Teradyne's Record Quarter Meets a 35% Drawdown

Memory records, a dual-vendor land grab and the first HBM stack-height reversal in years — all while the tape de-rates the entire AI data center complex.

TER · Earnings Call · 2026-10-28

Teradyne printed the kind of quarter that is supposed to end a bear case. Revenue topped $1.3 billion, up 70% year over year; non-GAAP EPS of $2.29 rose more than 150%; and for the first time in recent memory all three business groups — Semi Test, Product Test and Robotics — grew against the prior year. Memory set a record at $301 million, the fourth consecutive quarter above $200 million. In-system test set another at $95 million. Product Test book-to-bill came in at 1.4x. AI-driven revenue, by Greg Smith's own accounting, now represents “more than 60% of total revenue” — Gregory S. Smith, Chief Executive Officer (CEO) · 2026-10-28. And yet the shares sit 25% lower over the last 90 trading days, 35% below the June 30 high of $483.84. The blowout is not the story. The disagreement is.

The most interesting sentence on the call

Tucked into the memory discussion — between HBM and NAND — was a throwaway that inverts a narrative the Street has carried for two years. Stack heights were supposed to march 8-high, 12-high, 16-high forever, because that is how you cram more bits into the same silicon real estate. Smith says that is over:

I think it is pretty locked in that next generation, there was a trend towards higher stack heights... And I feel like that has reversed in 2026
Gregory S. Smith, Chief Executive Officer (CEO) · 2026-10-28

The mechanics matter more than the headline. The binding constraint is the total number of DRAM wafers that can be produced; out of a fixed wafer supply you can build twice as many 4-high stacks as 8-high stacks. So the industry is de-constraining accelerator shipments by going wider, not taller. And because Teradyne's share in DRAM wafer sort is materially below its share in HBM performance test, more stacks per wafer means more of the test insertion where the company is strongest. This is the rare company-unique signal: lower stack heights and high stacks both appear in Teradyne's keyword set for the first time this quarter, as a matched and opposed pair. Nobody else reporting this week has framed HBM that way.

What management stopped saying

The bigger change is rhetorical. In July, asked about the ATE TAM, Smith said flatly: “we are not talking about a TAM publicly because we feel really uncertain about which of those time lines we're living in” — Gregory Smith, CEO · 2026-04-29. Three months later he is promising “an evergreen model against a much larger ATE TAM” — Gregory S. Smith, Chief Executive Officer (CEO) · 2026-10-28 in January, with test intensity already climbing from 6.4% of semi CapEx in 2025 to 8.1% across the first eight months of 2026. Compare that with February, when the same CEO was managing expectations down:

We expect that we're in kind of a 2-, 3-quarter surge that may lead to a shorter period of digestion afterwards.
Gregory Smith, Chief Executive Officer · 2026-02-03

There has been no digestion. Instead, Smith now declares that “the patterns that we saw up through 2024 are over” — Gregory S. Smith, Chief Executive Officer (CEO) · 2026-10-28 — a seismic shift, in his words, in a market twice the size that simply cannot be read off a mobile-cycle playbook anymore. The language of share has shifted too, and this is where the call gets interesting. Dual vendor is now the single most prominent keyword for Teradyne — four customers staged across initial qualification, fast follower and mature sourcing, with the next big capacity buy flagged for mid-2027. But the win is not translating into aggregate share, and Smith concedes it: “the size of the compute market has grown so much that, that has essentially had a dilutive effect on overall share” — Analyst · 2026-10-28. The company's own momentum history agrees — gain share was among the sharpest decliners in the current keyword set after several quarters as a top mover. Gross margin is the other quiet reversal. In July the CFO said: “we are not anticipating any fundamental changes in gross margin” — Michelle L. Turner, Chief Financial Officer · 2026-07-29 for 2027. In October, that 2027 margin picture became explicitly two-sided — volume and mix on one side, supply-chain inflation and capacity expansion on the other — with Q4 guided to 57.5–58.5%, roughly 100bp below the trailing profile. The guidance is fine. The honesty is new.

The tape disagrees

Look at what the market is paying for. At the last filing Teradyne carried a price-to-revenue multiple of 16.9x, up 235% year over year, and 104x free cash flow. The drawdown has comped some, but not much. The business, meanwhile, is generating real money — free cash flow of $358 million in the June quarter, up 212% — at an operating margin of 33.4%, north of the 2021 peak, with R&D at roughly 12% of revenue. So the multiple is not the whole explanation. The AI capex complex itself has cooled. AI data center was a monster 360-day advancer — 45 names up, 11 down — but over 90 days it flips to 6 up and 50 down, dragging AMAT, LRCX, AVGO, MRVL, AMKR and COHR with it, while High bandwidth memory gives back 7% across a dozen suppliers. Teradyne's drawdown is beta, not indictment. Set against that, receivables at 83.5% of revenue, up 17 points year over year, are a reminder that test cells arrive in pieces and revenue can slip across quarter boundaries — the lumpiness the CFO keeps flagging.

Robotics is a 2028 story, not a 2027 one

Physical AI is where the retail narrative lives — automation of distribution center workflows, tripled revenue with the anchor customer in 2026 — and Smith is careful not to feed it. Asked when it inflects beyond that anchor, he is blunt: “I think it is probably more of a 2028 thing than a 2027 thing beyond our sort of major customer in robotics.” — Gregory S. Smith, Chief Executive Officer (CEO) · 2026-10-28 Robotics crossed into profitability on $96 million of revenue this quarter, but it is a pipeline business with short visibility, not a counted-noses ATE market. What resolves the tension? Three things: the January model refresh, the first genuine multi-socket capacity buy from the merchant GPU fast follower in 2027, and whether the memory franchise keeps compounding through AI accelerator complexity. Management told you “we have pretty good line of sight out a couple of quarters” — Michelle L. Turner, Chief Financial Officer (CFO) · 2026-10-28 and that every quarter of 2027 grows sequentially. The tape is betting on the multiple. Teradyne just bet on the testers.