Business

Know the Business

Auros is a sub-scale Korean specialist that holds one of only two industry-bench seats in wafer-overlay metrology — the photolithography measurement step every leading-edge DRAM and 3D-NAND fab must pass to hit yield. The base business is a hard-to-displace duopoly position at Samsung and SK Hynix that throws off 50-60% gross margins when shipments come; 70% of revenue rides on one product family from two customers, so a single quarter of Korean memory CapEx digestion can crater the P&L. The market treats this as a leveraged Korean memory cycle stock with an option on adjacency (thin-film + HBM packaging). The two open questions are how long the option takes to convert — four years of thin-film R&D, zero volume orders yet — and whether the cycle alone justifies a 5× EV/Sales multiple at trough margins.

1. How This Business Actually Works

Auros makes one thing well and sells it to two customers: a wafer-overlay metrology tool that fits inside a Korean memory fab's photolithography cell, measures how precisely each photoresist layer landed on top of the last, and lets the customer correct litho before billions of dollars of work scraps. A tool sells for roughly ₩2-4 bn, a wafer-demo takes months, and qualification has taken over a decade — but once installed, the tool generates a long service tail and gets functional upgrades that re-monetise the box. Profit comes from being one of only two qualified vendors at the customer (KLA is the other) and from R&D cadence — shipping a next-gen platform every 2-3 years matched to the next memory node.

The cost structure tells the rest of the story. Roughly 70% of materials cost is concentrated in three imported sub-systems — precision stages (₩9.2 bn FY25), optics/lasers (₩5.0 bn) and vision modules (₩3.8 bn). That fixes the gross-margin ceiling somewhere in the high 50s. The real swing variable is the SG&A line — and inside SG&A, R&D, which ran ₩19.5 bn in FY2025, or 36.7% of revenue, a five-year high and the reason FY2025 swung to a ₩8.4 bn operating loss despite still-positive gross profit. Management deliberately did not flex R&D in the downturn because the thin-film and back-end programmes need uninterrupted development if Auros is to graduate from "one-product specialist" to "two-or-three product specialist."

2. The Playing Field

Process control inside Wafer Fab Equipment is the most concentrated slice of the semi-cap industry. KLA controls roughly half the global profit pool; the rest is split among five-to-eight specialists. Inside wafer overlay specifically the industry is effectively a two-vendor club: KLA + Auros, with ASML (via Hermes Microvision) and Onto competing in adjacent niches but not currently selling a wafer-overlay system into Korean memory in volume. That tiny duopoly is Auros's high-quality asset. Everything else on the income statement reflects the cost of trying to leave it.

Two stylised facts. First, the metrology-peer scale curve runs from KLA at the top-right (operating-leverage prize at scale) down to a cluster of $0.5-1.0 bn specialists, then a steep drop to Auros — at $36 mn it is one-tenth the smallest US-listed peer and one-three-hundredth of KLA. Second, NVMI / CAMT / ONTO / Park Systems all earn 13-29% operating margins on similar gross margins to Auros; the gap is almost entirely R&D-as-percent-of-sales. The question for the equity is whether Auros can grow into that cluster's economics — not whether the addressable market is real.

What "good" looks like in this industry: R&D-intensity at 12-15% of sales, gross margin at 55-60%, ROE above 20%, service mix above 20%. Auros's R&D-intensity is more than double that of the cluster and its service mix is half — both gaps reflect installed-base size, not effort. The KLA → NVMI → Auros progression is the scale-economics ladder: every rung up adds installed base, which dilutes R&D, which lifts operating margin, which compounds into the equity multiple.

3. Is This Business Cyclical?

Yes — and the cycle hits four places at once: order timing, gross-margin absorption, R&D-as-percent-of-sales, and inventory. Auros has now been through one full mini-cycle in eight quarters: revenue rose from ₩3.8 bn in 1Q24 to ₩28.9 bn in 4Q24 (operating margin +35%), then halved to ₩7.6 bn by 3Q25 (operating margin -77%). Same plant, same R&D commitment, same two customers — only the shipment rate changed.

The pattern matters more than the numbers. Revenue is choppy. Gross profit tracks revenue almost one-for-one once unit cost normalises — gross-margin mix moves with absorption, not pricing. Operating income shows the violence: a single ₩9 bn revenue swing translates into a ₩7-8 bn operating-income swing because the fixed-cost base does not flex.

4. The Metrics That Actually Matter

Most readers will look at the P&L first. That is the wrong order for this business. The four numbers below predict the next twelve months better than any reported margin.

The scorecard reads as a single sentence: Auros has the right position on every dimension that compounds long-term value but the wrong magnitude on every dimension that drives short-term earnings.

5. What Is This Business Worth?

This is best valued as a single economic engine plus an option — not as sum-of-the-parts. There are no listed subsidiaries, no separable regulated assets, no holding-company stakes worth carving out. The recent equity investments in Konis (Korean development partner, Jan 2025) and Advanced Semiconductor Products (Singapore, Aug 2025) are immaterial and the US subsidiary is being wound down. The whole equity is one product family at two customers plus a multi-year bet to add two more product lines.

The right lens is therefore normalised cycle earnings on the base, plus an option-style premium for adjacency. At today's price (₩27,450, market cap ≈ ₩257 bn, EV ≈ ₩261 bn) the market is paying about 5× cycle-trough EV/Sales — roughly half of where the KLA-cluster peers trade and ~25% below where Park Systems trades despite Park Systems having far steadier earnings. The discount is real, but so is the reason for it.

The five-year band of operating margin is −16% to +10% on revenue of ₩35-61 bn. A central case of 5-7% op margin on ₩60 bn revenue produces ₩3-4 bn of op income — roughly ₩2-3 bn of mid-cycle net income to anchor to. This explains why peers' steady-state P/E of 25-40× cannot apply directly to a stock whose trough P/E is undefined.

6. What I'd Tell a Young Analyst

Read this stock as two stocks stapled together. The first is a small, defensible Korean monopoly in wafer-overlay metrology — high gross margins, quality customers, 14-year qualification moat, but capped revenue and lumpy quarters. The second is a four-year-old R&D project in thin-film and HBM-packaging metrology funded by burning the first business's earnings during trough quarters. The market multiple only makes sense if you believe both projects work. If you believe only the first, this is a ₩100-130 bn business trading above fair value; if you believe both, it is a ₩300+ bn business trading roughly fair.

Three things will tell you which world you are in within 18 months:

1. A second customer for thin-film. Pilot manufacturing has run since 2021. The filing window says no volume order has landed. First-volume HE-900 or METIS-FS shipment to SEC or SK Hynix is the only data point that can re-rate the option.
2. Samsung HE-900IR repeat orders. The November 2025 contract (₩6 bn, with up to ₩15 bn possible) is real but small. A second order — especially from SK Hynix — would convert "HBM optionality" from a press headline into a revenue line.
3. R&D-as-percent-of-sales coming down. Either revenue rises (cycle) or R&D pauses (it shouldn't). A sustained quarterly print below 22% would say scale has arrived. Anything above 30% for another year says R&D is funding diminishing returns.

Three things that would change the thesis decisively:

• KLA Archer share gain at SEC or SK Hynix. The single biggest risk to the base business. KLA's quarterly commentary on Korean memory wins is the clearest external signal. Auros's silence is not safety.
• Korean Value-Up payout commitment from FST. The 50.8% controlling block has been quiet; a public commitment to buybacks beyond the symbolic FY24 ₩1 bn programme would re-rate the stock independent of the operating story.
• A larger Korean specialist (Eugene Tech, Wonik IPS, or even Park Systems) acquiring Auros. The 33.54% FST stake is the only obstacle; the combination would solve Auros's scale problem in one transaction, and FST's chairman concurrently chairs Korea Lam Research, which is suggestive of strategic optionality.

One habit to avoid: do not value this on TTM earnings. The trough multiple is meaningless and the peak multiple flatters. Anchor to mid-cycle operating income and assess the option separately. If the option cannot be defended on evidence, the stock is too expensive on the base alone; if it can, the stock is interesting at any single-digit EV/Sales print of normalised revenue.