Onto’s EPS miss is the wrong fight: the real print is a packaging and GAA order inflection the Street did not pay for

The actionable read from this print is that Onto Innovation is being judged on the wrong quarter. What was priced in was essentially a clean Q4: Street revenue of $267.3 million and EPS of $1.28 left little room for a blemish, and ONTO delivered $266.9 million of revenue and $1.26 of EPS on that comparison basis, translating to surprises of -0.2% and -1.6%. That explains any initial disappointment. What actually surprised, and matters more for forward estimates, was not the reported Q4 line but the order and guide architecture around it: Michael Plisinski said, “We ended 2025 on a high note with orders from 2.5D packaging for AI devices more than doubling in the quarter, contributing to a record revenue of $267 million.” That wording matters because it ties the order acceleration directly to AI 2.5D packaging rather than a generic year-end shipment push, and because it frames the Q4 revenue base as the start of a demand step rather than a peak. The market may be mispricing a transition from a lumpy recovery stock to a process-control supplier with simultaneous exposure to 2.5D packaging, HBM4 DRAM, and gate-all-around, while still penalizing it for Q4 gross margin of 46.4% and diluted EPS of $0.21 in the company’s quarterly history.

The reason the miss is less important than it looks is that Q4 broke a three-quarter revenue deceleration while preserving a stronger backlog narrative than the income statement alone shows. Revenue had moved from $266.6 million in Q1 FY2025 to $253.6 million in Q2 FY2025 and $218.2 million in Q3 FY2025, before snapping back to $266.9 million in Q4 FY2025. The reported sequential growth of +22.3% did not just reverse the Q3 FY2025 decline of -14.0%; it restored revenue to the Q1 FY2025 level of $266.6 million and brought year-over-year growth back to +1.1% after Q3 FY2025 was -13.5%. The bear case is that this was a low-quality rebound because gross margin fell to 46.4%, the weakest gross margin in the table and below Q3 FY2025 at 50.7%. The counter is that the margin trough sits beside a management guide for Q1 revenue of $275 million to $285 million, operating margins of approximately 25.5% to 26.5%, and Q2 revenue exceeding $300 million. That combination says the company is absorbing mix and Semilab cost noise while demand is improving, not that demand is failing.

The capacity story explains why the Q2 guide matters more than the Q4 gross margin, because ONTO is telling investors the business can scale without a near-term physical bottleneck. Plisinski’s answer on capacity was unusually direct: “So we've said historically, our capacity, we're set up to be able to serve a $2 billion run rate.” That quote is worth using because it is a constraint statement, not a demand slogan. Against Q4 FY2025 revenue of $266.9 million and management’s Q2 expectation to surpass $300 million, the company is saying the next leg of revenue can flow through existing capacity rather than requiring a disruptive capex cycle. The historical revenue sequence supports that interpretation: Q1 FY2026 in the quarterly history shows $291.9 million of revenue, gross margin of 50.1%, revenue QoQ of +9.4%, revenue YoY of +9.5%, and diluted EPS of $0.67. Those figures do not erase the Q4 EPS problem, but they show the recovery path already re-approaching a 50.1% gross margin while revenue moves above the Q4 FY2025 level of $266.9 million. The variant perception is that a Q4 gross margin of 46.4% is the noisy trough investors can see, while the revenue slope toward $300 million is the more durable signal they may under-discount.

The mix evidence points to a narrower and more valuable cycle than “WFE is improving.” Brian Roberts said advanced packaging and specialty devices in the fourth quarter were approximately $145 million, slightly more than half of revenue, and that sales from 2.5D packaging doubled as compared to Q3. He also said approximately $9 million of Semilab revenue was included in that category, which matters because it prevents investors from treating every dollar of the Q4 packaging and specialty device increase as organic. For the full year, advanced packaging and specialty devices totaled $504 million of revenue, while advanced nodes more than doubled in 2025 to $308 million, with DRAM and logic together representing about 75% of that total. In Q4, advanced nodes grew sequentially by slightly over 30% to $72 million, primarily from pilot line sales tied to a new gate-all-around customer. The implication is that ONTO’s forward growth is not dependent on a single packaging pocket: it has 2.5D AI packaging, 3D bump metrology, GAA logic, and HBM4 DRAM all showing up in management’s narrative and product adoption commentary.

That mix is why the $240 million agreement is strategically more important than the Q4 EPS miss. Plisinski described the agreement as “valued at over $240 million, including over $60 million in systems for 3D bump metrology.” The wording is important because it separates total agreement value from the 3D bump metrology subset, which is the most direct read on advanced packaging complexity. A metrology supplier does not get pulled into 3D bump inspection because unit volumes are rising in a simple way; it gets pulled in because interconnect density, coplanarity, overlay, and yield windows are becoming harder. The company also said Atlas G6 is being adopted for new critical applications in both gate-all-around and HBM4 DRAM, which it expects will add to growth in 2026. The numerical anchor is the advanced nodes run-rate: $308 million in 2025, more than doubled, with Q4 at $72 million after slightly over 30% sequential growth. If the Street treats ONTO as a mid-cycle specialty tools name because full-year revenue finished at $1.005 billion, it risks missing that the growth vectors are coming from process steps that are becoming more measurement-intensive, not from a generic fab spending rebound.

The offset is real, and it is concentrated in power semi rather than hidden across the portfolio. Management said power semi revenue should decline around 10% in 2026, based on weakening demand for EVs and slowing infrastructure spending. That number is the main reason not to simply capitalize the AI packaging headlines at face value. But the same call also gives the offsetting scale: Q2 revenue is expected to surpass $300 million, resulting in 12% to 14% core growth in the first half of 2026 compared with the second half of 2025, per Roberts. The tension is therefore explicit. Power semi is a drag around 10%, while advanced packaging and advanced nodes are strong enough for a first-half core growth guide of 12% to 14%. That is the heart of the thesis: the market may be assigning too much signal value to the weakest end-market sentence and not enough to the quantitative bridge management gave for first-half revenue.

The read-through to customers is clearest for TSMC, Samsung, and Intel, because ONTO’s demand signals map to the exact process-control layers those manufacturers are trying to industrialize. For TSMC, the relevant read-through is process control metrology and inspection tied to AI 2.5D packaging, where ONTO said orders more than doubled in the quarter and Q4 advanced packaging and specialty devices were approximately $145 million. For Samsung, the relevant signal is metrology and inspection for GAA nanosheet process, with ONTO’s advanced nodes revenue growing sequentially by slightly over 30% to $72 million in Q4 and pilot line sales tied to a new gate-all-around customer. For Intel, the same $72 million advanced nodes figure and the $308 million full-year advanced nodes revenue base matter because advanced logic process control intensity rises as GAA adoption broadens. There are no named suppliers to ONTO in the data pack, so the supply-chain conclusion should stay on customer capex pull: this print is a positive demand read for customers spending on 2.5D packaging, HBM4 DRAM, and GAA process control, not a component supplier read-through.

The peer comparison reinforces that ONTO is not the highest-margin process-control asset, but it is no longer screening like a laggard on growth. In the latest reported quarter, ONTO’s $291.9 million of revenue, 50.1% gross margin, and +9.5% revenue YoY sit below KLAC’s $3,415.1 million of revenue, 61.1% gross margin, and +11.5% revenue YoY, and below NVMI’s 57.7% gross margin and +10.3% revenue YoY on $235.3 million of revenue. That relative gap matters because ONTO cannot be valued as if it already has KLAC-like margin structure. But compared with BRKR at $823.4 million of revenue, 47.3% gross margin, and +2.7% revenue YoY, ONTO’s 50.1% gross margin and +9.5% revenue YoY show a cleaner growth-margin combination within the broader process-control table. The comparative point is not that ONTO deserves the best multiple in the group; it is that a company guiding Q2 revenue above $300 million while recovering to 50.1% gross margin in Q1 FY2026 should not be valued solely off Q4’s 46.4% gross margin trough.

The call delivery supports the idea that management’s confidence improved, but the split between prepared remarks and Q&A keeps the thesis from being purely tone-driven. The tone history shows sentiment moving from 0.42 in Q4 FY2025 to 0.47 in Q1 FY2026, guidance_tone from 0.47 to 0.54, tone_confidence from 0.29 to 0.37, and uncertainty down from 90.9 to 72.4. That is a better delivery profile, especially because qa_evasiveness moved from 16.9 to -10.5, a call-over-call delta of -27.4. The caution is that prepared_sentiment was 0.80 in Q1 FY2026 while qa_sentiment was 0.30, so management sounded much more positive in scripted remarks than in analyst exchange. That split is not fatal, because qa_sentiment improved by +0.15 and uncertainty fell by -18.5, but it says investors should require the Q2 revenue number to materialize rather than buying the tone alone.

The cash and capital structure details also matter because they frame whether the Semilab acquisition is an earnings drag or a revenue bridge into a larger process-control platform. Roberts said ONTO generated $95 million of cash in the quarter, with cash conversion of approximately 150% of non-GAAP net income. He also said the new Tax Act produced cash tax savings of $19 million in 2025 and an additional estimated $14 million in cash savings in 2026. Against that, the Semilab close on November 17 required $445 million in cash and 641,771 shares of common stock, and Q1 operating expenses should approximate $80 million as the company realizes a full quarter of Semilab costs. This is where the Q4 EPS miss and the Q1 guide have to be held together: Semilab raises the operating expense base, and approximately $9 million of Q4 revenue related to the acquisition is already embedded in advanced packaging and specialty devices, but management also pointed to low $100 million to $110 million of revenue for 2026 in response to the organic-growth framing. The integration question is not whether Semilab adds revenue; it is whether ONTO can convert that added cost base into the higher-margin metrology opportunities implied by 3D bump, HBM4 DRAM, and GAA.

That leaves the investment debate with concrete markers rather than a vague AI story. The confirmation path is straightforward: Q1 revenue must land inside the $275 million to $285 million guide, Q1 operating margin must be approximately 25.5% to 26.5%, and Q1 EPS must fall in the $1.26 to $1.36 range using the company’s assumptions of approximately 16% tax rate and about 49.9 million shares outstanding. Then Q2 revenue needs to surpass $300 million, because that is the management commitment that underwrites 12% to 14% core growth in the first half of 2026 versus the second half of 2025. The break points are equally clear: if gross margin fails to hold near the Q1 FY2026 historical 50.1% level after Q4 FY2025’s 46.4%, if power semi weakness around 10% spreads into advanced packaging and advanced nodes, or if the over $240 million agreement does not translate into visible 3D bump metrology momentum beyond the over $60 million subset, the thesis weakens. Until those numbers break, the print should be read as a forward inflection masked by a Q4 Street miss, not as an earnings-quality warning.