Variant Perception

Where We Disagree With the Market

The market is pricing Auros as a Korean memory-cycle trough stock with an adjacency option starting to convert; the evidence says the adjacency option is decaying, the FY25 cash flow positivity is engineered, and the lone analyst's +55% FY26 revenue snap-back will not produce the operating leverage the multiple is paying for. The lone covering analyst's ₩31,000 price target embeds a ₩81 bn FY26 revenue rebound (back-of-envelope to a 2.87× forward EV/Sales) and a swing to ₩10 bn net income, anchored on the November 2025 Samsung HBM contract and the April 2026 trade-press confirmation of thin-film qualification. We disagree on three grounds: the ₩6 bn Samsung order is 1.2% of revenue while Camtek and Onto already own the same back-end niche at SK Hynix and Samsung at multiples of Auros's scale; the FY25 +₩1.7 bn operating cash flow was composed of ~₩11.7 bn of inventory unwind plus a related-party asset disposal; and FY24's record revenue year still printed −₩5.3 bn OCF because working capital absorbs every up-cycle. The debate is resolvable: a Samsung HE-900IR repeat order, a first thin-film volume contract, or an SK Hynix hybrid-bonding qualification flips the verdict toward consensus; their absence through 1H FY27, combined with another working-capital-absorbed up-cycle, confirms it.

Variant Perception Scorecard

Time to resolution: 6–18 months

Variant strength is 65 because the three disagreements are specific, testable, and contradict the lone-analyst model rather than a vague "watchlist" view — but they are not heretical. Consensus clarity is 60 because Auros has one covering analyst, no formal earnings guidance, and a Korean-only quarterly cadence; the consensus signal is the analyst's ₩81 bn FY26 model, MarketScreener forward multiples, and Korean trade-press tone. Evidence strength is 70 because the disagreement runs on disclosed accounting (cumulative FY21-FY25 FCF, related-party schedule, FY24 OCF) and on publicly documented competitive scale (Camtek $496 mn, Onto $1.0 bn already commercialised at SEC/SK Hynix). The 6-18 month window covers the next two reporting cycles, the Samsung extension window, and the thin-film volume-order window.

Consensus Map

The consensus is not loud but it is identifiable: a single analyst's snap-back model, supportive Korean trade press, and a holdout EV/Sales discount to a global metrology cluster. The variant question is whether the three load-bearing assumptions — that revenue is the binding constraint, that the Samsung HBM order is a recurring beachhead, and that thin-film qualification converts on a 12-24 month clock — actually hold under the report's evidence.

The Disagreement Ledger

Disagreement #1 — Option decay. Consensus reads the November 2025 Samsung HE-900IR contract and the April 2026 The Elec qualification confirmation as the moment the four-year pilot turns into a multi-product specialist. The evidence is harder: the Samsung order is 1.2% of revenue and represents Auros's third Samsung order in two years, but Camtek (FY25 revenue $496 mn) and Onto ($1.0 bn) already commercialise the same Korean HBM-inspection and thin-film TAM at scale, and both are qualified at SEC and SK Hynix. The market would have to concede that ~₩130 bn of embedded option value depends on volume, not on qualification. The cleanest disconfirming signal is a Samsung HE-900IR repeat order or a thin-film volume contract inside 12-18 months — both observable in DART filings or Korean trade press before the lone analyst can revise.

Disagreement #2 — Engineered cash flow. Consensus reads positive FY25 OCF (+₩1.7 bn) as the canonical late-cycle pattern — inventory unwinds, working capital flips, OCF leads revenue recovery. The evidence is that the OCF is composed of ₩6.3 bn inventory release plus a ₩5.4 bn related-party asset disposal to CM Technology (the same affiliate that holds 17.27%); underlying operations contributed approximately −₩10 bn. The forensic specialist graded this Red (Unsustainable CFO levers); external coverage does not. The market would have to concede that the cumulative −₩37 bn 5-year FCF is structure, not cycle, and that the FY26 working-capital cycle will absorb the snap-back rather than release it. The cleanest disconfirming signal is FY26 OCF trajectory net of working capital, observable in 1H FY26 분기보고서.

Disagreement #3 — Revenue snap-back without operating leverage. Consensus FY26 model translates +55% revenue (₩52 bn to ₩81 bn) into +₩16 bn of incremental net income (−₩6.4 bn to +₩10 bn), implying an operating margin equal to or above FY24's all-time peak of 9.9%. The evidence is that FY24's record year ran R&D at ₩16.6 bn while FY26 starts with R&D ₩19.5 bn (FY25 actual) — 18% higher fixed cost base. The market would have to concede that even at the lone analyst's revenue, op margin caps around 8-10% (op income ₩6-8 bn) rather than 12-15%, which makes the forward P/E 45-65× on positive earnings rather than the 25× that supports the ₩31,000 PT. The cleanest disconfirming signal is 1H FY26 gross-margin trajectory plus R&D-to-revenue TTM, both visible quarterly.

Disagreement #4 — Tape as leading evidence. Consensus reads the May 2026 21% drawdown as KOSDAQ distribution because no English-language coverage explains the cause and the lone analyst has not revised. The evidence is that the Q1 FY2026 분기보고서 was filed in Korean on 2026-05-15 — the same session as the −7.17% close — and three insider-holdings reports clustered 2026-05-07 just before the sell-down. Korean trade press has historically front-run DART by weeks. The market would have to concede that tape is processing knowable information before English-language coverage can. The cleanest disconfirming signal is the English-language read-through of the Q1 print inside 30 days.

Evidence That Changes the Odds

How This Gets Resolved

The two signals that resolve the long-term debate sit at rows 1 and 2 — Samsung HE-900IR extension and a first thin-film volume order. Row 4 (Q1 FY26 GM + net debt) resolves the near-term tape disagreement. Rows 3 and 5 (FY26 OCF composition and R&D-to-revenue trajectory) resolve the structural cash-conversion and operating-leverage disagreements. Row 7 (KLAC commentary) is a base-moat backstop the variant view does not contest but a PM should track because a failure there is larger than the option premium.

What Would Make Us Wrong

The variant's biggest analytical risk is time horizon error on the adjacency option. Five years of pilot status is unusual but not unprecedented in Korean semicap equipment qualification — Park Systems took roughly a decade to graduate from a niche AFM specialist to a ₩2 trillion market cap with 20% op margins, and several of those years also looked like option decay from the outside. The April 2026 The Elec confirmation is genuine evidence the option is closer to converting than the FY25 사업보고서 implies, and trade press in Korea consistently precedes formal DART disclosure by months. If the Samsung 5-unit extension lands inside the next 6-12 months, or if SK Hynix qualifies Auros on hybrid-bonding even without a Samsung repeat, disagreement #1 collapses on its own evidence and the option becomes substantially in the money. The PM check on this: take seriously any single Korean-press follow-up on Samsung HE-900IR or SK Hynix HBM4 vendor selection — those scoops are the leading-indicator channel and would refute the variant view before the lone analyst can revise.

The cash-conversion disagreement (#2) is fragile if Auros has structurally shifted its working-capital model post-FY24. The FY24 receivables build (+₩8.6 bn) and inventory build (+₩10.6 bn) was outsized because Q4 FY24 shipped ₩28.9 bn — 47% of full-year revenue in one quarter — and that customer concentration into year-end is not necessarily repeatable. If FY26 ships more evenly through the year, working capital may absorb cash less aggressively than the FY24 baseline implies, and the lone-analyst FY26 model could deliver +₩5-7 bn of FCF rather than the negative-to-zero we expect. The ₩5.4 bn CM Technology asset disposal is a one-off — its absence in FY26 is the base case, not a deterioration. The PM check on this: 1H FY26 DSO needs to print below 60 days alongside revenue growth; if it does, the cash-conversion variant weakens materially.

The operating-leverage disagreement (#3) overweights the R&D dollar trajectory. Management could pull a one-time R&D pause to deliver a clean FY26 P&L — the FY24 R&D-to-sales ratio of 27.1% shows the company already let the ratio breathe during the previous up-cycle. If FY26 R&D holds at ~₩18 bn while revenue snaps to ₩81 bn, the ratio compresses to 22.2% and op margin can plausibly reach 10-12% (op income ₩8-10 bn). The variant view assumes R&D dollars grow with the cycle; they may not. The PM check on this: 1H FY26 R&D run-rate. A flat or modestly declining quarterly print versus 4Q FY25 (₩4.8 bn) would weaken this disagreement.

The tape-as-evidence disagreement (#4) is the weakest of the four because KOSDAQ small-cap technical noise is real, and the absence of a follow-up Korean trade-press piece between 2026-04-03 and 2026-05-15 is itself a data point. If the English-language read-through of the Q1 print arrives benign in the next 30 days, this disagreement collapses immediately and only the structural variant views remain.

The first thing to watch is the English-language read-through of the Q1 FY2026 분기보고서 filed on 2026-05-15 — specifically the gross margin, net debt, and any commentary on Samsung HE-900IR + MT-30T installation completion at Cheonan / Onyang.