How Anti-Dumping Duties on Lithium Batteries Vary by Country

Commerce classifies China as a non-market economy under 19 USC section 1677b(c) and constructs normal value by pricing every factor of production using data from a surrogate country. The surrogate might be Indonesia, Thailand, Colombia, or South Africa for a given review period. Petitioner briefs for the surrogate with the highest costs; respondent briefs for the lowest. Commerce's Office of Policy evaluates against criteria that sound objective but leave wide discretionary space.

For lithium battery cells the valuation exercise is far more contentious than for steel or commodity chemicals. Battery inputs are specialized, differentiated by grade, and priced on thin or opaque markets.

Take electrolyte. Battery electrolyte is LiPF6 dissolved in a carbonate solvent blend with proprietary additives. Public pricing is scarce. If the surrogate country does not import finished electrolyte in meaningful volume, Commerce has to construct a value from the component chemicals. But the salt-to-solvent ratio varies by formulation, and Commerce cannot know the respondent's recipe without accepting the respondent's own data, which Commerce is simultaneously trying to replace with surrogate values. Commerce resolves this by accepting consumption ratios from the respondent while substituting surrogate prices for each component. Assumptions that cannot be verified get embedded in the result.

Separator creates a different kind of ambiguity. Bare PE separator and ceramic-coated separator differ in cost by a factor of three or four, but the surrogate country's import statistics under the relevant HS subheading may lump both types together. Commerce uses whatever data exists. Whether the resulting value is accurate for any individual respondent, nobody can say.

There are twenty or thirty more inputs where similar ambiguities arise. Graphite grade aggregation. Lithium hydroxide purity mixing. Copper foil gauge blending. Each ambiguous surrogate value introduces noise into the constructed cost. Over the full set of input valuations, the cumulative impact of these micro-decisions moves the margin by somewhere north of 8%, hard to pin down because Commerce does not publish sensitivity analyses. Electricity makes it worse. Commerce takes the respondent's total electricity consumption per cell, values it at the surrogate country's published industrial tariff, done. If the published tariff does not reflect volume discounts or demand charges that a real industrial consumer would negotiate, the constructed cost is off. Commerce does not audit electricity pricing in the surrogate country. For battery cell manufacturing, where electricity is consumed across formation cycling, dry room dehumidification (climate-dependent: a Guangdong factory runs dehumidifiers far harder than one in Inner Mongolia), coating ovens running at 120-150C for NMP evaporation, calendering, vacuum filling, and aging chamber climate control, the total per-cell consumption is substantial and facility-specific. Commerce collapses it all into one reported figure times one tariff number.

All of that imprecision becomes the dumping margin. In the U.S., the dumping margin becomes the duty, because Commerce calculates a number and that number is the rate. There is no further step asking whether a lower rate would be enough to protect domestic producers.

The EU caps duties at the "injury elimination level" under Article 7(2) of Regulation 2016/1036 whenever that level falls below the dumping margin. Even when the Commission's own cost construction (using the post-2017 "significant distortions" methodology, mechanically similar to the U.S. surrogate approach) produces a high normal value, the final duty gets capped. The constructed cost often ends up irrelevant to the final EU rate because the injury cap binds below it, while in Washington the constructed cost is the rate.

That single difference, presence or absence of a cap, explains more of the transatlantic rate gap than any specific disagreement about how to value any individual input.

The cap itself depends on the "target profit" the Commission assigns to EU producers. For European battery cell manufacturing this is a difficult judgment. Northvolt has not posted a profitable year. ACC, the Stellantis/Mercedes/TotalEnergies venture, is still commissioning lines. SVOLT's European operations are ramping. No historical period of normal profitability exists for this industry in Europe. The Commission estimates using analogous sectors or business plan projections. Whether it picks 4% or 9% on turnover moves the cap by several percentage points.

The Union interest test under Article 21 pushes in the same direction. BMW, Volkswagen, Stellantis, Mercedes have all submitted briefings arguing that high cell duties would raise EV production costs. The Commission has never formally rejected duties on Union interest alone, but DG Trade staff know that DG GROW and DG CLIMA may disagree on whether high battery duties serve European interests, and that tension affects how aggressively the Commission constructs normal value and where it sets the target profit. None of this shows up in the published regulation.

India has its own cap through the "non-injurious price," which pegs normative profit at 22% pre-tax return on capital employed. That comes from Indian case law. It has nothing to do with actual battery industry margins, which are single digits globally in a normal year. India also has a standing problem: almost no domestic cell manufacturing. Pack assemblers can petition, but the competitive dynamics are different. Chinese respondents frequently skip Indian proceedings because the market is small relative to compliance costs.

Surrogate methodology sets the price Commerce assigns to each input. But the quantity of each input consumed per cell comes from the respondent's questionnaire. Price times quantity equals constructed cost. The respondent's legal strategy on the quantity side, especially yield reporting, shapes the margin as much as the surrogate fights do on the price side.

Commerce's NME questionnaire demands physical consumption of every factor per unit of output, adjusted for scrap and rework. For cells, the critical variable is cumulative yield: what share of started cells emerge as shippable A-grade product?

Yield losses accumulate across the entire production sequence. Coated electrode gets scrapped at QC. Cells fail electrical testing after formation cycling or develop self-discharge problems during aging. Cumulative yield on a mature line is around 91%; on a newer line still being optimized, 82% or lower.

At 82%, the factory uses roughly 22% more of every input per good cell than at 91%.

That overconsumption, multiplied by surrogate values across all inputs, inflates the margin directly. And within what the production records support, experienced respondent's counsel finds real methodological room in how yield gets calculated. Do cells that fail initial formation but pass a second cycle count? If yes, yield goes up and per-unit consumption drops. Does recycled electrode slitting scrap generate a byproduct credit against copper foil consumption? If reported, net foil usage per cell falls. Are B-grade cells sold at discount to secondary markets counted as yield or treated as a separate product?

Commerce's regulations do not answer any of this. Respondents with counsel from the D.C. trade bar who have run dozens of NME electronics and battery cases make these methodological choices deliberately, document each one, and prepare to defend them at verification. Respondents without that experience hand data to an in-house accountant who uses whatever allocation seems intuitive. The resulting yield may be lower than a careful methodology would produce, and the margin comes out higher. Same factory, same data, different counsel, different rate.

Commerce sends two or three analysts plus a supervisor for five to seven days of verification. They pick transactions at random, trace each from purchase order through production records to payment receipt. Every figure must tie to a dated source document. They walk the production floor, check MES screens, pull electrode coating line logs, compare formation equipment records against reported electricity consumption, review QC databases.

A failure mode specific to battery manufacturing that comes up at verification: the gap between electrode coating yield and cell-level yield. A company reports 96% coating yield and 93% cell assembly/formation yield. Combined line yield should be around 89.3%. But if the coating QC system allows rework through a second calendering pass and reworked electrodes are not tracked separately in the MES, verifiers question whether 96% is a first-pass rate or includes unreported rework. If the company cannot demonstrate reworked electrodes meet first-pass specifications, Commerce may find the yield overstated. From there the entire cost submission becomes suspect.

The D.C. firms that represent major Chinese battery companies run multi-day mock verifications before the real one. Associates who have attended dozens of actual verifications play Commerce analysts, pulling random transactions from the sales database, demanding source documents in real time. Company personnel get coached on answering precisely without opening new investigative threads. Personnel who give discursive answers get reassigned or paired with interpreters trained to keep things tight.

A mid-sized Zhejiang manufacturer in its first proceeding does not have any of this. Its Beijing counsel may not know Commerce's formatting conventions or the tendencies of the analyst assigned to the case. The company's accountants may not understand that Commerce requires electricity in kWh per unit rather than in RMB, which embeds the Chinese rate Commerce is trying to discard. A formatting error that could be fixed in substance but goes uncaught before verification triggers a reliability finding. Entire cost submission rejected. AFA applied.

AFA rates in lithium battery cases have exceeded 100%.

That means the duty exceeds the merchandise value. And it applies to the vast majority of Chinese exporters because Commerce individually examines only three to five respondents per investigation. Everyone else gets either the all-others rate or, if they did not cooperate, AFA. The rate distribution is bimodal: a handful at 12% to 25%, everyone else at 80% to 110%+. That cliff between the two groups is driven by legal preparation, not by how aggressively the companies price their exports.

EU verification works differently. Commission teams flag discrepancies and allow on-site correction rather than treating each one as grounds for rejection. A documentation gap that triggers AFA in Washington gets resolved with a supplementary submission in Brussels.

Indian verification, when it happens, is shorter and higher-level. The rigor gap across jurisdictions feeds directly into the spread between best-case and worst-case margins.

A caveat that complicates the surrogate-versus-questionnaire framing. At the system level, surrogates generate the base rate gap between the U.S. and other jurisdictions. But for any individual company, the questionnaire and verification phase determines whether it lands in the 15% pool or the 100%+ pool. A well-prepared respondent can achieve mid-teens margins under the surrogate framework by controlling the quantity side and briefing aggressively on values. A poorly prepared respondent with the same factory gets AFA. The spread between 15% and 110%+ comes from the adversarial layer sitting on top of the surrogates.

Lithium carbonate on the Shanghai Metals Market hit $81,200/tonne on November 15, 2022. It had been at $8,500 in January 2021. By June 2024 it was below $14,000.

The POI in a U.S. case covers the four most recent calendar quarters before petition filing. Petitioner's counsel picks the filing date.

Filing during the November 2022 peak captures a POI where surrogate values for lithium hydroxide are at their highest. Export prices on existing contracts still reflect lower costs from when those contracts were signed months earlier. Constructed normal value soars while the export price lags behind. Filing nine months later captures falling costs and elevated export prices. The margin compresses.

D.C. trade bar counsel times filings around commodity cycles. Standard practice, openly discussed, legal. During 2021-2024 this timing variable dwarfed every other factor affecting lithium battery margins. A fourfold raw material price swing landing inside or outside the POI window moves the margin by 20 or 30 percentage points when the affected input is 40% to 50% of total cell cost.

The EU's POI ends around the initiation date. India sometimes uses longer periods. A three-month shift during 2021-2024 changed the cathode cost picture by a quarter or more. Different jurisdictions investigating the same cells during overlapping periods captured different commodity slices. Margin divergences reflected when each authority looked, not how it calculated.

Afterward, Commerce recalculates annually through administrative reviews. If the original investigation caught peak prices and yielded a 45% margin, the first review might catch collapsed prices and produce 8%. The rate bounces with the commodity cycle over the life of the order. A sourcing commitment made at 45% in 2023 faces 10% by 2026, or the reverse if lithium spikes again. Planning a battery supply chain against a duty that moves with quarterly commodity indices is unlike planning around steel or aluminum duties, where raw material pricing is less volatile.

EU reviews are dampened because the lesser duty rule means commodity swings affect both the margin and the injury cap, which move at different speeds.

Anti-dumping orders on lithium batteries from China will be renewed indefinitely at sunset. Chinese cell manufacturing capacity exceeds demand by a widening margin, with annual additions of 500-700 GWh against slower demand growth. Any forward-looking injury recurrence test will find a basis for continuation. EU expiry reviews at least allow rate recalibration through the lesser duty rule, so the continued order might carry a lower rate. U.S. orders persist at whatever rate the most recent administrative review calculated, which could be 8% or 45% depending on which slice of the commodity cycle the review period captured.

After the Appellate Body killed zeroing (US-Softwood Lumber V, US-Zeroing (EC), US-Zeroing (Japan), 2004-2007), Commerce developed the differential pricing analysis. Cohen's d plus a ratio test check whether export prices show "patterns" across purchasers, regions, or time periods. When found, Commerce switches from weighted-average-to-weighted-average comparison to average-to-transaction, where low-priced sales generate margins individually and high-priced sales produce no offset. The mathematical effect replicates zeroing through a different door.

For lithium batteries this fires almost every time because selling prismatic LFP storage cells at $0.07/Wh and cylindrical NMC automotive cells at $0.13/Wh produces exactly the "pattern" Cohen's d detects. That is not targeted pricing. That is selling chemically different products under a single product scope that ITC drew broadly in the injury phase. The margin inflation runs 5 to 15 percentage points over the standard method.

ITC's broad scope decision and Commerce's differential pricing test are administered by different agencies with different mandates, and neither accounts for the distortion the other creates. Narrower product scope would produce less price dispersion and trigger the test less often. Broader scope makes it fire routinely. The injury-phase scope decision ends up predetermining the margin methodology, which is a consequence neither agency's regulations contemplate.

EU, Indian, Australian, and Brazilian authorities all compare weighted averages without isolating low-priced transactions. Differential pricing is a purely American administrative creation, invisible in the statute, created through rulemaking, receiving remarkably little academic attention relative to its impact on rates.

Commerce runs ADD and CVD in parallel on the same merchandise. NME surrogate methodology already removes Chinese subsidies by substituting market-based costs. A parallel CVD then countervails those same subsidies a second time. Congress created a partial offset after the GPX International Tire Corp. litigation in 2012, but the offset is limited to overlap between specific identified subsidies and specific surrogate values. Subsidies affecting costs indirectly or embedded in surrogates without clean one-to-one mapping remain double-counted.

For lithium batteries, Chinese subsidies include discounted land for gigafactory sites, preferential electricity rates, state bank loans below market, and R&D grants. The CVD rate alone can be substantial, stacking on an ADD rate that already purged the same distortions through surrogates.

In the EU, the Commission adjusts normal value when parallel CVDs exist. India mostly avoids the issue by not running parallel proceedings. The double-counting therefore inflates the combined U.S. rate by an amount that cannot be precisely quantified without Commerce's internal working files, but the structural direction is clear: higher than it would be under a system that used either surrogates or CVDs but not both.

Chinese manufacturers ship cells to Vietnam, Thailand, Hungary, Morocco for pack assembly. If assembly adds enough value for substantial transformation, the finished pack owes no duty on Chinese cells. Commerce challenges this under 19 USC section 1677j, trending aggressive, but "sufficient value" is case-specific. BMS components designed in Shenzhen, assembled on Chinese-owned SMT lines in Hanoi with Chinese-sourced semiconductors: does that qualify as Vietnamese value? Commerce has not issued a general answer. Every importer structuring a Southeast Asian supply chain is betting on where Commerce draws the line, with exposure to retroactive liability.

Under Article 13, the EU's anti-circumvention investigation takes 9 to 12 months, during which circumvention flows freely. India's tools under Section 9A(1A) barely get used. Southeast Asian countries stay out entirely.

Scope classification creates additional leakage. An integrated energy storage system with inverter, fire suppression, thermal management, and software could be classified under HS 8507, 8504, or 8543. If the order covers "lithium-ion cells and modules," the integrated system might fall outside. Commerce has a scope ruling process under 19 CFR 351.225. Most other jurisdictions do not, leaving classification to customs officers at the port. Chinese product engineers design systems with scope language in mind.

New shipper reviews add another dimension. Anti-dumping orders let exporters that did not ship during the original POI request individual examination. Chinese battery capacity expanded so fast since 2020 that entire new companies with multi-GWh factories came online after the investigation period. Commerce scrutinizes whether "new" entities are independent from covered companies by examining ownership, management, facility sharing, procurement. Chinese corporate structures involving provincial government investment funds, shared industrial parks, and technology licensing make this hard. Counsel structures new entities to maximize distance on every factor Commerce examines.

The gap between AFA (100%+) and a new shipper rate (12% to 20%) means tens of millions of dollars annually. Some "new" entities are authentic entrants. Some are restructurings. Chinese capacity is expanding so fast that the landscape of producers changes faster than a five-year order can capture.

The absence of a lesser duty rule from U.S. law does more to explain the transatlantic rate divergence than everything else combined. Calculate identical dumping margins in Washington and Brussels. The EU rate still comes out 20 points lower because the injury cap binds. Add a cap to U.S. law and American rates drop. Remove it from the EU and European rates climb. India, Australia, and Brazil all have their own versions of the cap.

NME surrogate methodology amplifies the gap because its imprecision matters more with no cap to absorb it. Differential pricing inflates U.S. margins by 5 to 15 points through a mechanism absent from every other jurisdiction. ADD/CVD double-counting adds to the combined U.S. rate. AFA severity produces a bimodal distribution with a cliff between cooperating and non-cooperating respondents that does not exist in EU or Indian proceedings.

Then there is the question of how much of the published rate gets collected. Circumvention enforcement, scope disputes at the border, new shipper outcomes, sunset continuation decisions: these set the effective rate, and they vary by jurisdiction in ways harder to quantify but commercially as consequential as the headline numbers.

The gap persists because the legal frameworks producing it are not changing. Article 9.1 of the WTO Anti-Dumping Agreement makes the lesser duty rule optional. No international standard governs surrogate methodology. No convergence mechanism is under discussion. The same structures will be producing the same kind of divergence on whatever battery chemistry is being traded a decade from now.