Cognex: The Record Quarter the Tape Refused to Believe

A record quarter into a 33% drawdown

Cognex is an $8 billion machine-vision company, not an AI-infrastructure pure play, but the tape has stopped making that distinction. The stock is down 33.4% over the last 90 trading days, one clean down segment from the Aug. 4 peak of $71.52 to a 33.6% drawdown. That is the backdrop for a Q3 report that was, by the numbers, the best the company has produced in years: record revenue of $313 million, up 19%; adjusted EBITDA of $111 million, up 91%; a 35.5% EBITDA margin; and record adjusted EPS of $0.53, up 89%. “Revenue reached a record $313 million, increasing 19% year-over-year. This marked our ninth consecutive quarter of year-over-year revenue growth.” — Dennis Fehr, CFO · 2026-11-05 Management also raised full-year revenue to $1.14 billion-$1.16 billion and adjusted EPS to $1.68-$1.72. The disconnect is the story.

Q3 was a strong quarter for Cognex, with record revenue, a significant increase in profitability, and continued execution against our strategic objectives. Our results reflect broad-based demand strength...

What actually changed: breadth, not an AI-data-center kicker

The important shift is not that Cognex became an AI story. It is that growth became less dependent on any one story. Four of five end markets grew double digits in constant currency. Logistics delivered its 11th consecutive quarter of double-digit growth; packaging and electronics grew double digits; semiconductor grew strong double digits from a small base; automotive turned mid-single-digit positive and is stabilizing, with Europe still the weak spot. That is a factory automation recovery, powered by labor scarcity, direct material cost inflation and the need to automate quality — the same secular drivers management has cited for years, now showing up in the P&L.

The new product cycle is real but narrow: the In-Sight 1750 wafer and panel ID reader extends Cognex into semiconductor front-end and back-end inspection, a market still only about 10% of 2025 revenue. The second-ranked Q3 keyword was investment in AI, but the context is edge AI on the factory floor — AI Autotune, simpler deployment, complex code reading — not AI server racks. That distinction matters for how the company should be valued.

The keyword tell: data center faded, factory automation stepped forward

A quarter ago, data center was a top-two Cognex keyword. Management sized it as low single digits of revenue growing about 30%, tied to AI-factory build-outs. “We've been serving the data center market for several years, but it was always a smaller portion of our business... What's changed, obviously, is the very aggressive build-out of new facilities...” — Matt Moschner, CEO · 2026-08-06 In the Q3 call, data center dropped out of the top keyword set. It was mentioned only as one strand of electronics demand, not as a separate growth pillar. That is not a demand collapse; it is a narrative correction. The company is telling investors that the durable story is diversified factory automation, with data center as an adjacent kicker, not the main engine.

The tape, however, has been treating anything AI-adjacent as one trade. Globally, AI data centers fell about 31% over 90 days with 57 decliners, and high bandwidth memory fell roughly 37%. Cognex's electronics end market touches that supply chain — data center supply chain investment was called out as a driver — and its pending RealSense acquisition adds robotic perception. So the stock got swept into the same de-rating, even though its own results were accelerating.

Margins: the fear that didn't materialize

The cleanest evidence that execution is ahead of the narrative is gross margin. In the prior call, management warned about memory price pressure: “we would now say about 75 basis points of gross margin headwind being included in our Q3 guide.” — Dennis Fehr, CFO · 2026-08-06 Q3 gross margin instead expanded 360 basis points to 71.3%, with pricing and mix more than offsetting memory costs. Operating expenses fell 7% year over year, and roughly $35 million of annualized net cost reductions remain on track. This is operating leverage, not accounting. The latest 10-Q already showed operating margin at 30.7%, up 11.4 percentage points year over year, and gross margin at 70.6%. Q3 pushed both higher.

Dennis Fehr was explicit that this is not the end: “we are not at the end of our margin expansion story. There is still clearly more opportunity for us.” — Dennis Fehr, CFO · 2026-11-05 The 2027 framework is not yet formal, but the intent is to grow revenue with existing resources, letting R&D intensity run near 11% of revenue while revenue scales.

Valuation, not operations, is the problem

At the August peak, the market was paying 11.1x trailing revenue, up 93% year over year, and 44.9x trailing free cash flow. That is a valuation for a durable compounder with a new AI growth leg. The Q3 results support the earnings leg; they do not resolve the durability debate. Management declined to guide 2027, logistics faces tough comps, automotive Europe is still shrinking, and semiconductor is small. RealSense adds a new pillar but also integration work; the CFO said the near-term priority is “digesting the RealSense acquisitions” — Dennis Fehr, CFO · 2026-11-05 rather than deploying more capital.

So the stock is cheaper than it was, but not obviously cheap. The bull case is that Cognex is being mislabeled as an AI-infrastructure casualty when it is actually a broad-based factory-automation recovery with record margins and a new product cycle. The bear case is that the cycle is late, logistics growth must moderate, and the multiple still discounts a lot. The next proof points are Q4 — guided to $270 million-$290 million, about 11% growth at the midpoint — and the June 2027 Investor Day. Until then, the tape and the transcript are telling different stories.