are not identical products. At InfoLink's figures, the cell-to-system spread is RMB 0.13/Wh. At SMM's, RMB 0.065/Wh. Anyone building an integration cost assumption from one source's system quote and the other's cell quote will get a misleading number. Both have held flat for weeks. The divergence is structural, not directional.

Absorption removing buyer leverage

June CABIA data: Chinese battery production up 59.5% YoY, ESS battery sales up 67.5% YoY, exports up 48.7% YoY. The production-to-absorption gap this publication calculates from CABIA monthly figures narrowed through Q2, from approximately 46.0 GWh in April to 41.4 GWh in June (editorial calculation; methodology in the China section). Detailed demand analysis belongs elsewhere. The pricing implication is immediate: sellers are not under pressure to concede. Lithium's decline carries limited procurement leverage when absorption is tightening and export volumes are growing at nearly 50% YoY.

US BESS: still unpriced

No public, decision-grade US BESS equipment $/kWh surfaced this week. EIA's latest battery storage construction cost is $1,361/kW for 2023 installations; 2024 data expected September 2026. Lazard's July 2026 Levelized Cost of Storage analysis ($210–292/MWh for a 100 MW, 4-hour system) is a levelized cost inclusive of financing, degradation, and operations. It cannot be decomposed into a $/kWh equipment figure, but it is directionally useful: Lazard found costs rising, consistent with tariff layering and compliance friction. The standing position from Issue #6 holds: China EXW spot is negotiation gravity, not an executable US price.

Q3 floor reinforcement

Iron phosphate fee increases effective August 1. VAT export rebate already reduced from 9% to 6% since April 1, with full elimination January 1, 2027. GB38031 enforcement effective July 1 adding system-level compliance cost (this publication's prior estimate of 15–20% system cost uplift for compliant configurations; no named third-party estimate has surfaced publicly). Lithium carbonate at RMB 141–145K/MT is already near levels where marginal Chinese producers face cash-cost pressure, limiting further downside absent a demand shock.

The floor is structural. It is set by cost layers that lack real-time tickers, which is precisely why models anchored to the one input that does have a ticker are mispricing the trajectory. The sidebar covers which layers, and why they're sticky.