Teradyne’s AI test cycle is being underpriced as a 2025 exit-rate anomaly

The actionable read from this print is that investors are still treating Teradyne as a semiconductor capital equipment recovery stock when the numbers say it has become an AI infrastructure bottleneck supplier for test. What was priced in was a good Q4: the Street estimate already sat at $977.2 million of revenue and $1.38 of EPS, both above the run-rate implied by the softer first half of 2025. What actually surprised was the size and composition of the beat, with revenue at $1,083.3 million for a +10.9% surprise and EPS at $1.80 for a +30.4% surprise, followed by Q1 sales guidance of $1.15 billion to $1.25 billion and a 2026 model of about $6.0 billion of revenue. The variant perception is that this was not simply backlog conversion into a quarter that happened to be good; the company is telling investors that AI test intensity is pulling the whole model upward, and the operating leverage is now visible enough that the debate should shift to durability of AI-driven Semi Test demand rather than whether Teradyne has reached a cyclical peak.

That distinction matters because the revenue trajectory had already improved before Q4, but not in a way that forced a new multiple debate. Revenue moved from $685.7 million in Q1 FY2025 to $651.8 million in Q2 FY2025, then $769.2 million in Q3 FY2025, before jumping to $1,083.3 million in Q4 FY2025. The market could dismiss Q3’s +18.0% sequential growth as order timing after a -4.9% Q2, but Q4’s +40.8% sequential growth and +43.9% year-on-year growth breaks that explanation. The gross margin was not a clean upward line, at 60.6% in Q1 FY2025, 57.2% in Q2 FY2025, 58.4% in Q3 FY2025, and 57.5% in Q4 FY2025, so the print was not a simple story of mix lifting every financial metric at once. That is precisely why the stock’s setup is interesting: the top-line acceleration is already in the reported numbers, while the gross margin inflection is more a Q1 and 2026 argument than a Q4 proof point.

The margin story becomes more investable when Q4 is read as the volume bridge rather than the destination. Michelle Turner framed Q4 in company-accounting terms as “Fourth quarter sales were $1.083 billion with non-GAAP EPS of $1.80, both above the high end of our guidance range,” which matters because the company’s own reported basis aligns with the Street-comparison revenue of $1,083.3 million while EPS differs from the diluted EPS of $1.63 in quarterly history. The correct interpretation is not to blend those EPS figures, but to recognize that the non-GAAP earnings beat came despite a Q4 gross margin of 57.5%, below Q1 FY2025’s 60.6% and below Q4 FY2024’s 59.4%. Management then guided Q1 non-GAAP gross margin to 58.5% to 59.5%, with Turner specifying that this is “up 180 basis points at the midpoint of the guidance quarter-over-quarter.” That wording is useful because it commits to a margin recovery immediately after the volume breakout, rather than asking investors to wait for a second-half operating leverage story.

The segment mix explains why the market’s “great quarter, maybe peak” interpretation is too cautious. Semi Test revenue was $883 million in Q4, with SoC revenue of $647 million up 47% quarter-on-quarter and memory revenue of $206 million up 61% quarter-on-quarter, while Product Test Group at $110 million grew double digits sequentially and year-on-year and Robotics revenue of $89 million was up 19% from Q3. The company’s own language ties Semi Test to AI compute and memory demand, and those are the parts of the portfolio where the customer read-through is most relevant. The market was prepared for a tester supplier to benefit from AI accelerators, but the surprise is that both SoC and memory were accelerating at the same time, with memory marking a record sales quarter. That combination reduces the risk that Q4 was a one-product digestion event; it implies simultaneous test demand from AI compute devices and the memory content around them.

That customer read-through is the crux for the rest of the semiconductor complex. For TSMC, Samsung, ASE Group, Intel, and Amkor, Teradyne’s Q4 Semi Test revenue of $883 million is a direct signal that ATE demand for UltraFlex SoC, Magnum memory, J750 analog, and analog/mixed-signal test is being pulled by capacity plans tied to AI compute and memory. The magnitude is not a vague positive: SoC revenue was $647 million and up 47% quarter-on-quarter, while memory revenue was $206 million and up 61% quarter-on-quarter. For TSMC and Samsung, which are listed customers for ATE across UltraFlex SoC and Magnum memory, the print supports the view that advanced compute and memory test steps are scaling together. For ASE Group, Intel, and Amkor, the read-through is that backend and SoC test demand is not waiting for a broad-based semiconductor recovery, because Teradyne’s SoC test revenue already reached $647 million in Q4. On the supplier side, Leeno Industrial’s pogo pins and Yamaichi Electronics IC test sockets and connectors sit directly in the consumables and interface path of this ATE demand; the peer table shows Yamaichi at ¥13,103.0 million of revenue, 38.0% gross margin, and +43.0% revenue YoY, which is directionally consistent with Teradyne’s +43.9% revenue YoY in Q4 without needing to assume identical end-market exposure.

The competitive comparison also argues that Teradyne is participating in a subsector-wide test upcycle, but with a differentiated AI lever. In the Test_Assembly peer set, ATEYY reported ¥334,100.1 million of revenue, 67.4% gross margin, and +43.8% revenue YoY, while DSCSY reported ¥135,505.5 million of revenue, 70.8% gross margin, and +12.3% revenue YoY. Teradyne’s Q4 revenue YoY growth of +43.9% is essentially in line with ATEYY’s +43.8%, but Teradyne’s Q4 gross margin of 57.5% is below ATEYY’s 67.4% and DSCSY’s 70.8%. That gap is not a reason to fade the print; it is the operating leverage opportunity if AI demand remains mix-favorable and if Q1 gross margin reaches the 58.5% to 59.5% range. The peer data says the broader test complex is not uniformly accelerating, with 6125.T at -3.7% revenue YoY and 6140.T at +11.2%, so Teradyne’s +43.9% is not merely a rising-tide number. It places Teradyne in the high-growth cohort while still leaving margin headroom relative to the highest-margin peers in the table.

The call delivery supports the same conclusion, with one important nuance: management sounded more concrete in prepared remarks, while Q&A sentiment cooled. The tone history shows Q4 FY2025 sentiment at 0.33 versus Q3 FY2025 at 0.27, guidance_tone at 0.36 versus 0.42, tone_confidence at 0.40 versus 0.37, prepared_sentiment at 0.35 versus 0.52, qa_sentiment at 0.33 versus 0.11, ai_optimism at 0.46 versus 0.37, uncertainty at 83.6 versus 74.9, and qa_evasiveness at -25.4 versus 36.5. The Q1 FY2026 call-over-call delta versus Q4 FY2025 then shows sentiment -0.00, guidance_tone -0.04, tone_confidence +0.03, prepared_sentiment +0.33, qa_sentiment -0.17, ai_optimism -0.01, uncertainty -28.2, and qa_evasiveness +24.2. The useful signal is not that tone was uniformly better, because it was not: guidance_tone and qa_sentiment moved lower in the delta. The useful signal is that uncertainty fell by -28.2 while prepared_sentiment rose by +0.33, which matches a company moving from explaining a sudden demand surge to codifying it in targets.

That tone pattern matters because the 2026 framework is aggressive enough that delivery language is part of the evidence. Turner said, “While 2025 sales were 40% in the first half and 60% in the second half, based on what we know today, we expect 2026 sales to be in the inverse.” The phrase “based on what we know today” is the hedge, and investors should respect it, but the commitment is still significant because Q1 sales are guided to $1.15 billion to $1.25 billion and the company is framing first-half weighting as the opposite of 2025. Gregory Smith added the clearest end-market exposure point: “Looking forward to Q1 of 2026, we expect that upwards of 70% of our revenue will be driven by AI applications.” That is the sentence the market may be underweighting. If upwards of 70% of Q1 revenue is AI-linked on a $1.15 billion to $1.25 billion guide, then the question is not whether Teradyne has AI exposure; the question is whether customers are over-ordering test capacity or whether AI device complexity has structurally raised test intensity.

The earnings power evidence tilts toward structural uplift, although not without a margin caveat. Q4 non-GAAP operating expenses were $306 million, the non-GAAP operating profit rate was 29%, free cash flow was $219 million, and shareholder returns were $204 million through share repurchases and dividends. For the full year, revenue was $3.2 billion, up 13% from prior year, gross margin was 58.3%, OpEx was $1.2 billion, operating profit was 22%, free cash flow was $450 million, and shareholder returns were $785 million or 174% of free cash flow. That capital return number tells PMs that management is not hoarding cash for a speculative capacity build, but the more relevant point is the 2026 model: at an ATE TAM of $12 billion to $14 billion, the target model assumes roughly $6 billion of revenue, 30% to 34% operating profit, and non-GAAP EPS of $9.50 to $11.00. The caveat is that Q4 gross margin of 57.5% does not yet prove a 30% to 34% operating profit model by itself; the confirmation has to come from Q1 gross margin at 58.5% to 59.5% and OpEx contained at approximately 26% to 28% of first quarter sales.

The pricing question, then, is less about whether Q4 was good and more about whether estimates and multiples can absorb a revenue base that is being re-anchored. The quarterly history shows diluted EPS of $0.61 in Q1 FY2025, $0.49 in Q2 FY2025, $0.75 in Q3 FY2025, and $1.63 in Q4 FY2025, followed by $2.53 in Q1 FY2026. Even without calculating a run rate, the sequence shows that the earnings base changed abruptly once revenue moved above $1,083.3 million and gross margin moved to 60.9% in Q1 FY2026. The market may be reluctant to capitalize that because AI capex beneficiaries have already delivered several upside cycles, but Teradyne’s specific surprise was not just demand. It was the combination of +10.9% revenue surprise, +30.4% EPS surprise, Q1 sales guidance of $1.15 billion to $1.25 billion, Q1 EPS guidance of $1.89 to $2.25 on 158 million diluted shares, and a 2026 EPS target of $9.50 to $11.00. A PM fading the stock after the print has to believe either the Q1 guide is an air pocket before digestion or the $6.0 billion revenue framework embeds unsustainably high ATE TAM assumptions. The data pack supports the opposite view unless AI application revenue falls materially below the “upwards of 70%” Q1 marker.

What to watch next is precise. The thesis is confirmed if Q1 sales land within or above $1.15 billion to $1.25 billion, if non-GAAP gross margin lands within or above 58.5% to 59.5%, and if non-GAAP EPS lands within or above $1.89 to $2.25 on 158 million diluted shares. The thesis is strengthened if Q1 AI applications remain at or above the “upwards of 70%” revenue level, because that would validate the idea that Q4’s $883 million Semi Test revenue, $647 million SoC revenue, and $206 million memory revenue were not a one-quarter pull-forward. The thesis breaks if Q1 revenue falls below $1.15 billion, if gross margin fails to reach 58.5%, or if management walks back the about $6.0 billion 2026 revenue model, the 30% to 34% operating profit range, or the $9.50 to $11.00 non-GAAP EPS range. The next call should also be judged against the tone markers: a reversal in uncertainty from 55.4 back toward 83.6, or another decline in qa_sentiment from 0.16 after the -0.17 delta, would be an early warning that the attractive prepared framework is getting harder to defend under investor questioning.