China's capacity-approval mechanism uses utilization rates, among other metrics, to gate which battery producers can build new lines. A companion piece covers how that mechanism works. This piece reads H1 2026 filings from CATL, Gotion High-Tech, and REPT Battero against the gate's implicit financial requirements.
Sustaining high utilization in a softening market is a balance-sheet act. A producer that keeps lines running while finished goods accumulate needs working capital to absorb the inventory, a payables structure to fund it, and enough margin to avoid bleeding cash while doing so. Who can finance that, and for how long? Three producers at different scales give different answers.
The regulatory news broke over the US long weekend. Caixin reported on September 10 that authorities had surveyed both power-battery and ESS capacity and utilization; SCMP described the intervention as ESS-focused. As of September 7, ESS News reported no implementing instrument on government portals. The scoring criteria remain unpublished. What follows reads the filings against what we know of the gate's requirements, with the ambiguity left visible.
CATL: Running full and absorbing the surplus
CATL reported 94.86% utilization in H1 2026. As I covered in Issue #11, that rate coincided with finished goods rising 110.7% and contract liabilities, which are advance payments from customers and a rough proxy for forward order coverage, falling 25.9%. A producer running near nameplate while inventory accumulates and forward orders thin is producing ahead of demand. The filing shows how CATL pays for it.
Operating cash flow was RMB 60.2 billion. Capital spending on fixed assets, intangible assets, and other non-current assets came to RMB 25.1 billion, the closest capex proxy the filing supports, though it includes intangibles. The difference of roughly RMB 35.1 billion is my estimate of free cash generation; CATL does not report a free-cash-flow line.
The working-capital mechanics are the instructive part. CATL's cash-flow reconciliation shows inventories absorbing RMB 37.9 billion and operating receivables absorbing another RMB 33.0 billion. Both are substantial uses of cash, and both were offset almost exactly by a RMB 71.7 billion increase in operating payables. The inventory buildup that makes 94.86% utilization possible is funded substantially through extended supplier settlement terms. As I documented in Issue #10, CATL carried RMB 355.5 billion of combined trade and notes payables at mid-year, including RMB 57.8 billion in supplier-finance arrangements. Its suppliers are extending the working-capital credit that lets CATL keep producing cells it has not yet sold.
Whether CATL is compressing margins to hold that throughput, pricing down to move volume or absorbing higher unit costs from slack downstream demand, is the other half of the sustainability question. The filing offers only an indirect signal. Operating cash flow grew 2.61% year over year, from RMB 58.7 billion to RMB 60.2 billion, while the working-capital cycle absorbed and released comparable amounts through the inventory-payables offset. If margins were deteriorating materially, operating cash generation should have declined even with that offset holding. My reading is that CATL absorbed the inventory buildup without visible margin erosion, which extends the viable duration of this behavior well beyond what a producer discounting to chase utilization could sustain. That is an inference from a single aggregate, not a disclosed gross-margin series, and it should be weighted accordingly.
Cash and equivalents stood at RMB 340.6 billion at June 30, up RMB 40.7 billion over the half. That increase included RMB 24.3 billion of net financing inflows from the H-share placement and bond issuance. The RMB 440.2 billion liquidity figure appearing in some coverage includes trading financial assets alongside cash; the narrower filing-defined balance is RMB 340.6 billion.
Can CATL sustain this indefinitely? The filing does not support a finite answer. Estimating a duration would require assumptions about future inventory accumulation, payable growth, collection patterns, and construction spending, none of which CATL forecasts. What the filing does establish is that CATL generated positive operating cash flow, at apparently stable margins, during the same half in which inventory and receivables absorbed over RMB 70 billion of working capital, because payables grew by a comparable amount. Any competitor attempting to match that utilization score is measured against that financial structure, not just against the percentage.
Gotion: The score you can't demonstrate from public data
Gotion's H1 2026 filing leads with RMB 1.386 billion of attributable net profit, up 278% year over year. That figure includes government subsidies, asset disposals, and other one-time items, which Chinese GAAP filings categorize as "non-recurring gains and losses." Strip them out and attributable net profit was RMB 106.9 million.
That line sits below operating expenses, financing costs, taxes, and minority interests, so it is not operating profit. It is, however, the closest the filing comes to isolating what Gotion earns from making and selling batteries on a recurring basis. RMB 106.9 million against RMB 27.8 billion of recognized revenue is a margin that rounds to 0.4%.
Cash generation is a separate question. Reported net operating cash flow was RMB 428 million. Against that, net accounts receivable rose RMB 5.6 billion during the half, from RMB 16.3 billion to RMB 21.9 billion. Gross receivables reached RMB 25.2 billion with a bad-debt provision of RMB 3.3 billion. The aging schedule shows RMB 2.9 billion aged two years or more, including RMB 822 million over five years. Gotion does not disclose days-sales-outstanding or standard payment terms.
Revenue grew; cash came in slower than sales went out. Total cash and equivalents increased RMB 1.3 billion to RMB 14.1 billion, but that increase was funded by RMB 9.8 billion of net financing inflows, borrowing and equity proceeds, against RMB 8.5 billion of investing outflows. Strip out financing and the operating business barely covered its capital spending.
Gotion also does not disclose the metric the gate is said to score on.
Gotion's 223-page H1 report contains no company-wide production capacity figure, no production output figure, and no utilization rate. If the capacity gate scores producers on utilization, Gotion's competitive position cannot be assessed from its public filings.
The 28 GWh figure widely repeated in English-language coverage is power-battery installations in customer vehicles, sourced from CABIA. It measures what Gotion's customers deployed, not what Gotion's factories produced. Installation and production are different measurements with different denominators, and CABIA's production-sales residual is not an inventory identity, so it cannot be used to back into factory output. No investor presentation supplying the missing figures appears on Gotion's English-language IR pages.
I flagged this absence in Issue #12 and it remains unresolved. Outside analysts cannot construct the number, and neither can competitors. Whether Gotion reports utilization to regulators through non-public channels is unknowable from outside the process.
Taken together, the financials describe difficult economics for competing on throughput. At 0.4% recurring margin, the profit on an incremental cell is negligible. Receivables grew RMB 5.6 billion in six months against RMB 428 million of operating cash flow, so revenue is being recognized far faster than cash is collected. Producing more cells to improve a utilization score, under these terms, means adding volume that generates almost no recurring profit and whose proceeds may sit in receivables for a year or more before converting, if they convert at all. Gotion's throughput is funded by borrowing and subsidy income rather than by the margin on the cells.
REPT: An ESS pivot that may or may not count
REPT's H1 2026 results look like recovery. Revenue rose 57.2% to RMB 14.9 billion. Profit reached RMB 778.3 million against a RMB 62.7 million loss a year earlier. As I covered in Issue #13, the recovery is ESS-weighted: ESS shipments rose 43.9% to 27.2 GWh while power-battery shipments rose 14.8% to 15.5 GWh.
Whether ESS carries structurally different margins than power batteries cannot be determined from the interim announcement, because REPT does not break out segment-level profitability. The consolidated swing from loss to profit coincided with ESS rising to 63.7% of shipments by volume, but the filing does not isolate how much of the margin recovery came from ESS pricing rather than cost reduction or volume leverage. The full interim report, expected by end of September, may add segment detail. Until then the durability of the recovery rests on an undisclosed margin structure.
The cash picture is tighter than the profit improvement suggests. REPT generated RMB 3.46 billion of operating cash and spent RMB 3.30 billion on property, plant, and equipment, using REPT's own disclosed capex figure. The difference is RMB 161 million.
A correction: in Issue #12 I described REPT as spending roughly twice its operating cash on capital expenditure. That was imprecise. The 2x figure was total net investing outflow of RMB 7.0 billion, which included RMB 14.3 billion of financial-asset purchases partly offset by RMB 11.4 billion of disposals. On a narrow capex basis the ratio is closer to 1:1, which makes REPT's operations more nearly self-funding than the earlier framing implied.
The balance sheet is where the fragility sits. Interest-bearing borrowings rose to RMB 12.4 billion from RMB 9.7 billion at year-end, of which RMB 4.6 billion matures within twelve months. Cash and equivalents were RMB 5.7 billion, giving derived net debt of roughly RMB 6.7 billion, a figure REPT does not itself report. The liability-to-asset ratio reached 77.1%. Capital commitments for contracted construction not yet paid stood at RMB 2.9 billion. Net financing inflows of RMB 4.4 billion during the half, including RMB 5.2 billion of new bank loans, funded the gap between operating cash generation and total investing activity.
Then there is the scoring question. REPT's utilization story is overwhelmingly an ESS story: 27.2 GWh against 15.5 GWh of power-battery shipments. If the gate scores power-battery and ESS output separately, the ESS volumes may not help REPT's power-battery approval prospects at all. If it combines them, REPT's 42.7 GWh total tells a considerably stronger story.
Nothing in the public record resolves this. Caixin's September 10 report indicated authorities surveyed both power- and ESS-battery capacity and utilization, but said nothing about whether the GWh are combined into one score. For REPT that ambiguity compounds the rest: the pivot that restored profitability may or may not satisfy the gate, its segment economics are undisclosed, and either path requires continued bank lending into a balance sheet already at 77% gearing with RMB 4.6 billion due inside a year.
What the gate selects for
Whatever its stated intent, the mechanism reads from these filings as a financial filter with a specific bias. CATL sustains high utilization because its payables structure funds the inventory accumulation, its margins appear to hold through the buildup, and its cash generation stays positive across the cycle, with the cost distributed across its supply chain. Gotion cannot show a utilization figure from public data at all, and at 0.4% recurring margin it would be funding any throughput gain from borrowing rather than from operations. REPT can demonstrate throughput but depends on continued credit access, on ESS economics it does not disclose, and on a definitional question the regulator has not answered.
A gate that asks whether a producer can keep making cells, without asking whether it can sell them, selects for the balance sheet that was already largest.
Filing dates: CATL's H1 report was released July 26, 26 days after period end. Gotion's appeared August 24, 55 days after. REPT's interim announcement came August 26, 57 days after, with the full report expected by end of September. All figures are as reported in RMB unless otherwise noted.
- Capacity gate instrument: No ministry decree, NDRC circular, or MIIT gazette establishing the utilization-based approval mechanism had appeared on government portals as of September 7, so the scoring formula, threshold, and approving authority remain unverified.
- Grandfathered project treatment: Hunan's development commission approved a revised energy review in July for a 20 GWh project whose original 2023 plan had undergone major design changes, leaving open how substantially altered pre-May projects are classified under the reported freeze.
- CATL inventory composition: The H1 filing does not disclose the product age, format mix, or committed destination of the RMB 47.6 billion in finished goods, so whether the buildup reflects 314 Ah cells, large-format inventory, or timing stock remains unknown.
- Consumption tax incidence: InfoLink reported manufacturers using price letters and supplemental agreements to pass the September 1 tax to customers, but no public source has yet reported matched pre- and post-tax executed contracts that would establish how much of the 2% producers actually retain.