Short Interest & Thesis

Short Interest and Thesis — Goldkey Technology (3135)

Bottom line. No official reported short-interest position exists in this run: Taiwan's TWSE does publish per-stock margin-short (融券) and borrowed-share (借券) balances daily, but none were staged here, and no short-sale-volume, borrow-cost, public net-short, or peer-crowding rows are available. There is also no public short-seller report, activist campaign, accounting allegation, or litigation anywhere in the corpus — a search of every filing returns nothing, and the company itself discloses no pending legal or non-litigation proceedings [1]. So the honest institutional read is that measured short positioning is not decision-useful here. What is decision-useful is the market-structure and dilution setup you can build from the tape and the company's own disclosures: a newly-listed (6 Aug 2025) memory-module name that has run roughly five-fold on a retail-driven, repeatedly limit-up tape, a thin free float, deeply negative operating cash flow funded by serial capital raises, and a live convertible-bond overhang. Those are the real thesis risks — not a crowded short.

Reported positioning — what exists and what does not

Every structured short-interest input staged for 3135 is empty: the manifest reports zero reported-short-interest rows, zero short-sale-volume rows, zero public net-short rows, zero borrow rows, and zero peer rows, with no fetcher configured for this market in the v1 pipeline. That is a coverage gap, not a signal of low short interest. Taiwan is not a dark market — the TWSE and the OTC exchange publish per-security margin short-sale balances (融券餘額) and securities-borrowing-and-lending short sales (借券賣出) on a daily basis — so a reported short-interest series does exist in principle and is the single most valuable thing a later research pass could add. Two structural caveats matter for a name this young: newly-listed Taiwan shares face an initial restriction window before margin financing and short selling are broadly enabled, and Taiwan's 10% daily price limit means shorts caught in the repeated limit-up (漲停) sessions this stock has printed cannot always cover on the day.

Source: short-interest staging manifest and latest snapshot (reported short interest, short-sale volume, borrow — all staged empty).

Positioning read from the tape (proxy, not short interest)

With no position data, the price/volume tape is the only positioning read available — and it is a loud one. The stock listed on the TWSE main board on 6 August 2025 at an indicative NT$28, opened its first session up 15.54% to NT$32.35, then dropped to limit-down (NT$36.65) within days [2]. From there it has run in a near-vertical, repeatedly limit-up sequence to a NT$266 intraday peak in May 2026, last quoted NT$167 — roughly five times the offer price. This is textbook one-directional retail momentum, not a tape that suggests a crowded, comfortable short.

Source: daily price/volume feed as staged (month-end close and average daily share volume); listing detail per news summary [3].

The volume tells the crowding story better than any single price: average daily volume swelled from ~0.24M shares pre-listing to a 6.1–6.3M-share monthly pace during the January and March–May 2026 blow-off. Against roughly 77.5M shares outstanding, that is a daily turnover near 6% of the entire share count — the hallmark of a heavily day-traded retail vehicle. One local outlet flagged the setup directly, noting an AI-memory theme firing while "short-term chip (positioning) pressure remains" [4] — generic technical commentary about profit-taking and share supply, explicitly not a short-interest reading.

Crowding vs liquidity

Sources: shares outstanding 77,493,405 [5]; top-10 / family holdings [6]; ADV and turnover derived from the staged price/volume feed. Days to cover is not computable — no reported short interest exists.

Two facts define the crowding math. First, the register is concentrated: the top ten holders own about 49% of the shares, and a founder-family-related bloc — the chairman's investment vehicle at 18.52%, plus related-party vehicles and relatives — exceeds roughly 35% [7]. That leaves a genuinely thin tradable float. Second, that thin float is being turned over at ~6% per day. A low float turned violently is squeeze-prone in both directions: any short that did exist would be hard to cover into limit-up prints, but the same thin float is exactly what let the stock round-trip from NT$266 to NT$167 in six weeks. Days-to-cover cannot be calculated because there is no reported short interest to divide by ADV — stated plainly rather than proxied from volume.

The real overhang — dilution and convertible bonds

The substantive "positioning" risk on this name is not a short book; it is supply of new shares. Goldkey has raised capital repeatedly and carries a live convertible-bond overhang.

Source: FY2025 Annual Report, Capital and Shares (share-formation table) [8].

Paid-in shares reached 77,493,405 after the September 2025 IPO cash increase, against 200M authorized [9]. The chairman's stake illustrates the dilution: her vehicle held 28.15% at the last board election but only 18.52% now, as the count expanded [10]. On top of that sits the convertible overhang:

Source: FY2025 Annual Report, Corporate Bonds (CB #1 and CB #2 terms and conversion) [11]; CB #2 conversion price [12].

The first zero-coupon CB (NT$1bn, issued December 2025, conversion price NT$50.85) was struck well below the runaway share price and has been almost entirely converted — 17,356,828 new shares issued as of 30 April 2026, equal to about 22% of the post-IPO base, with only NT$117.4M left outstanding [13]. That is a large, recent expansion of the share count. The second zero-coupon CB (NT$1.5bn, issued May 2026, conversion price NT$129.9) is still entirely unconverted and is now in the money at the NT$167 tape — a fresh, sizeable dilution and share-supply overhang [14] [15]. Local reporting indicates the company is targeting NT$6–10bn of fundraising in 2026 to secure inventory, so the dilution path is likely not finished.

Short-thesis ledger — built from the company's own disclosures

With no external short report to adjudicate, the disciplined move is to construct the bear case a credible short would make from primary disclosures, and mark what the company says back. Each row separates the risk, the evidence, the company's response, and whether it is unresolved.

Sources: operating cash flow and FY2026 shortfall [16]; balance-sheet expansion and CB-funded growth [17]; margin/EPS [18]; CB terms [19]; governance and concentration [20] [21]; inflation/memory cost [22].

Two mitigants a short must respect. The company reports no material litigation, non-litigation, or administrative proceedings, and no supplier exceeding 50% of purchases and no sales over-concentration [23]. And because both sales and raw-material purchases are USD-denominated, the company runs a natural FX hedge, limiting currency risk [24]. Profitability is real — FY2025 revenue NT$7,704M, gross margin ~10%, EPS NT$6.33, ROE ~27% [25]. The bear case is a cash-flow, dilution, and cycle-timing case, not a fraud case.

Borrow pressure — no data; one inferred channel

No borrow-cost, utilization, lendable-supply, or hard-to-borrow data was staged, so borrow pressure cannot be assessed directly. One channel is worth flagging as inference, not measurement: zero-coupon convertibles like Goldkey's two CBs typically attract convertible-arbitrage buyers who hedge by borrowing and shorting the underlying stock. That mechanism can create genuine borrow demand and a technical short base tied to the bonds — most plausibly around the fresh, still-outstanding NT$1.5bn CB #2. This is a hypothesis about where any borrow demand would come from, not evidence that a position exists; confirming it requires the TWSE 借券 balance a later research pass should pull.

Evidence quality

Sources: staging manifests (short interest / borrow — not staged); corpus search (no short-seller report); FY2025 Annual Report for litigation, dilution, and ownership [26] [27].

Setup for the PM. Positioning is not a reason to own or avoid 3135 on measured short interest — there isn't any staged, and no credible short thesis is public. The reasons to size and time carefully are structural: a five-fold, limit-up momentum tape on a thin, founder-controlled float, deeply negative operating cash flow, a self-disclosed FY2026 funding gap plugged by a NT$1.5bn convertible, and continuing dilution. If a later pass pulls the TWSE 融券/借券 balances and they turn out elevated, this flips from a pure dilution-and-cycle story into a genuine crowding question; until then, the risk is de-rating and share supply, not a squeeze.