Semiconductors
PrivateA near-monopoly-tier process-control niche (metrology + advanced-packaging inspection) re-accelerating off a FY25 trough into a guided >30% FY26 — but the tape already pays ~47x forward non-GAAP / 12x sales for it, and the out-of-character $710M debt-funded Rigaku stake is the first capital-allocation question mark in a decade.
Research
The verdict
A near-monopoly-tier process-control niche (metrology + advanced-packaging inspection) re-accelerating off a FY25 trough into a guided >30% FY26 — but the tape already pays ~47x forward non-GAAP / 12x sales for it, and the out-of-character $710M debt-funded Rigaku stake is the first capital-allocation question mark in a decade.
Primary sources
Source documents — open to read in full
Onto Innovation is a pure-play semiconductor process-control company — it designs, builds and supports the metrology and inspection tools that fabs and packaging houses use to measure and police what they are building, plus the analytical software that turns that measurement data into yield decisions. It does not make the wafers or the chips; it makes the instruments that tell chipmakers whether their multi-billion-dollar lines are producing good die.
Product families (all ``):
Revenue model: ~84% systems & software, ~8% parts, ~8% services in FY25. This is a capital-equipment cadence, not recurring SaaS: metrology sales are lumpy and front-loaded to each new node ramp; advanced-packaging inspection is more volume-driven (tied to assembly throughput, not node transitions). >190 customers across 25+ countries in 2025. Warranty 12–14 months; direct sales/service in US, Korea, Japan, Taiwan, Malaysia, China, Vietnam, Singapore, Europe.
End-markets: logic (incl. gate-all-around), memory (DRAM/HBM, 3D-NAND with 700+ layers), advanced packaging (CoWoS, HBM stacks, 2.5D/3D, panel), plus specialty (SiC/GaN power, CIS, MEMS, RF). The two structural growth vectors are advanced packaging for AI and advanced-node metrology (GAA + HBM).
Upstream (inputs into Onto): precision optics and lasers are the critical, long-lead inputs — the 10-K states lead times for qualifying a new supplier of lasers and certain optics can be ~1 year, other components ~9 months, and some component shipments have >6-month lead times. A significant number of components/subassemblies are sole- or single-sourced, and Onto does not hold long-term contracts with many suppliers — a genuine chokepoint on its own ramp. Onto outsources non-differentiating assemblies to contract manufacturers and keeps differentiating modules in-house. The filing does not name individual suppliers (the lens stays partly generic here because the disclosure is generic — a known gap; named optics/laser suppliers were not sourced this run → n/a).
Onto's own manufacturing: Milpitas CA, Tucson AZ, Wilmington MA (HQ), Bloomington MN, plus contract manufacturers worldwide. Substantially all long-lived assets are in the US.
Downstream (Onto → end customer): Onto sells directly to the chipmakers and packaging houses. The named end-buyers are inferable from geography and the customer concentration: leading-edge logic foundries (Taiwan), memory makers (Korea — DRAM/HBM), OSAT/advanced-packaging houses, and a rising US/China cohort. Top-3 customers = 49% of FY25 revenue (see Lens 1/Lens 4). Onto's integrated-metrology (IMPULSE) systems bolt onto other WFE vendors' process tools — so a slice of revenue depends on third-party equipment suppliers' sales and their goodwill (they could build competing IM; Nova is the named competitor there).
Chokepoint summary: the binding constraints are (1) inbound optics/lasers (single-source, year-long requalification) and (2) customer concentration on a handful of leading-edge fabs. Onto is itself a chokepoint supplier into the AI buildout — see Lens 3.
Onto sits in a structurally attractive corner of WFE: process control is an oligopoly, and Onto is the clear #2–3 in several sub-niches behind KLA.
Onto reports as ONE business segment — so there is no segment-level EBIT disclosure. It does disaggregate revenue by source and geography, and management narrates end-market splits on calls.
By source (FY25, FY24, FY23):
| Source | FY25 | FY24 | FY23 |
|---|---|---|---|
| Systems & software | $847.8M (84%) | $850.4M (86%) | $683.3M (84%) |
| Parts | $84.2M (8%) | $76.6M (8%) | $74.6M (9%) |
| Services | $73.2M (8%) | $60.3M (6%) | $57.9M (7%) |
| Total | $1,005.3M | $987.3M | $815.9M |
Systems & software was flat FY25 (−$2.6M) — the trough — as AI-packaging inspection units dipped and were offset by advanced-node metrology + Semilab SiC. Parts/services grew (installed-base monetization).
By geography (% of revenue):
| Region | FY25 | FY24 | FY23 |
|---|---|---|---|
| Taiwan | 32% | 31% | 17% |
| South Korea | 28% | 29% | 21% |
| United States | 12% | 11% | 16% |
| Japan | 10% | 6% | 11% |
| China | 7% | 12% | 17% |
| SE Asia | 6% | 6% | 11% |
| Europe | 5% | 5% | 7% |
China collapsed from 17% (FY23) → 7% (FY25) — directly attributed to US export controls. Asia is the majority of revenue and management expects that to persist.
End-market trends (management, calls): advanced nodes (GAA + HBM4 DRAM via Atlas G6) more than doubled to ~$308M in FY25; advanced-packaging base ~$500M (CoWoS, HBM inspection, OSAT, panel); GAA + Iris films ~$100M in FY24. For FY26 management guides advanced packaging +50%, advanced nodes ~+25%, power semis −10% (weak EV).
Q1-2026 source mix: Systems&SW $247.2M (84.7%), Parts $26.6M (9.1%), Services $18.2M (6.2%) of $291.9M; Semilab contributed $24.0M.
The headline GAAP print looks ugly and is widely misread; the underlying story is a trough-to-acceleration inflection.
GAAP, Q1'26 vs Q1'25:
Vs consensus: revenue $291.9M came in light vs ~$297.9M Street; non-GAAP EPS $1.42 beat ~$1.41. Initial after-hours reaction was −6.4% on the revenue miss + a Q2 outlook some read as soft, then the stock recovered to record highs after the Rigaku deal landed.
Guidance (the real news): Q2'26 revenue $320–330M (~+28% YoY) and a full-year framework of >$1.3B revenue, >30% annual growth, >30% operating margin by Q4. That is a dramatic step-up from FY25's +2%, and it reframes FY25 as the cyclical bottom.
Balance-sheet flags (Q1'26): operating cash flow fell to $26.3M vs $92.0M YoY on a −$41.6M working-capital swing — inventory $298M→$316M and receivables $269M→$307M both built while revenue was only modestly up (watch this; see Lens 10). Cash+securities $654.2M, still effectively debt-free (total liabilities $264M, all operating). Buybacks paused in Q1'26 ($99.9M left on the $200M auth) — consistent with conserving dry powder for Rigaku.
No transcripts on the on-disk shelf (transcripts/ empty), so this lens is ``-grounded and lighter than ideal — flag: backfill FY25Q4 → FY26Q1 transcripts on next refresh.
Tone arc across the last ~4 calls:
Net: sentiment has shifted from "digesting a boom" to "next leg up," with management willing to commit to specific, falsifiable FY26 targets — which is itself a confidence tell (and a risk if they miss).
Peer set = the process-control / metrology oligopoly + adjacent WFE. Multiples are `` with date, or n/a. None are fabricated.
| Company | Ticker | Mkt cap | Fwd P/E | EV/EBITDA | P/E (TTM) | Notes |
|---|---|---|---|---|---|---|
| Onto Innovation | ONTO | ~$14–16B `` | ~47x `` | n/a | 77x ; 129–151x on other bases | Net cash pre-Rigaku |
| KLA | KLAC | n/a | n/a | ~38–41x `` | ~41–51x `` | Scale leader, dividend payer |
| Nova | NVMI | n/a | ~43–52x `` | n/a | n/a | Direct OCD/IM metrology rival; px ~$609; Q1 rev $235M |
| Camtek | CAMT | n/a | n/a | ~27x `` | ~31x `` | AP-inspection rival; ~9x EV/rev |
Derived ONTO multiples ``: shares 49.74M (10-Q); cash+sec $654.2M; EV ≈ $15.5B ($16.1B mktcap − $0.65B net cash); EV/Sales ≈ 15.0x on TTM ~$1.03B revenue, ≈11.9x on FY26 guide ≥$1.30B.
Read: Onto is not cheap on any lens. ~47x forward non-GAAP is roughly in line with Nova (~43–52x) and a touch above KLA (~38–41x EV/EBITDA), and richer than Camtek (~27x). The eye-popping 77–151x trailing GAAP P/E is an artifact of acquisition amortization depressing GAAP EPS — real, but not the number the market is paying on. The group as a whole is priced for the AI-WFE supercycle to keep compounding.
Mostly ``. The pattern over the last ~2 years:
Implication: to underwrite ONTO you are underwriting (a) the AI-packaging + GAA capex cycle and (b) management hitting its own aggressive FY26 guide. Earnings reactions hinge on forward commentary and AP/advanced-node growth rates, not the GAAP EPS line.
Forensic lens. Every figure labeled. Onto's accounting reads clean; the watch-items are working capital and acquisition optics.
Regulatory findings (required sub-section):
Built bottom-up from the latest actuals + management's own framework. Onto reports non-GAAP EPS as the Street number; I model non-GAAP and footnote the GAAP gap. No forecast.ts logged — --watchlist mode (skip per skill).
Anchors: FY25 revenue $1,005.3M; mgmt FY26 guide >$1.3B, >30% growth, >30% op-margin exiting Q4; Q2'26 guide $320–330M (~+28%); Street FY26 non-GAAP EPS ~$6.72–7.16; ~49.7M diluted shares, minimal dilution; net cash (pre-Rigaku).
| Scenario | FY26 rev | FY26 op-margin | FY26 non-GAAP EPS | FY27 EPS | FY28 EPS | Logic |
|---|---|---|---|---|---|---|
| Bull | ~$1.40B (+39%) | ~30% exit | ~$7.50 | ~$9.25 | ~$11.00 | AP +50% & advanced-nodes +25% both deliver; GM normalizes to ~53% as step-up rolls off; opex leverage. |
| Base | ~$1.32B (+31%) | ~28% | ~$6.90 | ~$8.20 | ~$9.50 | Hits the >$1.3B guide; GM ~52%; Rigaku slightly accretive late-FY26. Matches Street ~$6.7–7.2. |
| Bear | ~$1.15B (+14%) | ~24% | ~$5.40 | ~$5.75 | ~$6.25 | AI-packaging digestion / a key customer pushes out; China stays capped; more inventory write-downs; GM stuck ~49%. |
Arithmetic (base): FY26 rev $1.32B × ~28% op-margin ≈ $370M op-income; + ~$15M net interest (lower post-Semilab/Rigaku cash); − 14% tax; ÷ $1.50–2.00/sh lower** on the ~$80M/yr acquisition-amortization drag (growing with Rigaku) — i.e. GAAP FY26 ~$5.0–5.4.49.8M sh ≈ **$6.8–6.9 non-GAAP EPS**. GAAP EPS runs **
Brier forecast (not logged this run; candidate for a refresh): ONTO FY2026 non-GAAP EPS ≥ $6.50, p ≈ 0.70, resolves 2026-12-31.
Bull case. Onto is a picks-and-shovels chokepoint on the two highest-conviction semis vectors — AI advanced packaging (HBM, CoWoS, 2.5D logic; mgmt guides +50% FY26) and advanced-node metrology (GAA + HBM4 DRAM via Atlas G6; +25% FY26). FY25 was a cyclical trough (+2%); FY26 is a guided >30% reacceleration to >$1.3B with >30% exit op-margins — operating leverage on a largely fixed cost base. The moat is qualification/switching lock-in in an oligopoly where the only bigger player (KLA) can't be everywhere. Net-cash balance sheet, disciplined operator with a proven M&A record, ~$100M buyback dry powder, and Dragonfly G5 / Atlas G6 design wins de-risking the growth. Process control's content per wafer rises every node — a secular tailwind independent of unit volumes. If the guide lands, the stock is ~47x→ mid-30s forward as EPS catches up.
Bear case (permanent-impairment risks). (1) Customer + end-market concentration — top-3 = 49%, and the growth is levered to a handful of leading-edge AI-packaging and GAA programs; a single HBM/CoWoS push-out or a customer's in-sourcing of inspection guts the FY26 guide. (2) It's a #2/#3, not the franchise — KLA, Nova and Camtek are all well-capitalized and gunning for the same niches; share is contestable and pricing power is limited (the 10-K admits switching costs cut against Onto winning new sockets too). (3) The Rigaku pivot — $710M, 27%, non-controlling, funded by a $500M bridge loan, on a company that has been debt-free for a decade. A minority stake you can't consolidate, at ~$710M for 27% of a Japanese X-ray firm, is a strategic option, not an operating asset — and it converts the pristine balance sheet into net debt for an unproven payoff. Pre-mortem (18 months out, thesis broke): AI-packaging orders proved pull-forward, FY26 came in ~+12% not +30%, GM never reclaimed 52% as inventory written down again, the Rigaku bridge had to be termed out into real debt with no visible synergy, and a 47x multiple compressed to ~25x → the stock halves even though the business is "fine." Are multiples too high? Yes on any historical lens (10-yr median P/E ~30 vs ~47x fwd / 77–151x trailing GAAP). The price already embeds the guide working. There is little margin of safety if FY26 merely meets rather than beats. Contrarian view (what the market refuses to see): the bull crowd treats the >30% FY26 guide and AP +50% as near-certain and the Rigaku deal as obviously smart; the under-appreciated risk is that the same guide that is the bull case is also the bear trigger — management has staked credibility on specific numbers in a notoriously lumpy capex business, and the Rigaku debt removes the balance-sheet shock-absorber right as the cycle bet is placed.
Dismantling the bull case.
A 3-engine specialty-hardware roll-up wearing an "AI factory" costume — the AI-systems story (Advanced Computing) is the lowest-margin, most lumpy, most hyperscaler-concentrated leg, and the actual FY26 EPS beat is being driven by a cyclical memory (DRAM/Flash) price spike that the bulls are extrapolating as if it were the AI thesis; own the re-rating only if you trust the Shaikh-led non-hyperscaler pivot to convert before the memory cycle rolls.
The pure-play AOI/metrology pick on the HBM-and-chiplet inspection supercycle — >40% HBM-inspection share and 50% of revenue now AI-driven — but a ~50x forward multiple already prices the boom while 49% China revenue sits under a tightening export-control gun.
A best-in-class analog compounder mid-way through a violent cyclical recovery — the business is pristine, the cycle is real, but at ~35x forward / ~65x trailing the tape has already paid for the upturn; the edge is in the next destock, not at today's price.