Teradyne has just delivered the second-best quarter in its sixty-five-year history. Revenue doubled year-on-year, the balance sheet carries no debt, and return on equity is 29%. Nothing in the business is broken. But gross margin is now guided down rather than up, the higher operating expense base is structural, and management’s own most bullish public forecast, discounted, supports about $192 a share. The stock is at $325.
The author holds no position in TER as at the date of publication, and held none while the research was produced. No position will be taken for 72 hours after publication.
The author has received no compensation from Teradyne, Inc. or any affiliate, holds no position exceeding 0.5% of the issuer’s share capital, and has no business relationship with the company. This report was not shown to the issuer before publication.
| Dimension | Score | One line |
|---|---|---|
| Business quality | 8.5 / 10 | Duopoly in automated test, brutal switching costs, zero debt. |
| Financial strength | 9.0 / 10 | Net cash, 29% return on equity, thirteen straight years of dividends, no leverage. |
| Management quality | 4.0 / 5 | Credible operators, disciplined on buybacks, unusually forthcoming on long-range numbers. |
| Competitive moat | 5.9 / 10 | Excellent at the device level, but share gains are being granted rather than won. |
| Red flags | 5.2 / 10 | Clean accounting; structural concentration, cyclicality and now guided margin compression. |
| Valuation | 3.5 / 10 | Cheap-ish on revenue, expensive on earnings, and unsupported by the company’s own forecast. |
Take management’s own bull case at face value and you get $192. Greg Smith on the call: as wafer-fab equipment capex approaches $250bn by the end of the decade, he sees a path for the overall test market to reach or exceed $20bn. Give Teradyne a generous 40–45% of that — above its own stated 35–38% target — and discount it, and you get roughly $192 a share. The stock is at $325.
Margins are past peak, and it is guided rather than inferred. Third-quarter gross margin steps down to 58–59% from 59.8% delivered, with the full year around 59% — which implies roughly 57% in the fourth quarter. Operating expenses are guided at 29–30% of sales against 26.8%, and management named memory as a continuing strain on margins into 2027.
The market delivered its verdict in one session. TER opened up 14.9%, traded as high as +16.1%, and closed at $319.41 — below the prior close, one cent off the session low, on roughly three times average volume. A 14.2% peak-to-close reversal after a quarter like that is not profit-taking.
| 29 July 2026 session | Price | vs prior close |
|---|---|---|
| Open | $368.26 | +14.9% |
| Intraday high | $372.23 | +16.1% |
| Intraday low | $319.19 | −0.5% |
| Close | $319.41 | −0.39% |
| Volume | 6.23m | ~3x average |
| Peak-to-close reversal | −14.2% | 0.07% off the low |
Read that carefully, because it is the single most informative thing that happened this week. Teradyne reported the second-best quarter in its history, beat revenue by 7.6%, beat earnings by 17–21%, and guided the next quarter roughly 40% above consensus on the bottom line. The stock opened up 16% — and then sellers took every share offered to them for six and a half hours until it closed below where it started the week, on triple normal volume, one cent off the low.
That is not profit-taking. It is a distribution day: a session in which the marginal buyer was already positioned before the print and the marginal seller was simply waiting for liquidity. The broader tape was soft — both major indices were down around 0.9% — which explains part of it, but not a 14% round trip. The information in the reversal is that the good news was already in the price, and then some.
| Measure | Pre-call | Now | Comment |
|---|---|---|---|
| Implied volatility, annualised | 108.4% | 98.8% | About ten vol points of crush — far less than a typical post-event collapse |
| 30-day realised volatility | 103.0% | 110.6% | Rising. The 16.6% intraday range fed straight into it |
| IV percentile, 52-week | — | 92nd | Still near the top of its own one-year range, post-event |
| Implied minus realised | +5.4 pts | −11.8 pts | Implied now sits below realised — unusual, and not a comfortable premium-selling setup |
The event crush was shallow because the event did not resolve the uncertainty — it relocated it to fourth-quarter gross margin and the 2027 framing. With implied volatility now below trailing realised, options are not obviously rich here despite the 92nd-percentile reading. That is a materially different setup from the standard post-earnings vol-selling window, and worth knowing before treating a high percentile as a signal.
A ten-year discounted cash flow on a capital-equipment company is a blunt instrument. Terminal value runs from 65% of enterprise value in the bear case to 78% in the bull, and a cyclical’s terminal year is essentially unknowable. Treat the spread as the honest answer and the point estimates as illustration. The decision-relevant calculation is the management-TAM test below the model, not the number in the widget.
Every estimate here is built on the margin structure management has guided rather than the one it has just delivered, and on a revenue base cut directly from its own framing of second-half sales. Teradyne carries no debt, so the discount rate does all the damage rather than any interest expense — move the WACC slider first.
Terminal value is 73% of enterprise value. The management-TAM scenario below sits at $192 and is not one of the three presets.
Cash tax held at 16% and constant across cases. Net cash held at $517m against zero debt. Base-case FY2027E revenue of $6.15bn is our $5.12bn FY26 estimate grown at the 20% first-year rate; the FY28–32 CAGR of 10.8% is the 16% / 12% path fading toward 4%. The FY26 base is derived from management’s own framing of second-half revenue at 48–50% of full-year sales, which against H1 of $2.61bn implies $5.02–5.22bn. Implied free-cash margins across the three presets are 17.6% / 23.5% / 26.1%. The widget is a five-year constant-growth form, so the bear case’s single −18% digestion year and subsequent recovery are expressed as a smoothed +0.5% average rather than as the trough itself. The management-TAM scenario at $192 cannot be loaded as a preset — the model supports three — so it is set out in full below instead.
This is the most important calculation in the report, and it is deliberately constructed to be generous. Take management’s own long-range framing at face value and treat all of it as true. On the call, Greg Smith said that as wafer-fab equipment capex approaches $250bn by the end of the decade, he sees a path for the overall test market to reach or exceed $20bn. Grant that in full, then hand Teradyne 40–45% of it — deliberately above the company’s own stated target of 35–38% share in a $12–14bn market — plus continued growth in Product Test and Robotics, and the share gains it says begin in 2027.
| The test | Assumption | Result |
|---|---|---|
| Management’s own case | Teradyne to roughly $9.8bn of revenue by 2031 and $12.9bn by 2036, at a 24.5% free-cash margin, 10.5% discount rate, 3% terminal growth | $192 |
| What $324.97 requires | Roughly 18% revenue growth every year for ten years, on the same discount rate and terminal assumptions | ~$27bn by 2036 |
| The whole ATE market | Management’s own public forecast for the entire market by around 2030 | $20bn+ |
| Teradyne’s own share target | Against roughly 30% today. Our test is deliberately more generous than this. | 35–38% |
Management says the entire automated-test market gets to $20bn or more by around 2030. For the share price to be right, Teradyne alone must eventually approach $27bn — meaning either the market must be materially larger than management’s own bullish public forecast, or Teradyne must take a share of it that no supplier has ever held in this industry. That is not a bear talking down a good company. It is management’s own numbers, discounted.
| WACC ↓ / Terminal g → | 2.5% | 3.0% | 3.5% |
|---|---|---|---|
| 9.5% | $193 | $205 | $219 |
| 10.5% | $169 | $177 | $187 |
| 11.5% | $150 | $156 | $163 |
Robustness check. Even at the most generous corner of this grid — a 9.5% discount rate for a stock running 110% realised volatility, and 3.5% perpetual growth — the base case reaches only $219. Note also what the table does not contain: any cell reaching $325. The conclusion does not depend on a harsh cost of capital. It depends on the growth path.
| Measure | At $324.97 | Basis |
|---|---|---|
| Market capitalisation | $51.3bn | 157.7m diluted shares |
| Enterprise value | $50.7bn | Less $517m net cash, zero debt |
| P/E — trailing GAAP | 44.6x | Trailing twelve months |
| P/E — FY26E | 36.5x | Our FY26E non-GAAP EPS of roughly $8.90 |
| P/E — FY27E base | 31.1x | Our FY27E non-GAAP EPS of roughly $10.45 |
| EV / Sales — FY26E | 9.9x | Our FY26E revenue of $5.12bn |
| Price / Book | 14.9x | — |
| Return on equity, trailing | 29% | Disclosed on the call |
On revenue the shares are not the outlier they are often described as. At 9.9x forward EV/sales, TER sits below third-party screen figures near 11.7x for ASML and 11.6x for Applied Materials. It is on earnings that the premium shows, and it shows because margins are being guided down while the revenue base grows. Expensive on earnings, cheap-ish on revenue — which is exactly the signature of a company whose profitability is compressing. The valuation case here rests on the earnings multiple and the discounted cash flow, not on the revenue multiple.
Peer multiples are third-party screen data compiled between late June and 21 July 2026 and therefore pre-date second-quarter reporting across the group. They are directionally useful and should not be treated as precise. Teradyne’s own figures are calculated from the 28 July filings and the current quote.
Teradyne presented an updated long-term target earnings model at this call: revenue of roughly $6.0bn against fiscal 2025’s $3.2bn, gross margin of 59–61%, operating expenses falling to 27–29% of revenue, operating margin of 30–34% and non-GAAP earnings of $9.50–11.00. It assumes the test market expands to $12–14bn with Teradyne holding 35–38% of it. This is a target model, not 2026 guidance, and no year is attached to it — a distinction worth being careful about, because the same figures circulate in third-party coverage as though they were a full-year 2026 target. On the guidance actually delivered, our own FY26 estimate is $5.12bn of revenue and roughly $8.90 of earnings.
Same model, different inputs. Click any card to load it.
Sitting between base and bull is the path management itself described on the call, worked through in Section 02: wafer-fab equipment capex approaching $250bn, a $20bn-plus test market and Teradyne taking 40–45% of it — more than its own 35–38% target. It values the shares at approximately $192. It is the most decision-relevant number in this report, and it is deliberately the most generous reading of the company’s own guidance. The model widget holds three presets, so it cannot be loaded — but if you want to approximate it, set revenue growth near 13% and the free cash margin near 24.5%.
The presets reproduce the underlying ten-year model closely, at $62, $177 and $333, with implied free-cash margins of 17.6% / 23.5% / 26.1% against the 17.5% / 23.5% / 26.0% assumed. Note that the bull case is the only one of the four that reaches today’s price, and it does so with about 2% to spare — a narrower margin than this stock routinely moves in a single session.
The print itself was emphatic. Revenue of $1,329.0m was up 104% year-on-year and beat a roughly $1,235m consensus. Non-GAAP earnings of $2.47 beat a $2.04–2.11 consensus by 17–21%, the fourth consecutive beat. September revenue was guided to $1,200–1,300m against a Street near $1.0–1.1bn. Operating cash flow was 125% of net income, free cash flow was $378m in the quarter and $579m across the half, up 150%. There is no debt.
The call is where the shape changed. Management supplied, for the first time, the gross-margin and operating-expense guidance that the press release had left out — and both went the wrong way.
| Q3 2026 guidance | Guided | Q2 actual | Implication |
|---|---|---|---|
| Revenue | $1,200–1,300m | $1,329.0m | −5.9% sequentially at the midpoint; +63% year-on-year |
| Gross margin | 58–59% | 59.8% | Roughly 130bp headwind |
| Opex, % of sales | 29–30% | 26.8% | About $362–375m — above Q2 in absolute dollars |
| Non-GAAP EPS | $1.85–2.15 | $2.47 | −19% sequentially at the midpoint |
| Implied operating margin | ~29.0% | 33.7% | Roughly 470bp of sequential compression |
| Full-year gross margin | ~59% | H1: 60.3% | See the derived fourth-quarter figure below |
| H2 share of full-year revenue | 48–50% | — | Implies roughly $5.02–5.22bn for FY26 |
First-half gross profit was $1,575.6m on $2,611.5m, or 60.3%. The third quarter at the guide midpoint adds roughly $731m on $1,250m, or 58.5%. For the full year to land at “around 59%,” the fourth quarter has to come in near 57% on a $1.3–1.6bn revenue base — roughly 280bp below the June quarter. If management meant 58.5% for the year rather than 59.0%, implied fourth-quarter gross margin drops to about 55%. Nobody said “Q4 gross margin will be 57%” on the call. It falls out of the arithmetic of what was said, and it is the number to test against when the third quarter reports in late October.
| Question going in | What management said | Read |
|---|---|---|
| Is 59.8% gross margin the run-rate? | No — it steps down. Q3 to 58–59%, the full year to around 59%. CFO Michelle Turner attributed it to product mix and new product introductions, and named memory as a continuing strain on margins into 2027. | Negative |
| Is the opex step-up structural? | Yes. Q3 opex guided at 29–30% of sales, above Q2 in dollars, with increased investment in R&D, customer teams and manufacturing capacity. Fourth-quarter opex is expected to be comparable to the third. | Negative |
| Does backlog extend beyond twelve months? | Not directly quantified. Memory manufacturers are making test-capacity plans into 2027 and management does not expect memory lumpiness next year — supportive, but softer than a hard backlog figure. | Soft |
| Was there a full-year revenue frame? | No explicit figure, but second-half revenue was framed at 48–50% of full-year sales. Against H1 of $2.61bn that implies roughly $5.02–5.22bn. Our FY26E base of $5.12bn is cut directly from it. | Soft |
| Merchant GPU and share-gain progress? | Compute share “stabilising and starting to increase,” but gains are gradual with noticeable effects starting in 2027 — and driven by hyperscalers deliberately qualifying dual vendors for supply-chain resilience. | Mixed |
| 2027 outlook? | “Another year of healthy growth,” supported by transistor growth, memory demand and AI infrastructure spend. Constructive, but unquantified. | Positive |
On tone. Confident on the decade, deliberately unspecific on the next two quarters. The long-range disclosures were unusually generous and partly self-limiting — publishing the arithmetic that lets an outsider value your company is not something a promotional management team does. Credit where it is due. The near-term posture was the opposite: no explicit full-year revenue number beyond the second-half frame, no backlog figure, no quantified 2027.
The genuine positives were a 2026 memory market guided more than 40% larger than 2025, an explicit rejection of 2027 memory lumpiness, the internal coherence of the long-term market build, a 29% return on equity and a thirteen-year unbroken dividend record. Against those: gross margin guided down with memory named as a 2027 drag, opex confirmed structural and rising, share gains deferred to 2027 and framed as dual-sourcing rather than displacement, and AI-related revenue described as “more than 60%” of the June quarter against “nearly 70%” in March.
Every advanced processor, memory die and networking ASIC is loaded into an automated test system that runs it through thousands of electrical and functional patterns. Chips that fail are scrapped before they are assembled onto a server board worth tens of thousands of dollars. Teradyne makes those testers, the handlers that feed them and increasingly the software and data layer around them. It also owns a robotics business — Universal Robots collaborative arms and Mobile Industrial Robots autonomous carts — which is roughly 7.5% of revenue and should be treated as a free option rather than part of the thesis.
The economics work for one reason. Test programmes are written for a specific platform. Once a customer qualifies a device on Teradyne’s UltraFLEX architecture, moving it to a rival means rewriting and re-validating that programme at enormous cost and real schedule risk on a product with a two-year life. That is why this market has been a stable duopoly with Advantest for more than a decade, the two of them holding in excess of 80% of it between them.
| Segment | Q2’26 | % of total | What drives it |
|---|---|---|---|
| Semiconductor Test | $1,122m | 84.4% | SoC test for AI accelerators, networking and hyperscaler-designed compute; memory test across HBM, DRAM and NAND final test. The fastest-growing part — and, in memory, the lowest-margin part. |
| Product Test | $107m | 8.1% | Board test, defence and aerospace, wireless via LitePoint, photonics via Quantifi. Up 33.8% sequentially as a $139.9m multi-year US defence contract lands. |
| Robotics | $100m | 7.5% | Collaborative arms and autonomous mobile robots. Fifth consecutive sequential gain after a 2025 restructuring. |
Geographically the picture is concentrated and politically exposed. On fiscal 2025 figures, Taiwan was 36% of revenue, China 14%, Korea 14%, the United States 11% and Europe 7%. Roughly 89% of revenue is generated outside the United States, on a headcount of about 6,600. None of that was addressed in detail on the call.
| Advantage | Score | Our read |
|---|---|---|
| Switching costs | 9/10 | The core of the moat. Test programmes, correlation data and load-board hardware are platform-specific. Requalifying a high-volume device on a rival tester costs months of engineering and carries yield-excursion risk. Customers change platforms at generational transitions, not opportunistically. |
| Intellectual property | 8/10 | Six decades of instrument, pin-electronics and timing IP. Newer platforms show the R&D engine is still producing, and buying Infineon’s ATE technology team outright in 2025 was buying capability rather than revenue. |
| Customer loyalty | 7/10 | Very high retention — but read it alongside the concentration flag. Loyalty from a customer who is 19% of your revenue is also dependence. |
| Scale | 7/10 | Effectively a two-player market above 80% combined share. Scale funds R&D that sub-scale entrants cannot match at the leading edge. Teradyne is, however, the smaller of the two. |
| Brand & reputation | 6/10 | Strong technical reputation, but purchasing decisions are made by engineering teams on cost-of-test economics. Brand commands no price premium here. |
| Distribution & service | 6/10 | Direct global applications and service organisation embedded in customer fabs. Real, but table stakes for both duopolists. |
| Cost advantage | 5/10 | A 59.8% gross margin proves pricing power at the high end. But a designated second source competes on allocation rather than displacement, and Chinese entrants at 20–30% discounts change the calculus in trailing-edge nodes. |
| Data advantage | 5/10 | Test generates enormous yield data and both duopolists are embedding analytics. Early-stage, not yet durable, and customers increasingly want to own their own test data. |
| Network effects | 4/10 | Weak. A partner ecosystem exists — Tokyo Electron, MultiLane, Vention on the robotics side — but there is no genuine two-sided network. |
| Regulatory protection | 3/10 | Regulation is a net negative. Export controls restrict Teradyne’s addressable market in China more than they protect it, and they actively incentivise domestic substitution. |
The two scores that keep this out of the sevens are cost advantage and customer loyalty, and both come down to how the share question was answered on the call. Smith said large customers are qualifying multiple suppliers to reduce supply-chain risk, and management framed this formally as a dual-vendor lifecycle running through to mature-vendor status. Read it carefully: Teradyne’s incremental share is being granted by customers pursuing vendor diversification, not taken by Teradyne outcompeting Advantest on merit. A designated second source typically gets a capped allocation and less pricing leverage than a displacing winner. The same logic could later be applied to Teradyne in the segments where it is the incumbent.
In fairness, the company’s own framing is more ambitious than that reading — roughly 30% share today, a 35–38% target, and a mature-vendor range it puts considerably higher. And management’s confirmation that memory customers are planning test capacity into 2027 is real evidence of embeddedness in customer planning cycles.
The switching-cost moat is genuinely excellent and would justify a premium multiple in a stable market. What it does not do is protect against the demand side going quiet. A moat determines who wins the orders; it does not create the orders. In a buy-rate business that distinction is everything, and it is why a 9-out-of-10 switching-cost score sits inside a 5.9 composite. High-end gains are real but slower and less proprietary than the headline suggests; low-end Chinese erosion at 28nm continues.
We ran a deliberately hostile read of the filings and the market data. The shape of the result matters more than the score. Everything in the accounting, disclosure, leverage and dilution columns is clean — genuinely clean, not clean-because-we-did-not-look. The GAAP-to-non-GAAP gap in the June quarter was 3.9%. Stock compensation of $42.0m in the first half is 1.6% of revenue, very low for a technology company. Excess and obsolete inventory provisions are falling while revenue doubles. There is no debt and no litigation overhang.
| Area | Severity | Evidence, and why it matters |
|---|---|---|
| Customer concentration | 8/10 | The five largest direct customers were 44% of FY25 revenue, against 36% in 2024 and 32% in 2023 — up twelve points in two years. One direct customer alone was 19%. The dual-vendor dynamic increases dependence on a small number of hyperscalers making qualification decisions, and those buyers have enormous pricing leverage. |
| Margin trajectory | 7/10 | Not an inference from one quarter — it is guided. Gross margin to 58–59% in Q3 and around 59% for the year, implying roughly 57% in Q4; opex structurally at 29–30% of sales; memory named as a margin drag into 2027. Operating margin compresses about 470bp sequentially. Peak-margin risk in a cyclical is how multiples get destroyed. |
| Cyclicality & lumpiness | 6/10 | A buy-rate business with no meaningful contractual recurring revenue. Partly mitigated: management explicitly said it does not expect memory lumpiness in 2027 because customers are already planning test capacity into next year. But a 39% peak-to-trough drawdown occurred inside the strongest fundamental stretch in company history. |
| Geopolitical / export control | 6/10 | 89% of FY25 revenue booked outside the United States, Taiwan alone at 36%. The 10-K risk factors specifically call out Commerce Department rules on equipment destined for China, plus tariff exposure. A live policy variable, and not addressed in detail on the call. |
| Receivables / revenue quality | 4/10 | Receivables rose from $786.9m to $1,109.7m in six months and DSO moved from about 67 to 76 days. Ordinarily a warning. Here it is largely explained by revenue nearly doubling and by shipment timing, and it is strongly contradicted by cash flow — free cash flow of $378m in the quarter and $579m across the half, up 150%. Watch, do not panic. |
| Insider activity | 3/10 | Seventeen open-market transactions in six months, all sales, zero purchases, including a further CEO sale in mid-July. All appear to be pre-arranged 10b5-1 executions, which blunts the signal considerably — but a complete absence of open-market buying after a 3.4x run is worth noting. |
| Accounting & non-GAAP hygiene | 2/10 | Genuinely clean. A 3.9% GAAP-to-non-GAAP gap, ordinary add-backs, stock compensation at 1.6% of revenue, falling obsolescence provisions. |
| Debt & dilution | 1/10 | Zero debt, $517m cash, share count down 1.5% year-on-year, 29% return on equity, thirteen consecutive years of dividends. Nothing to see — and that is the point. |
This is not a forensic-accounting story. The books are clean, cash conversion validates the earnings, and there is no leverage. The score is driven almost entirely by structural exposure — a customer base getting more concentrated rather than less, a business model with no contractual revenue floor, 64% of revenue sourced from three geographies on the world’s most contested fault line, and margins now guided down on management’s own numbers. Combine that with a valuation that leaves no room for a bad quarter and the risk is real, even though the accounting is honest.
We score management 4.0 out of 5. The single most persuasive datapoint is not the revenue growth, which the cycle largely handed them. It is that they stopped buying back stock as it tripled — first-half repurchases of $74m against $275m a year earlier — repaid the revolver, and redirected the capital into MultiLane, TestInsight, Quantifi Photonics and Infineon’s ATE team instead. Most management teams do the reverse and call it confidence. Total goodwill of $664m against $1.4bn of annualised operating cash flow is not empire building, and diluted shares are down from 160.1m to 157.7m.
CFO credibility scores 4.5. Asked directly what was driving the 130bp gross-margin headwind, Michelle Turner answered specifically — mix, new product introductions — and then volunteered the harder fact, that memory will strain margins into 2027. Naming a multi-year headwind unprompted, on a morning the stock was up 16%, is the behaviour of a CFO managing credibility rather than the tape. Transparency also scores 4.5: management published a long-range market build specific enough that an outsider can check it — a wafer-fab equipment path, a test-market size, a share target and a target earnings model, all with numbers attached. Section 02 of this report is only possible because they did.
Two things hold the composite down. Guidance accuracy scores 3.0: four consecutive earnings beats, the last two by around 21%, with June revenue landing $79m above the top of the guided range. That is technically a forecasting failure however pleasant, and it is a large part of why this stock gaps 12–19% on results; 2027 was then described as “another year of healthy growth” without a number. And insider alignment scores 3.0: seventeen sales, zero purchases, no meaningful founder-style stake. Note too that the buyback discipline we credit above is simultaneously a statement about price.
BULLRevenue doubled year-on-year. Not a beat — a doubling. All three segments grew. September is guided up 63% year-on-year against a consensus that was $200m too low, and the stock is still 33% below its June high.
BEARRead the sequential line. September is 5.9% down quarter-on-quarter on revenue and 19% down on earnings. Semiconductor Test — 84% of the company — grew 1.0% sequentially. The year-on-year comparison flatters because the June 2025 quarter was a trough.
BULLWafer-fab equipment going toward $250bn by the end of the decade is a high-teens compound rate for the whole industry, and that supports a $20bn test market. Teradyne at 35–38% of that, plus Product Test and Robotics, is a far bigger company than today — and they just told us the plan.
BEARAgreed on all of it — and it still only gets you to about $192 a share discounted, using a share assumption more generous than their own target. The disclosure is credible and insufficient. When a company’s own bull case, taken at face value, values the stock 40% below the market price, the disclosure is a bear argument.
BULLShare is going up, not down. Roughly 30% today, a stated target of 35–38%, and a mature-vendor path management puts well beyond that. Compute share is stabilising and starting to increase. That is a company taking ground in the highest-value socket in the industry.
BEARTaking ground it was handed. Smith said customers are qualifying multiple suppliers to reduce supply-chain risk, and management calls it a dual-vendor lifecycle. A designated second source gets a capped allocation and less pricing power than a winner who displaced someone. And the effects start in 2027, so there is no near-term proof point either way.
BULLA 59% full-year gross margin is outstanding for capital equipment. Memory dilutes margin because memory is growing 40%-plus — that is a high-quality problem. Operating margin near 30% on a business with no debt is excellent, and June operating cash flow was 125% of net income.
BEARThe full-year guide implies roughly 57% in the fourth quarter, about 280bp below June. Opex is structurally at 29–30% of sales. Operating margin compresses 470bp sequentially in September. On a 36x forward multiple, mix-driven margin erosion is not a high-quality problem — it is the multiple’s problem. And top-five customers are 44% of revenue, up twelve points in two years, with one at 19%.
BULLOne session proves nothing. The tape was down 0.9% and the de-rating largely happened weeks ago. At 9.9x forward EV/sales the stock is cheaper than both ASML and Applied Materials on that measure.
BEARThe EV/sales point is fair, and it is the strongest thing the bull has said. But margins are guided down, so revenue multiples are the wrong lens. And a 16% gap higher closing at the low on triple volume, after the second-best quarter in company history and a 40% guidance beat, is about as clear a distribution signal as this market produces.
CHAIRThe bear has the stronger evidence; the bull has the stronger story. Where the bull is clearly right: the business is excellent, the accounting is clean, management is capable and unusually candid, and AI test demand is a real structural driver rather than a narrative. None of that is seriously contestable. Where the bear is clearly right: management published a long-range forecast specific enough to be discounted, and discounted it comes to roughly $192. Grant them the whole $20bn market, hand them more share than they ask for, and the number still lands well below the quote. Excellent business. Not 36x-forward-earnings excellent, on our numbers.
One. A test-market forecast materially above the $20bn management has put in public — the single assumption doing the most work in Section 02. Two. Fourth-quarter gross margin landing at or above 59% rather than the roughly 57% the full-year guide implies, which would be evidence the memory drag is smaller than stated. Three. A quantified 2027 revenue frame implying more than 25% growth, which would beat our base case by a wide enough margin to reset the model. Four. Merchant GPU converting to multiple high-volume SKUs rather than second-source allocation.
We rate TER overvalued — constructive on the company, unwilling to underwrite the price. This is a high-quality business with clean accounts, capable management and a real structural tailwind, trading at roughly 84% above our base case. That gap does not close on a modest overshoot; it requires the bull case to become the base case. Still genuinely uncertain: the duration of the AI capital-expenditure cycle. Everything above is downstream of it, and nobody in this debate knows.
Where we differ from the Street is wide, and we should be plain about it. Published targets run from $270 to $550, with a median near $443 across ten analysts — all of which pre-date this print and are therefore stale. Our base case is below every one of them. We are not arguing the analysts are wrong about the business. We are arguing that their targets require the AI capex cycle to run for years without a digestion phase, and in a buy-rate business that has never once behaved that way, that is an assumption rather than a base case.
Think of Teradyne as the quality-control department for the chip industry, except it sells the equipment rather than doing the work. Before a chip ships, it has to be tested — plugged into a machine that runs it through thousands of checks. Teradyne builds those machines. If a $30,000 AI accelerator turns out to be faulty after it has been soldered onto a server board, the whole board is scrap. So chipmakers really, really care about test.
It sells testers — expensive, complex machines — plus the service and support around them. About 84% of revenue comes from testing semiconductors. There is a smaller side business selling robot arms and warehouse robots, but that is roughly 8% of the company. Don’t buy this for the robots.
The company reported a fantastic quarter and said the next one would also be much better than expected. The stock jumped 16% at the opening bell. And then it fell all day and finished slightly lower than where it started, on triple the usual trading volume. Three things from the conference call explain why.
First, profit margins are going down, not up. They make about 60 cents of gross profit per dollar of sales today; that drops to 58–59 cents next quarter and probably around 57 by the end of the year, because memory chips — their fastest-growing business — are also their least profitable one. Second, costs are going up permanently. They’re hiring, building factories and spending more on research. Good long-term; it squeezes profits now. Third, growth slows from here. Next quarter’s sales are guided to be smaller than this quarter’s, and the market-share gains management talks about do not really show up until 2027 — partly because big customers want a second supplier for safety, not because Teradyne beat anyone.
Not remotely. No debt, plenty of cash, a 29% return on equity, thirteen straight years of paying dividends, and a business growing over 60% this year. It is a genuinely excellent company.
The price. Here is the cleanest way to see it. We took management’s own optimistic forecast — the chip-testing market reaching $20bn-plus by around 2030, with Teradyne taking an even bigger slice than it says it is aiming for — and calculated what the shares would be worth. The answer is about $192. The stock is $325. For $325 to be right, Teradyne would eventually need to be bigger than management thinks the whole market will be.
If you take one thing away: this is a high-quality business attached to a high-risk share price. Those are two separate judgements, and it is a common mistake to let the first one settle the second. Teradyne being an excellent company does not tell you whether $325 is a good price for it — and on our numbers, and on management’s own published forecast, it probably isn’t. If you like the story, position size and entry price matter far more here than in an ordinary stock, because a 40% drawdown has already happened once this year while everything was going right.
Our standing rule is that every figure traces to a primary source and every derived input is labelled as derived. This report leans on an earnings call, so it is worth being explicit about which category each load-bearing number falls into.
| Status | What it covers |
|---|---|
| Verified from Teradyne | All Q2’26, Q1’26, Q2’25, H1 and FY25 financial statement figures; segment revenue; Q3’26 revenue guidance of $1,200–1,300m and non-GAAP EPS of $1.85–2.15; Q2 free cash flow of $378m and H1 of $579m, up 150%; cash and securities of $517m; zero debt; thirteen consecutive dividend years; H1 capital allocation of $74m repurchases, $41m dividends and $176m acquisitions; AI-related revenue above 60% of the quarter; Smith’s wafer-fab-equipment and $20bn test-market framing; the long-term target earnings model; customer concentration and geography from the FY25 Form 10-K. |
| Our estimate or judgement | FY26E and FY27E revenue and earnings; the derived ~57% fourth-quarter gross margin; every DCF input and output, including the management-TAM scenario, the 40–45% share assumption and the reverse-DCF implied growth rate; all scores in Sections 06 and 07 and the management assessment in Section 08; the reading of the 29 July reversal as distribution rather than profit-taking. |
| Still open | Teradyne had not published the Q2’26 slide deck, prepared remarks or GAAP reconciliation on its investor site at the time of writing — only the earnings release and an audio replay. Guidance detail sourced from the call rather than the release, including the 58–59% gross-margin range, the 29–30% opex range and the 48–50% second-half revenue frame, will be reconciled against those documents and the Form 10-Q when they publish. |