Concentration
Concentration
Bottom line. Goldkey's revenue rests on a narrow base. Two customers supplied 41% of FY2025 sales, and the company buys its raw material from a three-firm memory-chip oligopoly, its two largest suppliers around 42% of purchases. The auditor's sole key audit matter for FY2025 is the genuineness of revenue from a few large, fast-growing customers — not inventory. Concentration eased as Taiwan sales tripled, but the base a 6%-share price-taker depends on stays thin at both ends.
The customer base
Two customers each cleared 10% of Goldkey's FY2025 revenue and together accounted for NT$3,174m of the NT$7,704m top line — 41.2% [1]. Customer A alone was NT$1,826m (23.7%); Customer B NT$1,348m (17.5%) [2]. Neither is named in the filings.
Customer A, % of revenue
Customer B, % of revenue
Top two combined
US-located revenue
Source: FY2025 Audited Financial Report, Note 28 (major customers) [3] and Note 34 (revenue by customer location) [4].
The direction of travel matters as much as the level. In FY2024, Customer A supplied NT$2,279m — 41.4% of that year's revenue on its own — and the top two together were 60.1% [5]. The FY2025 surge did not come from those two accounts; Customer A's revenue actually fell in absolute terms. It came from elsewhere, so the base widened even as it stayed concentrated.
Source: FY2025 Audited Financial Report, Note 28 — customer revenue as a share of the annual total [6].
Where that widening came from is visible in the geographic split. Revenue by customer location shifted decisively toward Taiwan — from NT$2,416m (43.9%) in FY2024 to NT$4,630m (60.1%) in FY2025 — while US-located revenue fell from NT$2,312m to NT$1,894m even as the total grew [7].
Source: FY2025 Audited Financial Report, Note 34 — revenue disaggregated by customer location [8].
One inference is worth drawing. Customer A's revenue tracks US-located revenue almost exactly in both years — NT$1,826m against NT$1,894m in FY2025, and NT$2,279m against NT$2,312m in FY2024 — which points to the largest single account being the US channel. That leaves the biggest customer relationship exposed to US trade and inventory policy on top of the ordinary risk of a concentrated account. The filings do not confirm the identity, so this is a read from the numbers, not a disclosed fact.
The supplier side
Concentration runs through the supply side as well. Goldkey's raw material — DRAM and flash ICs — comes from what its own prospectus calls an oligopolistic ("寡佔") market supplied by a handful of large wafer makers, principally SK Hynix, Samsung and Micron [9]. Those three set memory prices; Goldkey, a NT$7.7bn module assembler, takes them.
Purchases are as concentrated as sales. In FY2024 the two largest suppliers were 42.1% of total purchases; in FY2023, 48.6% [10]. The company reports no single supplier above 50% of purchases and describes a multi-source strategy [11], and the top supplier does rotate — Supplier A led in FY2023, a different Supplier C in FY2024, with Supplier B second in both years [12].
Source: Listing Application Prospectus 2025 — suppliers exceeding 10% of annual purchases [13].
The picture is a firm with ~6% domestic share (Competitive Standing) sitting between a supply side it cannot influence and a demand side of a few large accounts. Its gross margin is the spread between two concentrated ends it does not control. That is why the FY2025 record reads as a good pass-through year on a rising memory price rather than pricing power. The structure affords Goldkey little of the latter.
The auditor's key audit matter
For FY2025 the auditor, Deloitte, named a single key audit matter: the genuineness ("真實性") of revenue from specific customers. Its stated reason is that certain customers' annual sales were material and rose sharply against the prior year — FY2025 revenue grew NT$2,196m — so it treated the occurrence of that revenue as the year's principal audit risk and sampled those accounts against credit-limit approvals, original orders, shipping documents and cash receipt [14].
The FY2025 audit opinion is unqualified. What is notable is the choice of matter: with inventory at 50% of assets and its valuation the obvious candidate, the auditor instead singled out revenue occurrence from a few large, fast-growing accounts — the same concentration the customer note describes.
A key audit matter is a focus disclosure, not a finding of misstatement; the accounts passed. But the auditor and the concentration data point the same way. When 41% of revenue runs through two unnamed accounts and one of them appears to be the US channel, the quality of that revenue — that it occurred, on arm's-length terms, and converts to cash — is exactly where a skeptic looks first. That it did not convert to cash in FY2025 is the subject of Cash Conversion.
Receivables
The concentration carries into the balance sheet. At year-end the five largest trade-receivable balances were NT$1,180m of the NT$1,500m book — 79% — with the single largest at NT$332m [15]. Credit quality itself looks clean — almost all receivables are current and the loss allowance is negligible — but the exposure is lumpy: a single large account going bad would matter. To fund the working-capital build, Goldkey also factored US$4.5m of receivables with recourse, retaining the credit risk and keeping NT$224m of factored receivables on the balance sheet as collateralized borrowing [16].
The read
Concentration on both sides is real and, for a sub-scale price-taker, structural — it is the shape of the business, not a passing feature. The strongest fact against reading it as an acute risk is direction: the FY2025 growth came from outside the two largest accounts, the top customer's share fell from 41% to 24%, and the top-two supplier share eased from 49% to 42%, so the base is widening as the company grows. The company also markets an AI industrial-control division, set up in FY2024, as the source of newer, more diversified demand [17].
What would change the read is disclosure the corpus does not yet contain: who Customers A and B are, and whether the Q1 FY2026 revenue tripling (Priced for Peak) was broad or ran back through the same one or two accounts. A concentrated surge would sharpen the risk the auditor flagged; a broad one would confirm the base is genuinely widening. Until then, the honest statement is that Goldkey earns a thin, cyclical spread between an oligopoly it buys from and a short list of customers it sells to — and that its auditor named the genuineness of those sales the sole key audit matter for FY2025.