Chinese Utility Tenders Are Splitting ESS Cell Supply by Format Faster Than Producers Can Follow

NARI Group's June 2026 ESS cell procurement divided approximately 4 GWh into two lots: roughly 2 GWh requiring cells of at least 314 Ah, and roughly 2 GWh requiring at least 587 Ah. Both lots demanded Grade A cells and laboratory test reports against GB/T 36276-2023, the Chinese national standard governing safety and performance for grid-storage lithium cells. The 314 Ah lot required the full report at bid submission. The 587 Ah lot could submit an interim one.

Same buyer, same procurement cycle, two different evidentiary thresholds sorted by cell format. A buyer who relaxes its own qualification standard for one lot is telling you it wants that product badly enough to accept less proof, and that the supply base cannot yet furnish the proof it routinely extracts from 314 Ah producers.

What the tenders show

Five major state-owned enterprise procurements from early-to-mid 2026, reviewed against published tender documents and award reports, sort into two structural approaches.

Format-separated lots with relaxed qualification for the larger format. Huaneng's 4 GWh framework split volumes across 314–450 Ah and 450–700 Ah bands, permitting the large-cell lot to defer full test documentation. CEEC's 7 GWh framework separated 5 GWh at 314 Ah+ from 2 GWh at 500 Ah+, specifying approximate cell dimensions for the smaller format but not the larger.

Minimum 314 Ah threshold, no separate large-cell lot. Huadian's 12 GWh system framework and SPIC's 7 GWh procurement both set 314 Ah as the floor. SPIC added a 10 GWh cumulative-supply threshold and a 5 GWh minimum annual capacity requirement, filtering by production scale rather than by format ambition.

Two things the pattern does not support. It does not show 314 Ah being phased out: every reviewed tender continues to procure multi-GWh quantities at that format. And it does not show buyers pricing the new format's risk. No reviewed tender imposes longer warranties, different liability terms, or a discount on 500 Ah+ cells. The format gate operates through specification and qualification paperwork, not through commercial terms.

Pricing from the CEEC tender gives an early read on the premium. Reported candidate prices: RMB 0.340–0.394/Wh for the 314 Ah+ lot, RMB 0.360–0.383/Wh for the 500 Ah+ lot. The ranges overlap, but the 500 Ah+ floor sits RMB 0.02/Wh above the 314 Ah+ floor. EVE and CALB appeared in both categories.

Penetration is rising; the shape of the curve is unmeasured

InfoLink estimated 500 Ah+ cells at approximately 5% of the market in Q1 2026, projecting roughly 20% for the full year. One observation plus one projection cannot establish whether penetration is accelerating, flat-rate, or front-loaded into the first half. Nobody publishes quarterly format-split shipment data for Chinese ESS cells, and the absence is itself worth registering: without it, every 2026 penetration figure circulating in the market traces back to a small number of estimates built on producer conversations rather than shipment reconciliation.

The tender lot ratios are the closest available proxy for how hard individual buyers are pulling. Among the three format-separated procurements, the large-format share ranges from roughly 25% (Huaneng, 1 GWh at 450–700 Ah out of 4 GWh) to 50% (NARI, approximately 2 GWh at 587 Ah+ out of 4 GWh), with CEEC at roughly 29% (2 GWh at 500 Ah+ out of 7 GWh). The spread is more informative than the average. Buyers within the same procurement year and the same broad category of state-owned utility are setting materially different format allocations, which means the pull is real but uncoordinated. Set against InfoLink's 20% full-year projection, the largest buyers are specifying a higher share of the new format than the market as a whole has yet absorbed.

SMM argued that 314 Ah supply had turned structurally tight because manufacturers were converting lines toward 500 Ah+ while customer deliveries remained anchored to the older format. Plausible, and unverifiable from published data. Neither SMM nor the tender documents disclose matched figures for qualified 314 Ah output, line downtime during conversion, or unfilled order books, and SMM's tightness claim and InfoLink's penetration estimate are not measuring compatible objects: one is about availability at a format, the other about share of shipments. If the tightness is real, its cause is producers converting capacity ahead of demand that tender structures have signalled but not yet delivered in volume.

CATL's warning and what it implies for everyone below it

In a March 2026 investor-relations record, CATL said larger ESS cells can improve system energy density and simplify system architecture, then added:

CATL described its 587 Ah cell as the current "sweet spot" and claimed lifetime energy throughput nearly 50% above mainstream 314 Ah products. Its June 2025 product launch claimed a parts-per-billion defect rate and "7.5-sigma" safety control for that cell, which is marketing language rather than an audited production metric.

CATL quantified nothing that would help a buyer size the risk: no yield penalty, no defect-rate comparison against 314 Ah, no ramp duration to stable output on a 587 Ah line. The warning carries weight because of who issued it. As covered in Issue #7, CATL runs a 23.96% ESS-system gross margin, which is room enough to absorb ramp costs that would be existential for a producer working at mid-single-digit margins, and its process-engineering bench is the deepest in the industry. Whether or not CATL has solved the problem on its own lines, a public caution from the best-resourced manufacturer in the sector sets a floor on what less capitalised producers should be assumed to be facing.

Why scaling format is a process-control problem

Three manufacturing difficulties scale with cell format in ways the technical literature documents, though no published study quantifies the yield penalty for a specific 314-to-500 Ah conversion.

Electrolyte wetting. Larger electrode stacks take longer to wet uniformly. A 2025 JES study identifies wetting time as a manufacturing bottleneck in large-format cells. A separate review reports that inadequate wetting can produce localised current concentration and lithium plating. The step cannot simply be run faster without risking incomplete saturation, so a converted line carries a longer cycle time at the same nominal throughput.

Formation and aging. Formation, the controlled first charge-discharge cycling that builds the initial interphase layers, plus post-formation aging, can extend for weeks. A Nature Communications perspective on battery manufacturing quality notes that equipment or process recalibration during these steps cuts throughput and raises cost. Larger cells require recalibrated formation protocols; the recipes do not transfer across formats.

Defect scaling. The same paper draws a distinction worth carrying into supplier conversations: electrode-level defects may occur more frequently per cell as electrode area grows, while cell-level defects such as weld failures can favour larger formats, because a pack of a given size needs fewer individual cells. The authors state that the industry lacks a clear empirical relationship between form factor and quality. A Chinese industry whitepaper argues that larger winding structures make tab deformation, curved-edge stress, and electrolyte distribution harder to control.

The risk is documented and plant-specific. Its magnitude depends on an individual producer's process control, equipment vintage, and engineering depth, which is why producer-level evidence is worth more here than any aggregate penetration forecast.

Producer positioning by filing evidence

China Energy News reported in June 2026 that CATL, EVE Energy, Hithium, and Envision AESC had achieved stable production of 500 Ah+ cells, while mid-tier suppliers were converting existing lines or adding one or two new ones. It published no company-by-company yields or conversion costs.

For procurement purposes the useful sort is not tier but evidence state: a deployed project with operating hours behind it is a different object from an announced line with a target date, which is different again from a launched product with no disclosed production status.

Deployed and independently corroborated. EVE Energy's 2025 annual report describes a 200 MW/400 MWh station using its 628 Ah cells, energised in September 2025. Fujian's provincial industry authority separately identified EVE among companies achieving mass production above 600 Ah. Hithium reported a 500 MW/2 GWh project on 587 Ah cells reaching full grid connection in April 2026. Neither source discloses line yield or conversion cost.

Announced ramp, not yet at volume. Ganfeng Lithium said in June 2026 that its Nanchang 588 Ah line was in final preparation, with ramp-up expected in Q3 and gradual volume production by year-end; existing ESS lines were running near 100% utilisation. Great Power said in January that its 314 Ah lines were fully loaded and that it planned 587 Ah output during 2026, selecting a winding process for yield and cost reasons. In February it approved RMB 1.2 billion for a Zhumadian project combining one 587 Ah line and one 120 Ah line, without disclosing GWh capacities or a stable-yield date.

Product launched, production status unclear. Pylontech has released 588 Ah and 601 Ah products but says batch-production timing depends on line commissioning, customer certification, and demand. Narada Power's 2025 annual report lists 587 Ah as "released and being implemented," with current shipments mainly 314 Ah and 392 Ah. Weight Narada's disclosures accordingly: two independent directors declined to assure the accuracy of that annual report, and the auditor issued a qualified opinion with going-concern language.

None of these filings provide format-specific first-pass yield, scrap rates during conversion, customer-qualification pass rates, or any reconciliation between announced capacity and sustained saleable output. The distance between "we have a 587 Ah product" and "we are shipping qualified 587 Ah cells at stable yield" stays wide for most producers outside the top four, and no public disclosure regime currently closes it.

Conversion economics from a single estimate

The only format-specific line-conversion cost I have located comes from a July 2026 Economic Observer report citing an unnamed production manager at a central-China cell factory: converting a 1 GWh 314 Ah line to 500 Ah+ means replacing or modifying coating, rolling, winding, electrolyte-injection, formation, and grading equipment, at above RMB 50 million per GWh, with nearly two months of shutdown.

One anonymous estimate. The available company capex figures cannot corroborate it, because they bundle too much: Great Power's RMB 1.2 billion covers a mixed two-line project, Ganfeng's approximately RMB 4.07 billion covers a mixed 10 GWh complex. Treat the RMB 50 million figure as provisional, not as an industry benchmark.

Even so, the cost components are worth assembling, because their sum is what determines which producers can afford to follow the tender specifications. At CEEC's reported 314 Ah pricing floor of RMB 0.34/Wh, a 1 GWh line at full utilisation generates roughly RMB 340 million in annual cell revenue. Two months of shutdown removes about RMB 57 million of that while depreciation, lease, and staffing costs continue. The conversion capex adds RMB 50 million or more. Then the ramp: Ganfeng's own disclosed timeline, Q3 start to gradual volume production by year-end, implies at least two quarters of below-nameplate output and elevated scrap. Each quarter at half utilisation costs roughly another RMB 43 million in foregone revenue against a fully loaded line. These are my estimates from partial data and margin structures differ by producer, but the components together point to something well above RMB 150 million per GWh of converted capacity before the line produces stable, qualified output.

That number lands very differently depending on the balance sheet absorbing it. Ganfeng has lithium-mining cash flow behind its battery division and can fund a conversion from a separate revenue base. Great Power approved RMB 1.2 billion for Zhumadian while disclosing that financing arrangements were not yet finalised. For a producer whose entire business is cell manufacturing, the conversion cost may approach or exceed a year's gross profit from the line being converted. Tender qualification thresholds filter on cumulative supply history and annual capacity, which is to say on the past, and capture none of this.

Provincial subsidies do not resolve the format question

The provincial and municipal programmes reviewed in Guangdong, Dongguan, and Yibin, Sichuan offer equipment-investment rewards (20–30% of eligible equipment cost, capped at RMB 15 million per project in Guangdong's case), interest subsidies, certification reimbursement, and industrial-park infrastructure. Dongguan adds R&D support prioritising sodium-ion, solid-state, and silicon anodes.

None of them tie assistance to a cell format. The equipment-reward percentages that applied to a producer's original 314 Ah line apply equally to conversion equipment. At Guangdong's 20% rate, a RMB 50 million conversion could attract up to RMB 10 million, assuming the project clears eligibility thresholds, which reduces the capex line but touches neither the revenue lost during shutdown nor the ramp-period yield loss, the two larger components. Nor do the reviewed programmes impose format-specific clawback, so a producer that took equipment rewards for 314 Ah lines has no contractual reason to stay on that format. The subsidy structure neither traps producers on the old format nor meaningfully accelerates the move off it.

A mid-tier producer deciding whether to convert is therefore making the investment case on its own commercial merits, or on its reading of where tender specifications will sit in eighteen months.

What follows for qualification and timing

In Issue #3 I argued that the Chinese cell price had stopped being one number, because format, bankability, and compliance status each carry their own floor. The format gate adds an axis, and this one is set by demand structure rather than by what technology exists.

Three working conclusions for anyone qualifying Chinese ESS cell suppliers or timing procurement over the next twelve to eighteen months.

On 314 Ah availability, assume gradual narrowing rather than a cliff. SMM's tightness claim is plausible and unverified, and every major 2026 SOE tender reviewed still procures multi-GWh quantities at that format. The exposure is a shrinking qualified-supplier list as producers move engineering attention and line capacity toward the new format, which shows up first as longer lead times and thinner competitive bidding rather than as unavailability.

On 500 Ah+ qualification, demand the evidence the tenders themselves are currently waiving. NARI and Huaneng accepted interim or deferred test documentation for their large-cell lots, which is a reasonable accommodation for a buyer with a portfolio and a poor one for a buyer qualifying a single supplier. Ask for format-specific first-pass yield, scrap during ramp, and operating hours on a deployed project at that format. Outside the top four producers, the public record does not show any of it.

On supplier financial screening, the conversion arithmetic is the more useful filter this year. A producer facing something above RMB 150 million per GWh in combined capex, shutdown, and ramp cost, against a cell price floor of RMB 0.34/Wh, is making a decision that its balance sheet may not support. That is the stratification dynamic described in Issue #7: the format gate, the warranty-accountability filter, and the balance-sheet filter compound, and the buyer writing the specification has no obligation to wait for anyone to clear all three.