Cash Conversion

Cash Conversion

Bottom line. Goldkey's record FY2025 profit did not become cash. Net income of NT$446m sat against an operating cash outflow of NT$1,774m, a gap funded by borrowing and IPO proceeds. Almost the entire outflow is a working-capital build: inventory more than tripled to NT$2,619m — half the balance sheet — while receivables rose NT$600m. The earnings are real accounting profit, but low quality as cash, and the pattern recurs across the memory cycle rather than being a one-off.

Record profit, negative cash

FY2025 reported net income of NT$446m on revenue of NT$7,704m [1]. The consolidated cash-flow statement tells a different story about the same year: operating activities used NT$1,774m of cash [2]. The NT$2.2bn distance between the two is the subject of this chapter.

Source: FY2025 Annual Report, financial performance and cash-flow analysis [3] [4].

The audited cash-flow reconciliation shows where the money went. Starting from pre-tax profit of NT$561m, the dominant adjustments are not accounting add-backs but movements in working capital: inventory absorbed NT$1,879m of cash and receivables another NT$600m, offset only partly by NT$153m more owed to suppliers. Cash generated from operations was NT$1,703m negative before interest and tax [5].

Source: FY2025 audited cash-flow statement; non-cash items and payables/other grouped for clarity [6].

Management is candid about the cause. The annual report attributes the swing directly to "advance stocking to build inventory in response to anticipated future market demand" [7]. This is not an earnings-recognition problem — the profit is audited and the receivables are collectible (below). It is a capital-intensity problem: the way this business grows is by tying up cash in memory chips.

Nearly two hundred days of working capital

The clearest way to see the strain is the cash conversion cycle — how long a dollar is locked in inventory and receivables before it comes back as cash, net of what suppliers finance. In one year it nearly doubled, from about 106 days to about 199 days, driven overwhelmingly by inventory days rising from 50 to 138.

Source: derived from FY2025 audited balance sheet and income statement — inventory NT$2,619m, receivables NT$1,499m, payables NT$187m against COGS NT$6,930m and revenue NT$7,704m [8].

Two features stand out. Payables are strikingly short — under ten days — because the upstream chip makers extend little credit; Goldkey pays for silicon well before it collects from customers, so almost none of the inventory is supplier-financed. And the receivables, though large, are clean: at year-end effectively all of the NT$1,500m book was current, with only NT$15m overdue and under thirty days, against a loss allowance of less than NT$1m [9]. The cash is not stuck because customers won't pay; it is stuck because the model carries months of chip inventory and settles with suppliers first.

The inventory question

Inventory is now the defining line on the balance sheet: NT$2,619m, or 50% of total assets, up from NT$714m (28%) a year earlier [10]. It is carried at the lower of weighted-average cost and net realisable value, with a valuation allowance for obsolescence and price declines that closed the year at NT$117m, up from NT$92m [11].

That allowance is where cyclicality meets the income statement. When memory prices fall, Goldkey writes inventory down and the charge lands in cost of sales; when prices recover, the earlier write-down is reversed and credited back to cost of sales, lifting reported gross margin. FY2025 gross profit of NT$774m includes a NT$25.5m net reversal of prior write-downs [12]. Excluding it, the gross margin is about 9.7% rather than the reported 10.0% — the reversal is real but modest, roughly a third of a percentage point.

The longer record shows the mechanism working in both directions. In the FY2022 downturn the company took a NT$76m inventory write-down that helped push that year's gross margin to 2.0% [13]; across the FY2023–FY2025 recovery it released reserves back — NT$62m, NT$15m, then NT$26m — each release a tailwind to margin [14].

Source: cost-of-sales notes, FY2020–FY2025 audited statements; a positive bar reduced gross margin, a negative bar lifted it [15] [16].

The disclosed reversal is small; the undisclosed exposure is not. Under weighted-average costing, a rising memory price also flatters margin as chips bought earlier and cheaper are sold at today's higher prices — a holding gain embedded in the 10% margin but not separated out. The same arithmetic runs in reverse if prices roll over. The FY2022 write-down of NT$76m was taken on an inventory base of well under NT$1bn; the NT$2,619m now on the books was accumulated during a price ramp, so a downturn comparable to FY2022 would apply to a base several times larger — the recovery-year margin and the recovery-year inventory bet are the same wager.

A cash consumer across the cycle

Over the full seven years the relationship is close to inverse: operating cash is strongest when profit is weakest, and weakest when profit is strongest. FY2022 earned only NT$18m but generated NT$636m of cash as a falling market unwound working capital; FY2025 earned NT$446m and consumed NT$1,774m as a rising market rebuilt it.

Source: reported net income and operating cash flow, FY2019–FY2025 filings [17] [18].

Summed across the cycle, the two diverge sharply: about NT$1,131m of cumulative reported net income against roughly NT$1,895m of cumulative operating cash outflow over FY2019–FY2025. With capital spending negligible — the manufacturing is outsourced, so FY2025 capex was only NT$27m [19] — this is neither a plant-investment story nor a bad-debt story. It is purely working capital: a low-margin distributor that reinvests each up-cycle's profit, and then some, into more inventory.

There is a genuine two-sided reading here. The bull case is that the cash burn is the cost of growth — revenue rose 84% from NT$4.2bn in FY2023 to NT$7.7bn in FY2025, and a growing distributor must permanently fund a larger book; FY2022 shows the model does release cash when growth stops. The bear case is that the reinvestment happens at the worst possible time — buying the most inventory when prices are highest — so reported profit is chronically trapped in stock bought near the peak. What would move the read is a full cycle in which cumulative operating cash turns positive, or evidence that working-capital intensity per dollar of sales is structurally falling; so far it has risen.

What management projects for FY2026

The most direct evidence on durability is the company's own forward liquidity plan. The annual report projects that FY2026 operating activities will again be cash-negative, at about NT$645m outflow, and that after investing and financing needs the year opens with a projected NT$1,177m shortfall against a starting cash balance of only NT$61m — to be met by issuing a second domestic convertible bond [20].

Source: FY2025 Annual Report, next-year cash-liquidity analysis; remedy stated as a second unsecured convertible bond issue [21].

The strain already shows in three places. Year-end cash of NT$61m stands against NT$1,221m of short-term borrowings, so the company is running on rolled-over bank lines [22]. It has begun factoring receivables with recourse — US$4.5m of trade receivables pledged for NT$141m of bank funding in FY2025, retained on the books because the credit risk is not transferred [23]. And the rising debt is now visible in the income statement: finance costs climbed to NT$43m in FY2025 from NT$28m in FY2024 [24], and will carry a full year of the enlarged borrowings and the convertible bond into FY2026.

None of this makes the FY2025 profit fictitious. The measured read is narrower: Goldkey's earnings are real but structurally cash-absorbing, and the FY2025 record was accompanied by the largest inventory bet in its history, funded by debt and equity rather than by the business. Whether that bet reads as durable earning power or a cyclical windfall is most sensitive to memory prices, which will decide both the value of the inventory and the margin on what it sells.