Teradyne
The maker of the machines that check every finished chip works before it ships, now riding the rush to build AI data centres.
The scores
Growth3/4 ↓
Not Compounding: under two years of record, narrowing to AI compute and memory, and Taiwan plus Korea rose from 42% of revenue to 60%.
Business quality3/4 ↓
Not Protected: Advantest holds the larger system-on-chip share at twice the operating margin, and buyers are deliberately dual-sourcing to cut the switching cost.
Financial position4/4 ↑
Net cash, no drawn debt since the $200.0M was repaid, a new undrawn $1.0B facility to 2031, and every plan funded from operations.
Execution record3/4 ↑
Not Delivers: revenue above the guided midpoint eight quarters running and a well-timed Technoprobe stake, but Robotics has lost money for years and was restructured.
Per share capture3/4 ↓
Not Intact: the share count rose 0.19% in six months as the buyback fell to $74.2M, so issuance is no longer outrun.
Potential2/4 →
Expected, not committed: no backlog figure is disclosed anywhere, and 2027 rests on somebody else's capital budget and a share gain still at correlation.
Risk profile2/4 ↓
Beyond Concentrated: Taiwan and Korea are 60% of revenue, one customer 19%, and the filing says trade restrictions already limit competitiveness in certain markets.
Valuation2/4 ↓
Not Fair: the price needs peak-rate operating profit to double again, while management says the ratio that doubled its market stops here.
Evidence quality3/4 →
Not Corroborated: the ratio the whole argument turns on is a management claim from one call, and no independent measure of chip-equipment spending was reachable.
If it works 0.5× · if it fails you keep about 47%.
The Verdict
We rate Teradyne unattractive. It makes the machines that test finished chips, and the business is solid: sales doubled in a year on the AI build-out. But the market grew while Teradyne's share did not, and its chief executive says the trend that doubled the market stops rising here. The price treats this peak as a plateau, so it is full, not cheap. As a bet it is priced for success: chip testing covers under half the price, and robots growing as promised add little. The view hinges on whether test's slice of chip-equipment spending, or Teradyne's own share, keeps rising through 2027.
The long caseIf factories building for AI keep automating assembly and test, robot arms that work beside people become standard kit, and Teradyne sells one of the leading ones.
The evidence under this - 9 claims
- Advantest's own results show consolidated sales up 44.7% in its year to March 2026 at a 44.2% operating margin, so the leader is compounding at least as fast in the same market.
- The chief executive volunteered that Advantest 'believed that they were going to continue to gain share in SoC' and agreed only about memory, which is a supplier conceding the point on the segment that matters.
- He also said the share change should start to show in 2027 and is 'a socket by socket thing', so flat today is not a forecast of flat.
- No independent measurement of total semiconductor capital equipment spending was reachable: semi.org returns HTTP 403 to curl and to an agent fetch alike, so the denominator in this ratio rests entirely on a forecast Teradyne cites and does not name.
- The same executive calls the relationship 'correlation, not causation' with about three quarters of lag, which means the ratio measured within a single year mixes two different vintages of spending.
- 2026-10-28editedthe Q3 2026 call gave the test share for the first eight months of 2026 as 8.1%, against 6.4% for 2025
- The same remarks argue the opposite for the medium term: advanced packaging raises test intensity per die because one latent defect destroys a more valuable multi-chip package, and both package volumes and dies per package are said to keep rising through the end of the decade.
- A chief executive guiding down the ratio his own market depends on has an incentive to set an expectation he can beat, which cuts against reading it as a ceiling.
- Annualising a half year assumes no seasonality, and Teradyne's own guidance says the second half will be lower, so $32.45 flatters the price rather than the reverse.
- Against consensus 2028 earnings per share of $15.46 the same price is about 24.6 times, which is not an extreme multiple for a capital-equipment maker if the cycle runs that far.
- The thesis argues this is near a cycle peak: 2025's operating profit was $650.1 million (C10), so holding $1.40 billion as the core flatters it if test's share of chip spending reverts (C15).
- Q2 2026 alone annualises higher, at about $1.82 billion for the first half (C29), so the trailing year understates the run rate if the level holds.
- Robotics revenue fell 15.5% in 2025 and the segment lost money in each of the last three years (C55), so 'in proportion' is a promise against a record of shrinking.
- No order, backlog or customer figure for robotics is disclosed; the end-market growth is management's word alone.
The company
Every chip is tested before it ships, twice: once on the silicon wafer and again after it is packaged. Teradyne's testers are racks of instruments that drive signals into the chip at full speed and decide pass or fail.
The buyers are chipmakers, the foundries that make chips for others, and the firms that package and test them. Its annual report names Advantest, SPEA and Cohu as rivals.
The evidence under this - 3 claims
How it makes money
Teradyne is paid when a customer buys a tester, and again for servicing it. So its sales follow how much chipmakers spend on new equipment, and that swings with the chip cycle. From 2021 to 2023 sales fell 27.7% and operating profit fell by more than half.
The evidence under this - 2 claims
Where the money comes from, stream by stream
| Stream | Share of sales (Q2 2026) | How it earns it |
|---|---|---|
| Semiconductor Test: system-on-a-chip | 63.4% | Testers for logic chips - AI compute, networking, mobile and auto processors - sold to chipmakers and outsourced test houses, plus service. |
| Semiconductor Test: memory | 16.0% | Testers for DRAM, high-bandwidth and flash memory, sold mostly to the Korean and other Asian memory makers. |
| Semiconductor Test: integrated system test | 5.0% | System-level and storage test equipment that checks a finished chip or drive working as it would in use. |
| Product Test | 8.1% | Test systems for defence and aerospace electronics, circuit boards, wireless devices, high-speed links and silicon photonics. |
| Robotics | 7.5% | Collaborative robot arms (Universal Robots) and autonomous mobile robots sold to factories and warehouses, largely through distributors. |
Who runs it
Gregory Smith has run Teradyne since 2023, after leading its chip-testing and robot businesses; Michelle Turner became finance chief in late 2025. Under Smith, sales have beaten the middle of the company's own forecast every quarter for two years. Insiders own well under 1% of the company. Their bonus rides mostly on sales growth, profit margin and internal goals the company does not publish; their long-term shares on profit and on how the stock does against the market.
The evidence under this - 4 claims
Half the bonus rides on goals whose targets the proxy withholds as competitively sensitive, so outsiders cannot check how hard they were; no metric pays on return on capital, and cash appears only as an inventory-efficiency Vital Goal.
Why it wins, and who could beat it
Why customers come back
A customer writes its test programme for one platform's software, then proves chip by chip that it gives the same answer as the reference machine. Moving to a rival means doing all of that again, which Teradyne's chief executive says takes nine to twelve months.
Amkor, one of the largest packaging-and-test firms, confirms it from outside: its annual report says it concentrates spending on standard tester platforms to keep them busy. The flip side is that share barely moves either way. Teradyne's sits near 37%.
The evidence under this - 4 claims
- The same passage says those customers are running the process precisely to derisk their supply chain, so the nine-to-twelve-month cost is a barrier the buyers have decided to pay - which makes it a switching cost being deliberately dismantled rather than one holding.
- Amkor's filing says many of its own orders are cancellable, and Teradyne's 10-K says the same of its backlog, so a qualification does not convert into a committed purchase.
- Advantest's own results show consolidated sales up 44.7% in its year to March 2026 at a 44.2% operating margin, so the leader is compounding at least as fast in the same market.
- The chief executive volunteered that Advantest 'believed that they were going to continue to gain share in SoC' and agreed only about memory, which is a supplier conceding the point on the segment that matters.
- He also said the share change should start to show in 2027 and is 'a socket by socket thing', so flat today is not a forecast of flat.
The threat
Advantest, the market leader, is the rival that matters: its sales rose 44.7% in the year to March 2026, at an operating margin of 44.2%. Cohu, which already sells the equipment around Teradyne's machines, now calls itself a growing tester maker. And several customers build their own.
The evidence under this - 2 claims
The rivals: COHU, ROK, KEYS, and 2 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| Advantest CorporationPriced against it: Listed abroad | Semiconductor Test | On sale | Listed abroad | |
| COHU | Cohu, Inc.Priced against it: Same product | Semiconductor Test | On sale | Same product |
| FANUC, ABB and YaskawaPriced against it: Listed abroad | Robotics | On sale | Listed abroad | |
| ROK | Rockwell AutomationPriced against it: One segment only | Robotics | On sale | One segment only |
| KEYS | Keysight TechnologiesPriced against it: One segment only | Product Test | On sale | One segment only |
How big it could get
How big this could get
The market is the testers chipmakers buy each year, a slice of everything they spend on equipment. That slice rose from near 4% in 2023 to 8% in the first five months of 2026.
The chief executive says it settles at 7% to 9% and may slip to 6% or 7% in 2027.
The evidence under this - 3 claims
- No independent measurement of total semiconductor capital equipment spending was reachable: semi.org returns HTTP 403 to curl and to an agent fetch alike, so the denominator in this ratio rests entirely on a forecast Teradyne cites and does not name.
- The same executive calls the relationship 'correlation, not causation' with about three quarters of lag, which means the ratio measured within a single year mixes two different vintages of spending.
- 2026-10-28editedthe Q3 2026 call gave the test share for the first eight months of 2026 as 8.1%, against 6.4% for 2025
- The same remarks argue the opposite for the medium term: advanced packaging raises test intensity per die because one latent defect destroys a more valuable multi-chip package, and both package volumes and dies per package are said to keep rising through the end of the decade.
- A chief executive guiding down the ratio his own market depends on has an incentive to set an expectation he can beat, which cuts against reading it as a ceiling.
Who has signed
Nothing is binding. Teradyne shipped a first order to a maker of graphics chips for AI in the June quarter, value undisclosed, and memory makers ordered more than twice what they were billed. Both are orders a customer may cancel, and no backlog figure is published.
The evidence under this - 3 claims
- Fast follower is the phase after correlation, but the capacity buys are dated to the second half of 2027, so no second-customer production ramp is running in 2026.
- No volume, term or counterparty was disclosed for any of the programmes, and the same executive said Teradyne's compute share has been diluted by the market growing faster than its wins.
- 2026-10-28editedthe Q3 2026 call restated each customer's dual-vendor phase; the second hyperscaler is past qualification with capacity buys expected in the second half of 2027
- Memory was $212 million of the quarter, so a ratio over two on that base is roughly $430 million of orders against a company guiding the next quarter down.
- The same 10-K says orders may be cancelled without advanced notice, so a book-to-bill ratio is not a committed quantity.
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| Product | 90% | +127.9% | +115.2% | +68.2% |
| Service | 10% | +6.6% | +9.6% | +10.1% |
| Whole company | 100% | +103.9% | +95.3% | +57.9% |
The filed split through 2026-06-28. These rates are not carried forward: the reading below says why.
How we read it: Trusts the trailing blend over the mix's doubling: the half-year rate is AI compute and memory off a weak 2025 base, not acquisitions, and guidance flattens the second half amid cyclicality.
The results
The numbers
Sales grew 58% over the twelve months to Jun 2026, faster than the 13% in the year to Dec 2025.
Operating margin is 30%, up from 19% in the year to Dec 2023.
Running costs take 29¢ of each sales dollar, up from 27¢ in the year to Dec 2021.
Each dollar put into the business earns 37% a year after tax, down from 44% in the year to Dec 2021.
It turned 18% of sales into free cash, $800M over the last twelve months.
It holds $350M of cash against $0 of debt, $350M more cash than it owes.
The share count fell 14% in 4.5 years, and pay in shares costs 1.7% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $3.70B | $3.16B | $2.68B | $2.82B | $3.19B | $4.46B |
| Sales growth | +19% | −15% | −15% | +5.4% | +13% | +58% |
| Profit | ||||||
| Gross margin | 60% | 59% | 57% | 58% | 58% | 59% |
| Operating margin | 32% | 26% | 19% | 21% | 20% | 30% |
| Net margin | 27% | 23% | 17% | 19% | 17% | 26% |
| What growth costs, per sales dollar | ||||||
| Research and development | 12% | 14% | 16% | 16% | 16% | 13% |
| Sales and marketing | - | 9.8% | 12% | 12% | 11% | 8.8% |
| General and admin | - | 4.9% | 6.0% | 5.9% | 5.3% | 3.8% |
| Pay in shares | 1.2% | 1.5% | 2.2% | 2.1% | 2.0% | 1.7% |
| Capital spending | 3.6% | 5.2% | 6.0% | 7.0% | 7.0% | 5.9% |
| Cash | ||||||
| Cash from operations | $1.10B | $578M | $585M | $672M | $674M | $1.06B |
| Free cash flow | $966M | $415M | $426M | $474M | $450M | $800M |
| Balance sheet | ||||||
| Cash and short-term investments | $1.12B | $855M | $758M | $553M | $294M | $350M |
| Debt | - | - | - | - | $200M | $0 |
| Net cash | - | - | - | - | $93.8M | $350M |
| Returns and shares | ||||||
| Return on capital | 44% | 28% | 17% | 20% | 20% | 37% |
| Shares (diluted) | 184M | 170M | 164M | 163M | 160M | 158M |
What growth costs
Growth makes Teradyne more profitable. The cost of each machine rises with sales, but engineering, selling and office costs barely move, so a bigger year keeps more of each sale. The same costs bite when sales fall, as the last downturn showed.
The evidence under this - 2 claims
The same fixed costs cut the other way when sales fell from 2021 to 2023: operating margin went from 32.4% to 18.7% (C31), so the leverage is the cycle's and runs in both directions.
The last report (28 Jul)
Quarter ended June 2026, filed figures.
The September quarter brought $1.31B of sales, a little below June's $1.33B, and 31.6 cents of every sales dollar was operating profit. The company's own guide for the fourth quarter is $1.25B to $1.35B, so its midpoint of $1.30B is again under June.
The call added two things. Chip testing was 8.1% of chipmakers' equipment spending in the first eight months of 2026, inside the 7% to 9% range its chief executive has said it settles in. And the second big cloud customer has passed testing and is expected to buy capacity in the second half of 2027, not before.
What it did to the thesisThe guide landed between our two lines, so the view holds. June is still the biggest quarter of 2026 on the company's own midpoint, and margins are guided lower. Growth next year is promised, but dated to late 2027.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $2.47 against $2.09 expected; $2.38 as filed, the gap being what the company adds back.
The evidence under this - 4 claims
Every one of the last seven quarters beat the guidance midpoint, so the $1,300M midpoint may understate Q4; the top of the range, $1,350M, is above Q2 2026's $1,329.0M.
- Every one of the last seven quarters beat the guidance midpoint, so the midpoint is the wrong central case if the pattern holds.
- Management says the next surge at its compute customers is expected in the first half of 2027, so a 2026 plateau is consistent with its own growth case rather than contradicting it.
- 2026-10-28editedthe fourth-quarter guide was published: $1,250M to $1,350M, midpoint $1,300M, above the $1.26B this claim estimated and still below Q2 2026
- Fast follower is the phase after correlation, but the capacity buys are dated to the second half of 2027, so no second-customer production ramp is running in 2026.
- No volume, term or counterparty was disclosed for any of the programmes, and the same executive said Teradyne's compute share has been diluted by the market growing faster than its wins.
- 2026-10-28editedthe Q3 2026 call restated each customer's dual-vendor phase; the second hyperscaler is past qualification with capacity buys expected in the second half of 2027
- No independent measurement of total semiconductor capital equipment spending was reachable: semi.org returns HTTP 403 to curl and to an agent fetch alike, so the denominator in this ratio rests entirely on a forecast Teradyne cites and does not name.
- The same executive calls the relationship 'correlation, not causation' with about three quarters of lag, which means the ratio measured within a single year mixes two different vintages of spending.
- 2026-10-28editedthe Q3 2026 call gave the test share for the first eight months of 2026 as 8.1%, against 6.4% for 2025
What management did with the money
It bought back far fewer shares as the price rose: $74.2M in the first half of 2026, at an average $341.89, against $277.3M a year earlier at $93.67. Its best use of money was a 10% stake in Technoprobe, bought for 483.1 million euros in 2024 and quoted at $2.46B in June.
Insiders in 2026Q2: 16 purchases and 16 sales, from their Form 4 filings.
The evidence under this - 3 claims
The price, and where we disagree
In short. Full. A good business priced as if its best stretch were permanent. If the cycle turns as it did after 2021, the profit this price rests on disappears.
The test has been read: on 27 October Teradyne reported the September quarter and guided the fourth to a $1.30B midpoint, under June's $1.33B. The next one is the January results, with the updated long-term target model.
The spending underneath is real. TSMC, the largest foundry, spent US$40.9B on equipment and buildings in 2025, a third more than in 2024, and guides US$52B to US$56B for 2026. But its wafer capacity barely moved, at about 17 million a year: the money bought newer technology, not more chips to test.
The evidence under this - 4 claims
Every one of the last seven quarters beat the guidance midpoint, so the $1,300M midpoint may understate Q4; the top of the range, $1,350M, is above Q2 2026's $1,329.0M.
- The same sentence says the figure 'may be adjusted later' depending on market conditions, and TSMC has a record of revising capital plans intra-year.
- One customer's capital budget is not the market: TSMC is the largest single buyer of front-end equipment but Teradyne's testers are bought largely by outsourced assembly and test houses and by memory makers, whose plans TSMC's filing says nothing about.
How we read the multiple
Counting its debt and cash, buyers pay about $44 for each $1 of last year's operating profit, the profit from the business itself. On the first half's faster pace it is about $32.
But that pace is a peak:
- It is about 52% above the last peak, in 2021.
- After that peak, sales fell about 28% and operating profit more than halved within two years.
- So for today's price to be right, a peak has to become the floor.
Against its own history. Far dearer than its own normal, on margins it has not held through a full year lately, so the gap measures optimism rather than a better business.
The evidence under this - 4 claims
- Annualising a half year assumes no seasonality, and Teradyne's own guidance says the second half will be lower, so $32.45 flatters the price rather than the reverse.
- Against consensus 2028 earnings per share of $15.46 the same price is about 24.6 times, which is not an extreme multiple for a capital-equipment maker if the cycle runs that far.
Priced against its rivals
Only Cohu sells the same thing to the same buyers as a large part of what it does. A price comparison needs at least two, so the price is judged against its own history instead.
Private or foreign rivals that cannot be priced: Advantest Corporation, SPEA S.p.A., Test Research, Inc., Rohde & Schwarz GmbH & Co. KG, Anritsu Company, National Instruments Corporation, Welzek / iTest, KUKA Robotics Corporation, ABB Ltd / FANUC / Yaskawa Electric / Omron / KION Group, Staubli / Techman / Doosan Robotics / Jaka / AUBO Robotics / Junion / HikRobot / Agilox.
The bet: what it pays if it works
The bet is on Robotics alone: robot arms that work beside people and factory robots that move goods. It lost money for three years and shrank last year, and only now nears breaking even.
Chip and product testing, held at last year's profit, valued at an ordinary equipment maker's price, cover under half of today's price. Even if robots grow as management promises, the whole company would be worth about half of today: an expensive bet.
If robots never earn, shareholders keep the testing businesses and the cash.
| Robotics operating profit in two to three years | Million dollars of Robotics revenue a year needed | At the stated multiple |
|---|---|---|
| $40.00M | about 400 at 10% kept | 0.5× |
| $53.00M (the big win) | about 530 at 10% kept | 0.5× |
| $108.00M | about 1.1K at 10% kept | 0.5× |
To pay 2× (a Fair bet) it would need about $4.52B of Robotics operating profit (about 45.2K million dollars of Robotics revenue a year at 10% kept); to pay 5× (a Cheap bet), about $13.38B of Robotics operating profit (about 133.8K million dollars of Robotics revenue a year at 10% kept), at the same multiple.
Each row is Robotics alone, plus the rest of the company (Semiconductor Test and Product Test) counted at about $27.93B.
The steps it has to take
- Done Robotics restructured: about 150 jobs cut in 2025
- Done Robotics near break-even: $2.6M lost in Q2 2026
- In progress US manufacturing centre opens due 2026-12
- Not started A full year of Robotics profit before tax due 2027-12
- Not started Robotics revenue about $530M a year at a 10% margin due 2028-12
The detail behind the label
The evidence under this - 6 claims
The 2025 loss carries the restructuring that cut about 150 Robotics jobs, so part of it is a one-off charge rather than the running cost of the business (C56).
The segment has not yet reported a profitable quarter since the restructuring, and the first half of 2025 carried $10.0 million of Robotics severance, so part of the swing is a charge not repeating.
- Robotics revenue fell 15.5% in 2025 and the segment lost money in each of the last three years (C55), so 'in proportion' is a promise against a record of shrinking.
- No order, backlog or customer figure for robotics is disclosed; the end-market growth is management's word alone.
- The thesis argues this is near a cycle peak: 2025's operating profit was $650.1 million (C10), so holding $1.40 billion as the core flatters it if test's share of chip spending reverts (C15).
- Q2 2026 alone annualises higher, at about $1.82 billion for the first half (C29), so the trailing year understates the run rate if the level holds.
Rockwell's segment is drives, motion and sensing at a far larger scale, not collaborative robots, so its margin is what a mature automation-hardware business earns rather than what Teradyne's robots have shown.
Where we disagree with the market
This is a multi-year AI growth phase, not a cycle peak, and analysts keep raising their ratings and targets.
The doubling came from two ratios that have stopped moving: test's slice of equipment spending, and Teradyne's own share, which never moved.
Consensus puts 2027 sales 21% above 2026. What it may be missing sits in the July call, not the results release: a chief executive capping his own market, and a weighting that turns the guidance into a plateau nobody wrote down.
The evidence against us: positive ratings rose from 60.0% to 70.6% since May, though the panel shrank from 20 analysts to 17. And a second large cloud customer has passed correlation; a production ramp there would move share for the first time.
What analysts say, and the evidence
The evidence under this - 6 claims
- No independent measurement of total semiconductor capital equipment spending was reachable: semi.org returns HTTP 403 to curl and to an agent fetch alike, so the denominator in this ratio rests entirely on a forecast Teradyne cites and does not name.
- The same executive calls the relationship 'correlation, not causation' with about three quarters of lag, which means the ratio measured within a single year mixes two different vintages of spending.
- 2026-10-28editedthe Q3 2026 call gave the test share for the first eight months of 2026 as 8.1%, against 6.4% for 2025
- The same remarks argue the opposite for the medium term: advanced packaging raises test intensity per die because one latent defect destroys a more valuable multi-chip package, and both package volumes and dies per package are said to keep rising through the end of the decade.
- A chief executive guiding down the ratio his own market depends on has an incentive to set an expectation he can beat, which cuts against reading it as a ceiling.
- Management framed the same sentence as a strengthening: it said the second-half outlook had improved and the range narrowed on increased visibility, so the plateau is a plateau at a level it had previously feared would be lower.
- Teradyne's revenue has landed above the midpoint of its own guidance in each of the last eight quarters and above the top of the range in the last three, so its framework has recently been set to be exceeded.
- 2026-09-26editedcounter-evidence corrected: it said revenue beat the high end of guidance in each of seven quarters by 0.2% to 10.9%; those are the beats against analysts' consensus, and against the company's own range revenue topped the high end only in the last three of eight quarters (C48)
- Fast follower is the phase after correlation, but the capacity buys are dated to the second half of 2027, so no second-customer production ramp is running in 2026.
- No volume, term or counterparty was disclosed for any of the programmes, and the same executive said Teradyne's compute share has been diluted by the market growing faster than its wins.
- 2026-10-28editedthe Q3 2026 call restated each customer's dual-vendor phase; the second hyperscaler is past qualification with capacity buys expected in the second half of 2027
- Advantest's own results show consolidated sales up 44.7% in its year to March 2026 at a 44.2% operating margin, so the leader is compounding at least as fast in the same market.
- The chief executive volunteered that Advantest 'believed that they were going to continue to gain share in SoC' and agreed only about memory, which is a supplier conceding the point on the segment that matters.
- He also said the share change should start to show in 2027 and is 'a socket by socket thing', so flat today is not a forecast of flat.
What could go wrong, and how we would know
The risks
- The cycle turns. From 2021 to 2023 sales fell 27.7% and operating profit fell from $1.20B to $501.1M.
- A handful of buyers. One customer was 19% of 2025 sales, and two others 12% and 10%.
- Trade rules. Export restrictions already limit it where some rivals are not bound by them, its annual report says.
- Customers build their own. Several make their own testers; Teradyne bought Infineon's in-house test team in 2025.
The evidence under this - 6 claims
- The same remarks argue the opposite for the medium term: advanced packaging raises test intensity per die because one latent defect destroys a more valuable multi-chip package, and both package volumes and dies per package are said to keep rising through the end of the decade.
- A chief executive guiding down the ratio his own market depends on has an incentive to set an expectation he can beat, which cuts against reading it as a ceiling.
What we still do not know
- Test's real slice of chipmakers' equipment spending, measured by anyone other than Teradyne's chief executive.
- Who the 19% customer is, and whether it overlaps the two customers at 12% and 10%.
- Whether the second large cloud customer, now past testing with capacity buys expected in the second half of 2027, ramps earlier.
The evidence under this - 3 claims
- No independent measurement of total semiconductor capital equipment spending was reachable: semi.org returns HTTP 403 to curl and to an agent fetch alike, so the denominator in this ratio rests entirely on a forecast Teradyne cites and does not name.
- The same executive calls the relationship 'correlation, not causation' with about three quarters of lag, which means the ratio measured within a single year mixes two different vintages of spending.
- 2026-10-28editedthe Q3 2026 call gave the test share for the first eight months of 2026 as 8.1%, against 6.4% for 2025
- Fast follower is the phase after correlation, but the capacity buys are dated to the second half of 2027, so no second-customer production ramp is running in 2026.
- No volume, term or counterparty was disclosed for any of the programmes, and the same executive said Teradyne's compute share has been diluted by the market growing faster than its wins.
- 2026-10-28editedthe Q3 2026 call restated each customer's dual-vendor phase; the second hyperscaler is past qualification with capacity buys expected in the second half of 2027
What would change our view
- Unattractive→AttractiveThe level survives a down quarter, at its own normal multiple
It needs all of:
- ○Operating margin above 30% for a full year with a weak quarter in it, and sales above June 2026's three quarters running, none bought.checked at the next results, due by 27 Jan
- ○The shares fall to about 6.74 times sales, the company's own six-year normal, from 13.30 when we wrote this.about $192.43 today, 49% below the last close
- Reopens the researchTest's share of equipment spending structurally rises
It needs any of:
- ○Test's slice of chip-equipment spending rises above 9%, past the chief executive's own ceiling, and stays there through 2027.checked at the next results, due by 27 Jan
- ○Teradyne's stated share of testing for complex processor chips rises above 40% in its 2027 results.checked at the next results, due by 27 Jan
We would also be wrong if
- The two smaller businesses, product testing and robots, grow to a quarter of pre-tax profit, making this more than one business.
What is coming
- 27 October 2026 Q3 2026 results: Whether Q4 is guided above Q2's $1,329.0M, gross margin holds above the 58% floor, and a compute dual-vendor programme reaches production ramp rather than correlation.
Every dated event ahead
- 27 October 2026 Q3 2026 results: Whether Q4 is guided above Q2's $1,329.0M, gross margin holds above the 58% floor, and a compute dual-vendor programme reaches production ramp rather than correlation.
- January 2027 TSMC Q4 2026 results and its 2027 capital-spending guidance: Whether the largest chip buyer of test keeps raising equipment spending after US$52-56B in 2026, the base the test-share argument multiplies.
- February 2027 Q4 2026 results and the 2027 outlook: Whether the second half matched the first, as the 50-52% weighting implies, and whether the 2027 compute surge stays in the first half.
Since we wrote this
- 28 OctEarningsFourth quarter guided to $1.30B, below June's $1.33BThe call dated the second hyperscaler's buys to late 2027 and margins are guided down, so the view holds.
- 27 OctEarningsTeradyne sold $1,306M, just above its guide; fourth-quarter guide unread. View heldQ3 2026 results.
- 11 OctThesis reviewCo-packaged optics test kept out of the Robotics betThe owner's rule of 2026-10-11: a thesis carries one bet, so each early line is folded into it or said not to be.
- 6 OctNewsTeradyne takes a stake in Bright Machines; size undisclosed, robots go in its factories. View heldTeradyne invests in Bright Machines, an AI-infrastructure contract manufacturer.
- 5 OctThesis reviewThe long case now sits under the verdictAfter the five-name pilot, the owner asked for the long case on every thesis.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 26 SepThesis reviewGuidance record corrected: above the top of the range only three timesThe thesis said revenue beat the top of Teradyne's guidance seven quarters running. The releases show it beat the midpoint eight quarters running and the top only in the last three.
- 13 SepThesis writtenUnattractive, medium convictionFirst look at the name: sales had doubled in a year, but on the market growing rather than on Teradyne winning share.
Also checked 3 times with nothing new to act on.
Sources 25
Sources - 25
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-19Item 1 business and product descriptions, the Competition section, customer and receivable concentration, the segment reorganisation into Semiconductor Test / Product Test / Robotics, the Infineon and Quantifi acquisitions, the MultiLane joint venture, the statement that no backlog figure is meaningful, the trade-restriction language, and the audited 2023-2025 income statementsbacksC01C04C10C11C12C22C24C30C39C42C44C45C47C50C54C55C59
- S02primarycompany filing2026-07-31segment revenue and pre-tax income for the quarter and half year, revenue by country, the balance sheet, share count, buyback detail, the Technoprobe carrying value against its quoted market value, cash flow, and the revolving credit facility notebacksC02C03C20C26C27C30C32C37C38C39C40C41C42C43C47C49C50C56C61
- S03primarycompany filing2026-07-28the headline quarter, the segment split as management framed it, and the third-quarter revenue and earnings guidancebacksC02C26C33C34C48
- S04primarytranscript2026-07-29Second quarter 2026 earnings call transcript, held July 29, 2026the prepared remarks' figures with no filed home - the Semi Test product split, the AI share of revenue, the memory book-to-bill - and, in the question section, the chief executive's own decomposition of test as a share of chip-equipment spending, his statement on system-on-chip share, the four-phase dual-vendor qualification, and the chief financial officer's first-half revenue weighting. The 37% share figure is the analyst's, not management's - Teradyne states no share number itself - and is recorded that way in the ledger
if you look back to 2023, it was down near 4% of the total CapEx spent on test equipment. And by 2025, that had gone up to 7% The first five months of 2026, it is 8% of total semi CapEx has been on test equipment. So it is kind of this rocket trajectory. ... we think it is going to settle somewhere in the 7% to 9% range, but it is not going to continue to go up from there. ... [Timothy Arcuri, UBS] I get your test share for this year is up just a touch, but it is still basically flat sort of in the 37% range. ... [Gregory Smith] I think in terms of SoC, it is probably going to be pretty flat, maybe a slight incremental gain for us year-on-year. ... our share in the compute segment is actually stabilizing and starting to inflect upward. ... it is possible that next year that could revert down to, you know, 6, 7% or so. ... In Q1, we received our first order from a merchant GPU customer. Which was delivered in the second quarter. ... we completed correlation at a second hyperscaler ... driving our book-to-bill ratio in the quarter over two. ... we are updating our first-half weighted revenue to 50% to 52% of annual revenue.
backsC06C07C08C14C15C19C21C23C28C34C57C61 - S05primarycompany filing2026-02-03the 2025 full-year result and the first-quarter 2026 guidance issued with it - the annual guidance moment the source audit exists to catchbacksC48
- S06primarycompany filing2026-08-10the $1.0 billion aggregate revolving commitment, the August 7, 2031 maturity and the $100 million letter-of-credit sublimitbacksC40
- S07secondaryvendor data2026-09-13Research pack, frozen at content/TER/research-pack.jsonprice, enterprise value, market capitalisation, the volatility bucket, the analyst rating series and its shrinking panel, price-target drift, insider transactions and their base rate, short interest, consensus estimates, the vendor multiple series against its own six-year medians, and the filer's own XBRL statement series back to 2008backsC30C32C33C43C46
- S08primarycompany filing2026-04-16the driver's rate measured by the buyer rather than the seller, and the fact that capacity in 12-inch equivalent wafers was roughly flat while that spending rose a third
Our capital expenditures in 2023 , 2024 and 2025 were NT$ 949,817 million, NT$ 956,007 million and NT$ 1,272,411 million (US$ 40,895 million, translated from a weighted average exchange rate of NT$ 31.11 to US$1.00), respectively. Our capital expenditures in 2026 are expected to be between US$52 billion and US$56 billion
backsC16C17C18 - S09primarycompany filing2026-04-27Advantest Corporation, FY2025 Consolidated Financial Results (year ended March 31, 2026) Advantest Corporationthe market leader's own audited scale and margin, and its own statement that the tester market is expected to reach its largest scale on AI-related complexity and volume. The quoted paragraph is the parent-company section; the consolidated figures cited in the ledger come from the Consolidated Financial Results table on page 1 of the same document, which extracts from the PDF as an unpunctuated run of digits and cannot be quoted as prose
Against the backdrop of a significant expansion in demand for testers used in high performance semiconductors, primarily for AI-related applications, net sales reached (Y) 988,232 million (46.8% increase in comparison to the corresponding period of the previous fiscal year), operating income amounted to (Y) 405,874 million (86.7% increase in comparison to the corresponding period of the previous fiscal year)
backsC09C13 - S10primarycompany filing2026-02-17a competitor's own naming of the same two-company core, its serviceable market of about $3 billion, and its statement that many orders are cancellable with limited or no penalty
Cohu's primary competitors include established global players in semiconductor test and inspection, such as Advantest, Teradyne, Hon Precision, KLA, and other Asia-based manufacturers.
read, but no claim rests on it - S11primarycompany filing2026-02the buyer's own account of why it standardises on a tester platform - the switching-cost mechanism stated by the party paying it - and its filed capital spending, which rose 21.6% in 2025
Test equipment is generally a more capital-intensive activity than packaging, and test equipment tends to have longer delivery lead times than most types of packaging equipment. We focus our capital expenditures on standardized tester platforms to maximize test equipment utilization where possible. In some cases, our customers will consign test equipment to us.
backsC05C25 - S12primarycompany filing2025-12-17the evidence that a named Product Test rival sells into design laboratories rather than production floors, which is why it is recorded as a segment-only, immaterial-perimeter candidateread, but no claim rests on it
- S13primarycompany filing2025-11-12a named Robotics rival's own competitor set, which contains no test-equipment company at all; and the Intelligent Devices segment operating margin, a mature automation-hardware benchmark for the Robotics betbacksC58
- S14primarycompany filing2024-02-22the audited 2021, 2022 and 2023 income statements - the last completed cycle, peak to trough, in filed figuresbacksC31C50
- S15secondarynews2026-09-13News digest for TER, trusted-source filter applied118 articles fetched, 105 dropped by the source allowlist, two developments kept: the Q2 2026 print and the September 1 UltraFLEXplus instrument launchread, but no claim rests on it
- S16primarycompany filing2024-07-24the third-quarter 2024 revenue guidance of $680 million to $740 millionbacksC48
- S17primarycompany filing2024-10-23third-quarter 2024 revenue of $737 million and the fourth-quarter 2024 guidance of $710 million to $760 millionbacksC48
- S18primarycompany filing2025-01-29fourth-quarter 2024 revenue of $753 million and the first-quarter 2025 guidance of $660 million to $700 millionbacksC48
- S19primarycompany filing2025-04-28first-quarter 2025 revenue of $686 million and the second-quarter 2025 guidance of $610 million to $680 millionbacksC48
- S20primarycompany filing2025-07-29second-quarter 2025 revenue of $652 million and the third-quarter 2025 guidance of $710 million to $770 millionbacksC48
- S21primarycompany filing2025-10-28third-quarter 2025 revenue of $769 million and the fourth-quarter 2025 guidance of $920 million to $1,000 millionbacksC48
- S22primarycompany filing2026-04-28first-quarter 2026 revenue of $1,282 million and the second-quarter 2026 guidance of $1,150 million to $1,250 millionbacksC48
- S23primarycompany filing2026-03-27beneficial ownership as of March 13, 2026, executive officers and their prior roles, the 2025 variable cash and profit bonus plans with their goals and payouts, the PSU design and the 2026 change to an S&P 500 TSR comparisonbacksC51C52C53
- S24primarytranscript2026-10-28Third quarter 2026 earnings call transcript, held October 28, 2026the fourth-quarter guide's gross margin and operating-margin ranges, the dual-vendor phases by customer, the test share of semiconductor capital spending for the first eight months of 2026, the 1.4 product-test book-to-bill and the 2027 quarter-by-quarter growth statement
The portion of semi CapEx associated with test is increasing as well. From 6.4% in 2025 to 8.1% for the first 8 months of 2026. ... Our first hyperscaler customer is in the mature phase, and we expect our next major ramp in the middle of 27. Our merchant GPU customer is in the fast follower phase, with multiple devices in development, and we expect to see additional revenue in 2027. Our second hyperscaler customer has past qualification and is now in fast follower with an expectation of capacity buys in the second half of 27. We also have a third hyperscaler in initial qualification. ... Gross margins are expected to be in the range of 57.5% to 58.5% ... The non GAAP operating profit rate is expected to be between 28-30%. Our Q4 guidance midpoint of $1.3 billion represents growth of approximately 20% over the last year's fourth quarter. Which would bring full year revenue to approximately $5.2 billion up over 60%. ... [Michelle Turner] we expect the quarters next year to represent growth quarter on quarter when you look at next year.
backsC07C14C34C62 - S25primarycompany filing2026-10-28Q3 2026 revenue of $1,306M, the fourth-quarter 2026 guide of $1,250M to $1,350M and non-GAAP earnings of $1.95 to $2.25 a sharebacksC34C62