Cohu
It makes the machines that move chips through testing and hold them at the right temperature, plus the contacts and parts those machines wear out.
Our viewUnattractive Medium conviction
The scores
The profile · 1-4
Growth3/4 ↑
Sales up 38% in Q2 on a real new driver. Not 4: a rebound from a 55% fall, FY2025 lagged its market, one product.
Business quality2/4 ↑
Kits sell into 25,000 installed systems, but in the handler carrying the case Hon Precision is incumbent at about 50% margins.
Financial position4/4 →
About $194M of net cash, positive operating cash flow, capital spending near 2% of sales, no maturity before 2031.
Execution record3/4 ↑
Beat its Q2 revenue guide and raised the year twice. Not 4: missed per-share estimates in two of the last three prints.
Per share capture2/4 ↓
The convertible lands about 10% more shares at today's price, above what the capped call covers, and buybacks are too small to offset it.
Potential2/4 ↑
The upside is expected, not committed: $190M of qualified opportunity, no order figure filed since the ramp, and capacity that depends on suppliers.
Risk profile3/4 →
One serious concentration: the growth line leans on few buyers in a socket one incumbent holds. Otherwise spread, no binary event.
Valuation2/4 ↑
At 10 times its record year's operating profit the price still needs the AI ramp to work and the cycle to stay near its peak.
Evidence quality3/4 →
Financials reconcile to the filings and SEMI backs the market. Not 4: every computing figure is management's; the rival's are press and vendor.
The bet · 2-3 years
Priced for success0.8×
If it works 0.8× · if it fails you keep about 42%.
The Verdict
We rate Cohu unattractive. It makes the machines that test finished chips, and its new handler for hot-running AI processors is winning its first customers. But a Taiwanese rival already holds most of those sockets at far higher profits, and Cohu's sales more than halved in the last downturn, so the business is mixed. The price is full: it assumes the AI ramp works and the cycle stays near its best year, so a miss would hurt. As a bet it is fragile: even if the AI line hits management's capacity target, the company is worth little more than it costs, and if it fails well under half is left.
The long caseIf AI chips keep running hotter, buyers want a second thermal handler supplier, and Cohu could win enough sockets to change what it earns.
The evidence under this - 7 claims
C01facthigh confidence
Cohu sells test handlers that move packaged chips into a tester's sockets and hold them at a set temperature while they are tested, the contactors and probe cards that make the electrical connection, its own testers, optical inspection systems, analytics software and service. It reports one segment.
C09management claimmedium confidence
Management raised its FY2026 high-performance computing revenue estimate to about $120M, from $100M-$110M a quarter earlier, $80M-$100M before that and $60M-$85M at the start of the year, almost all of it the Eclipse thermal handler.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ1 2026 earnings call transcript, held April 30, 2026primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primary
against this
- The figure is the company's own and is not reported as a line in any filing: Cohu reports one segment and splits revenue only into systems and recurring, so nothing filed will confirm or refute it.
- The $26M single-customer Eclipse order taken early in Q3 is about a quarter of the whole year's estimate, so the estimate leans on few buyers.
- 2026-10-28editedthe Q3 release raised the fiscal-2026 computing revenue outlook to about $120M
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
C17factmedium confidence
Hon Precision, the Taiwanese test-handler and thermal-control maker, more than doubled 2025 revenue to NT$30.27B with operating income of NT$15.00B, a margin of about 50%, and grew first-half 2026 revenue another 91.6% to NT$24.52B.
C06estimatehigh confidence
At $46.26 the whole enterprise costs $2.28B: 36 times the $62.8M seven-year average operating income before amortisation, and 9.6 times the $236.9M Cohu earned in FY2021, its best year.
sourceResearch pack, COHU, 2026-10-06secondaryDaily closes to 22 October 2026 (FMP end-of-day prices)secondary
against this
- The vendor's enterprise value nets off only part of the $498.2M of cash and investments Cohu reports. On the filed balance sheet the net cash is about $194M and enterprise value nearer $1.99B, which takes the multiples to 32 and 8.4.
- If the computing handler business is a lasting step up, the seven-year average understates what Cohu now earns mid-cycle, and 36 times is a fact about the window rather than the business.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
C05estimatehigh confidence
Over FY2019 to FY2025 Cohu earned an average of $62.8M a year of operating income before amortisation of purchased intangibles. Its best year, FY2021, was $236.9M.
derivedC04: the seven filed operating income figures sum to $179.9M and the seven amortisation charges to $259.9M, a total of $439.8M over seven years; FY2021 is $201.5M plus $35.4M.
against this
- Seven years is a choice. It starts at a trough year, FY2019, and contains two loss years at the 2024-2025 bottom. Dropping FY2019 lifts the average to $81.1M, which moves the multiple from 57 to 44 rather than removing it.
- Adding back amortisation flatters the figure: the intangibles were bought with real money, mostly Xcerra in 2018, and the charge is a cost of the business as acquired. Struck on GAAP operating income the seven-year average is $25.7M and the multiple near 140.
C54estimatemedium confidence
Computing handler revenue near $105M in FY2026 would be about a sixth of the roughly $612M year, leaving about $507M for everything else - below Cohu's FY2019 to FY2025 average revenue of $630M.
derivedC08's 35% growth on C34's $453.0M is about $612M; C09's midpoint is $105M; the seven filed revenue years in C03 average $630.1M.
against this
The core is mid-recovery: management puts test utilisation at 80% and an upcycle at six quarters, so the rest of the business should keep growing into 2027.
The company
Before a chip ships, it is tested. A handler picks each packaged chip, presses it into a socket wired to a tester, holds it at a set temperature while the test runs, and sorts it good or bad. A processor tested too hot is scrapped though good; too cool, it may fail later.
Cohu sells those handlers, the contacts in the socket that wear out, its own testers for simpler chips, inspection machines and software. Teradyne, a much larger tester maker, names Cohu as a competitor.
The evidence under this - 3 claims
C01facthigh confidence
Cohu sells test handlers that move packaged chips into a tester's sockets and hold them at a set temperature while they are tested, the contactors and probe cards that make the electrical connection, its own testers, optical inspection systems, analytics software and service. It reports one segment.
C51facthigh confidence
Teradyne's own annual report names Cohu among its Semiconductor Test competitors, with Advantest and SPEA.
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
How it makes money
Two ways. Machines are sold when chip testers add capacity, and stop selling when they don't: total sales fell from $887.2M in 2021 to $401.8M in 2024. Steadier money comes from the contacts, kits, spares and service the installed machines keep needing, which were 60% of sales in 2025 and about half of the June quarter's.
The evidence under this - 3 claims
C22facthigh confidence
Recurring revenue - interface products, spares, kits sold apart from systems, software and service - was 49% of sales in 2023, 65% in 2024 and 60% in 2025; systems were the rest.
C53facthigh confidence
Systems revenue was $70.7M in Q2 2026, up 78% from $39.6M; non-systems revenue was $78.3M, up 15% from $68.0M.
C03facthigh confidence
Revenue peaked at $887.2M in FY2021, fell to $401.8M in FY2024, 54.7% lower, and recovered to $453.0M in FY2025. FY2019, the first full year after Cohu bought Xcerra in October 2018, was $583.3M.
Where the money comes from, stream by stream
| Stream | Share of sales (Q2 2026) | How it earns it |
|---|---|---|
| Non-systems | 52.6% | Contactors, probe cards, change kits, spares, service and software bought by the installed base, steadier through the cycle. |
| Systems | 47.4% | Test handlers, testers and inspection systems, bought when customers add capacity; this is where the computing handler sits. |
Who runs it
Luis Müller, a former Teradyne executive, has run Cohu since 2014 and steered it toward AI handlers; finance chief Jeffrey Jones has been there since 2007. Insiders own about 3%. Both sold shares near $46 in May outside any pre-set plan, and no insider bought in the year. The bonus pays for sales growth and profit; most long-term shares pay only if the stock beats small companies.
The evidence under this - 4 claims
C43facthigh confidence
Luis Müller, formerly of Teradyne, has been chief executive since December 2014; Jeffrey Jones has been chief financial officer since 2007; Klaus Ilgenfritz became chief product officer in January 2025 after running the test handler group.
C40facthigh confidence
Directors and executive officers owned 3.05% of Cohu at March 16, 2026; the chief executive held 464,672 shares and the chief financial officer 203,590. BlackRock held 14.58% and Vanguard 8.87%.
C42facthigh confidence
The annual bonus pays on year-over-year sales growth, four-year sales growth and profitability; it paid 32% of target for 2025 and nothing for 2024. Sixty per cent of executive equity vests on three-year shareholder return against the Russell 2000. The chief executive's salary is $715,000.
C39facthigh confidence
In the year to 5 October 2026 no insider bought shares in the open market. Twenty-one sales totalled 238,772 shares; seven, 101,273 shares, were outside a Rule 10b5-1 plan, including the chief executive's 45,154 at $46.62 and the chief financial officer's 13,000 at $46.84 on 22 May. The chief executive then sold 104,692 more under a plan between 21 September and 2 October at $60 to $73.
Why it wins, and who could beat it
Why customers come back
Each new chip design needs its own set of socket contacts, and they wear out. So every one of Cohu's more than 25,000 installed machines keeps buying parts as long as it runs, and a test floor that runs well does not swap machine makers mid-programme.
The evidence: from 2023 to 2024 machine sales fell about 57%, but these recurring sales fell only about 16%. Steadier, but not immune.
The evidence under this - 3 claims
C22facthigh confidence
Recurring revenue - interface products, spares, kits sold apart from systems, software and service - was 49% of sales in 2023, 65% in 2024 and 60% in 2025; systems were the rest.
C23inferencemedium confidence
The interface kits are a switching cost that sells back into the installed base: contactors are specific to each chip design and wear out, so every new device on an installed handler buys new ones. In dollars recurring revenue fell about 16% from 2023 to 2024 while systems fell about 57%.
derivedC22 applied to C03: recurring about $312M in 2023 and $261M in 2024; systems about $325M and $141M. C24 for the installed base the kits sell into; the 10-K says test interfaces are specific to individual device designs and need to be replaced frequently.
against this
- Recurring revenue still fell 16% in the downturn: it tracks how many devices are being tested, so it is a utilisation stream as much as a lock-in.
- The 10-K names other Asia-based manufacturers across its lines including interface, and contactors are made by many suppliers; nothing filed shows the kits command a premium.
C24management claimhigh confidence
Cohu's active installed base exceeds 25,000 systems at over 280 high-volume manufacturing sites, across 108 customers in 31 countries.
The threat
The AI handler sockets Cohu wants already have an owner. Its chief executive names Hon Precision as essentially the only rival there, the long-standing supplier to the firms that test AI chips. Hon more than doubled its 2025 sales, earned about 50% operating margins on them, and grew another 91.6% in the first half of 2026. Advantest, the biggest tester maker, sells handlers too.
The evidence under this - 5 claims
C17factmedium confidence
Hon Precision, the Taiwanese test-handler and thermal-control maker, more than doubled 2025 revenue to NT$30.27B with operating income of NT$15.00B, a margin of about 50%, and grew first-half 2026 revenue another 91.6% to NT$24.52B.
C18management claimhigh confidence
Cohu's chief executive names Hon Precision as primarily the single competitor for computing handler sockets, the long-standing supplier at the test subcontractors.
against this
- The chief executive allowed there is arguably a second competitor without naming it, and the 10-K lists Advantest, which sells handlers too.
- Hon Precision's strength is at the test subcontractors; chip designers and hyperscalers buying directly may be less tied to it.
C19factmedium confidence
Hon Precision primarily supplies test handlers for chips packaged by chip-on-wafer-on-substrate, the advanced packaging used for AI accelerators.
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
C52facthigh confidence
Cohu's 10-K names Advantest, Teradyne, Hon Precision, KLA and other Asia-based manufacturers as its primary competitors.
The rivals: TER, KLAC, 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 |
|---|---|---|---|---|
| Hon Precision thermal test handlersPriced against it: Listed abroad | High-performance computing thermal handlers | On sale | Listed abroad | |
| Advantest test handlers and testersPriced against it: Listed abroad | Test handlers outside computing, interface, testers, inspection, software and services | On sale | Listed abroad | |
| TER | Teradyne testersPriced against it: Same product | Test handlers outside computing, interface, testers, inspection, software and services | On sale | Same product |
| KLAC | KLAPriced against it: One segment only | - | On sale | One segment only |
How big it could get
How big this could get
SEMI, the industry's trade body, expects $15.3B of spending on chip test equipment in 2026, up 31.0%, and $20.8B by 2028. Most of that is testers Cohu does not sell. Cohu says about $3B is open to it, and counts about $850M a year of AI processor and memory spending at the customers it is chasing. It sold $453.0M in 2025.
The evidence under this - 4 claims
C15factmedium confidence
SEMI, the industry's trade association, says test equipment sales rose 55.3% in 2025 and forecasts a 31.0% rise to $15.3B in 2026 and $20.8B by 2028, citing AI devices' performance and reliability requirements.
sourceSEMI - Mid-Year Total Semiconductor Equipment Forecast, test…primarySEMI - Mid-Year Forecast, the 2028 test equipment lineprimary
against this
- SEMI's test line is dominated by the testers themselves, where Advantest and Teradyne lead and Cohu is small, so it measures Cohu's market only loosely.
- SEMI revised 2026 up sharply from its own December forecast: a two-year-out figure moves a great deal.
C46management claimlow confidence
Cohu puts its serviceable available market at about $3 billion.
against this
No outside body sizes handlers, interface and inspection together; the figure is the company's own and no method is given.
C10management claimmedium confidence
Cohu puts its annual computing opportunity pipeline at about $1.0 billion, up 18% from about $850M a quarter earlier: $230M of annual revenue potential is qualified and another $440M is in qualification, across 13 ranked customer programs, and seven more programs worth about $340M a year reach qualification milestones between late November and the first quarter of 2027; several hundred Eclipse systems are forecast for 2027.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ2 2026 earnings call - the pipeline by stageprimaryQ2 2026 earnings call - capacity and the HPC revenue it…primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primaryQ3 2026 earnings call transcript (FMP), held October 28, 2026primary
against this
- A pipeline is the company's tally of what its targeted customers spend on this class of equipment, not what Cohu will win: management itself calls it a served market rather than a total one.
- Three months earlier the same measure was $750M including about $100M of memory inspection, so part of the rise is reclassification and better visibility rather than new demand.
- 2026-10-28editedthe release puts the annual computing pipeline at about $1.0B, up 18% from the quarter before; the split by stage is in slides and a call not yet read
- 2026-10-29editedthe Q3 call gave the split by stage ($230M qualified, $440M in qualification, 13 programs) the release left out; the count of qualified customers is still not stated
C34facthigh confidence
FY2025 sales were $453.0M, up 13%, at a 42.7% gross margin, with a GAAP net loss of $74.3M, $1.59 a share; Q4 2025 sales were $122.2M, up 30%.
Who has signed
Little is on file. One unnamed customer ordered $26M of AI handlers in July, mostly for delivery this quarter. The last reported orders on hand were $165.1M at December 2025, before the AI ramp, and many orders can be cancelled with little or no penalty.
The evidence under this - 2 claims
C14management claimmedium confidence
Early in Q3 2026 a single customer ordered $26M of Eclipse systems, mostly to ship in Q4.
against this
- One order is about a quarter of the year's computing estimate, which shows how few buyers the line rests on.
- Orders in this industry can be cancelled or rescheduled with limited or no penalty, by the 10-K's own account.
C45facthigh confidence
Backlog was $165.1M at December 27, 2025 against $138.0M a year earlier, and many orders can be cancelled or rescheduled with limited or no penalty.
The results
The numbers
Is it growing?Fast
Sales grew 34% over the twelve months to Sep 2026, faster than the 13% in the year to Dec 2025.
Is it profitable?Losing
Operating margin is −2.4%, down from 6.8% in the year to Dec 2023.
How steady is it?Swingy
Yearly sales growth has ranged from −37% to +39% over the last 5 years, with sales falling in 3 of them.
Does it earn good returns?Poor
Shareholders' money lost 2.8% a year, down from 24% in the year to Dec 2021.
Is it making cash?Solid
It turned 9.8% of sales into free cash, $55.9M over the last twelve months.
Is the balance sheet safe?Net cash
It holds $508M of cash against $302M of debt, $206M more cash than it owes.
Are shares being diluted?Creeping
The share count rose 12% in 4.8 years, and pay in shares costs 4.0% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-29. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Sep 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $887M | $813M | $636M | $402M | $453M | $569M |
| Sales growth | +39% | −8.4% | −22% | −37% | +13% | +34% |
| Profit | ||||||
| Operating margin | 23% | 15% | 6.8% | −18% | −15% | −2.4% |
| Net margin | 19% | 12% | 4.4% | −17% | −16% | −4.0% |
| What growth costs, per sales dollar | ||||||
| Research and development | 10% | 11% | 14% | 21% | 20% | 18% |
| Selling, general and admin | 14% | 16% | 21% | 32% | 27% | 23% |
| Pay in shares | 1.6% | 1.8% | 2.7% | 5.2% | 5.1% | 4.0% |
| Capital spending | 1.4% | 1.8% | 2.5% | 2.6% | 4.6% | 1.8% |
| Cash | ||||||
| Cash from operations | $97.9M | $113M | $101M | $2.78M | $31.7M | $66.2M |
| Free cash flow | $85.9M | $98.1M | $85.4M | −$7.86M | $10.7M | $55.9M |
| Balance sheet | ||||||
| Cash and short-term investments | $380M | $386M | $336M | $262M | $484M | $508M |
| Debt | $118M | $79.0M | $40.6M | $8.80M | $296M | $302M |
| Net cash | $262M | $307M | $295M | $253M | $188M | $206M |
| Returns and shares | ||||||
| Return on capital | 19% | 9.6% | 2.7% | - | - | - |
| Shares (diluted) | 48.5M | 48.8M | 48.0M | 46.9M | 46.7M | 54.4M |
What growth costs
Most of Cohu's spending on research, selling and running the company stays about the same whatever it sells, so as sales recover, much of each extra dollar becomes profit: management expects about 40% of the next quarter's added sales to. The same fixed costs turned the 2024 slump into a $71.7M operating loss.
The evidence under this - 3 claims
C55inferencemedium confidence
Operating expenses are a mostly fixed base: management expects about 40% of the Q3 sequential revenue increase to reach operating profit, so margins widen quickly as sales recover.
derivedC32 for the cost base that held in dollars through the downturn; C08 for the Q3 guide of $170M with operating expenses of about $54M against $52.7M in Q2; the chief financial officer's statement of about 40% conversion.
against this
- The same fixed base turned FY2024's 37% revenue fall into a $71.7M operating loss; leverage cuts both ways.
- Rising component costs and computing hiring could absorb part of the drop-through.
C32facthigh confidence
R&D rose from 10.4% of sales in FY2021 to 20.4% in FY2025 and 18.6% over the four quarters to June 2026; selling and administration from 14.3% to 27.3% and 25.4%. R&D was $97.2M over those four quarters, above FY2021's $92.0M in dollars.
C04facthigh confidence
GAAP operating income from FY2019 to FY2025 ran -$52.3M, $3.3M, $201.5M, $125.6M, $43.3M, -$71.7M and -$69.8M. Amortisation of purchased intangibles charged against it ran $35.4M to $39.6M a year.
The last report (30 Jul)
Quarter ended June 2026, filed figures.
38%
Sales growth
against the same quarter a year earlier
19.1%
Growth from the quarter before
against 2.4% the quarter before that
0¢
Operating profit
per dollar of sales
The September quarter brought $173.0M of sales, up 37%, against a guide of about $170M. Orders rose 126%, with systems up 318%, and Cohu raised this year's computing revenue estimate to about $120M from up to $110M.
It guides the fourth quarter up to $178M and puts its pipeline of computing sales at about $1.0B, of which $230M a year is qualified. On the call it said it will build to ship about three times as many AI handlers in 2027. Even so, only 7.4 cents of each sales dollar was operating profit on the accounting basis.
What it did to the thesisThe quarter and the call leaned against our doubt: a bigger qualified pipeline, triple the capacity and a 2027 computing line above $200M. But our routes also need a far higher operating margin and more qualified customers, and neither is shown, so the view stays Unattractive.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.26 against $0.14 expected; $0.00 as filed, the gap being what the company adds back.
The evidence under this - 5 claims
C56facthigh confidence
Cohu's Q3 2026 net sales were $173.0M, up 37% from a year earlier, with gross margin 45.5%, a GAAP operating income of $12.8M and non-GAAP earnings of $0.43 a share; bookings rose 126% (systems up 318%, recurring up 36%), estimated test-cell utilisation rose to 82%, and cash and investments were $507.9M with no shares repurchased in the quarter.
against this
GAAP operating income of $12.8M is 7.4% of sales, well under the 15% company margin route P1 requires, and non-GAAP earnings exclude $7.2M of amortisation of purchased intangibles and share pay.
C08management claimhigh confidence
Cohu guided Q3 2026 sales to about $170M, plus or minus $7M, and reported $173.0M; it guides Q4 2026 to $178M, plus or minus $7M, up 46% from a year earlier and 3% above the third quarter.
sourceQ2 2026 results release, Exhibit 99.1 to Form 8-KprimaryQ2 2026 earnings call transcript, held July 30, 2026primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primary
against this
- Q2 came in at $149.0M against a $144M midpoint, so the guide has recently been conservative.
- Management says the computing handler line is capacity-bound through Q4, so the flat fourth quarter is a supply ceiling rather than a demand reading, and the guide carries no information about 2027.
- 2026-10-28editedthe guided quarter was reported at $173.0M and the fourth quarter is now guided up to $178M, not flat
C09management claimmedium confidence
Management raised its FY2026 high-performance computing revenue estimate to about $120M, from $100M-$110M a quarter earlier, $80M-$100M before that and $60M-$85M at the start of the year, almost all of it the Eclipse thermal handler.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ1 2026 earnings call transcript, held April 30, 2026primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primary
against this
- The figure is the company's own and is not reported as a line in any filing: Cohu reports one segment and splits revenue only into systems and recurring, so nothing filed will confirm or refute it.
- The $26M single-customer Eclipse order taken early in Q3 is about a quarter of the whole year's estimate, so the estimate leans on few buyers.
- 2026-10-28editedthe Q3 release raised the fiscal-2026 computing revenue outlook to about $120M
C10management claimmedium confidence
Cohu puts its annual computing opportunity pipeline at about $1.0 billion, up 18% from about $850M a quarter earlier: $230M of annual revenue potential is qualified and another $440M is in qualification, across 13 ranked customer programs, and seven more programs worth about $340M a year reach qualification milestones between late November and the first quarter of 2027; several hundred Eclipse systems are forecast for 2027.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ2 2026 earnings call - the pipeline by stageprimaryQ2 2026 earnings call - capacity and the HPC revenue it…primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primaryQ3 2026 earnings call transcript (FMP), held October 28, 2026primary
against this
- A pipeline is the company's tally of what its targeted customers spend on this class of equipment, not what Cohu will win: management itself calls it a served market rather than a total one.
- Three months earlier the same measure was $750M including about $100M of memory inspection, so part of the rise is reclassification and better visibility rather than new demand.
- 2026-10-28editedthe release puts the annual computing pipeline at about $1.0B, up 18% from the quarter before; the split by stage is in slides and a call not yet read
- 2026-10-29editedthe Q3 call gave the split by stage ($230M qualified, $440M in qualification, 13 programs) the release left out; the count of qualified customers is still not stated
C58management claimmedium confidence
On the Q3 2026 call management gave a preliminary 2027 view: sales up 30% to 40%, gross margin 46% to 48%, quarterly operating costs of $54M to $56M and about half of each added sales dollar reaching operating income; the chief financial officer put computing revenue at the higher end of $200M to $250M. One computing customer was more than 10% of Q3 sales.
against this
- It is a preliminary view that depends on qualification timing and on suppliers ramping with Cohu; management itself said it has no view of the fourth quarter of 2027.
- The margin arithmetic is on the company's adjusted basis, which leaves out amortisation of purchased intangibles and share pay; the GAAP margin is lower, 7.4% in Q3 2026.
What management did with the money
It bought Xcerra in 2018, a small software firm in 2025, and repurchased little stock. In 2025 it borrowed $287.5M through bonds that turn into shares, which at today's price means about a tenth more shares later.
Insiders in 2026Q3: 2 purchases and 9 sales, from their Form 4 filings.
The evidence under this - 3 claims
C44facthigh confidence
Cohu bought Xcerra in October 2018 and Tignis, an analytics software company, in January 2025, and raised the convertible notes in September 2025 while buying a capped call; its term loan for Xcerra was repaid.
C28facthigh confidence
The $287.5M of 1.50% convertible notes due January 2031 convert at about $27.18 a share, about 10.58M shares; a capped call bought for $31.4M offsets dilution up to $41.02. Principal is settled in cash and the excess in shares, and the notes were convertible from July 28 to September 8, 2026 because the share price had cleared its trigger.
C31facthigh confidence
Diluted shares were 48.46M in FY2021 and 46.72M in FY2025. Cohu bought back 915,504 shares for $27.0M in 2024 and 432,288 for $8.6M in 2025, none in Q2 2026, and has paid no dividend since suspending it in May 2020; stock pay ran $24.0M over the four quarters to June 2026.
The price, and where we disagree
In short. Full. The price pays for the AI ramp working and for Cohu staying near its best year ever.
Share price
$73.73
close,
Company value for each $1 of cash profit
157.6
its 6-year normal 8.7
Price for each $1 of sales
6.6
its 6-year normal 2.5
The shares have nearly quadrupled in a year, and most of the rise came from investors paying more for each dollar of sales rather than from the sales themselves. That is the market pricing in the AI ramp.
The first check came on 28 October: sales of $173.0M, a computing outlook raised to about $120M and a fourth quarter guided up to $178M; the call added a 2027 view of 30% to 40% growth. The next is the meeting with investors on 10 November. Management says test floors are about 82% busy and that upturns usually last about six quarters.
The evidence under this - 4 claims
C56facthigh confidence
Cohu's Q3 2026 net sales were $173.0M, up 37% from a year earlier, with gross margin 45.5%, a GAAP operating income of $12.8M and non-GAAP earnings of $0.43 a share; bookings rose 126% (systems up 318%, recurring up 36%), estimated test-cell utilisation rose to 82%, and cash and investments were $507.9M with no shares repurchased in the quarter.
against this
GAAP operating income of $12.8M is 7.4% of sales, well under the 15% company margin route P1 requires, and non-GAAP earnings exclude $7.2M of amortisation of purchased intangibles and share pay.
C08management claimhigh confidence
Cohu guided Q3 2026 sales to about $170M, plus or minus $7M, and reported $173.0M; it guides Q4 2026 to $178M, plus or minus $7M, up 46% from a year earlier and 3% above the third quarter.
sourceQ2 2026 results release, Exhibit 99.1 to Form 8-KprimaryQ2 2026 earnings call transcript, held July 30, 2026primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primary
against this
- Q2 came in at $149.0M against a $144M midpoint, so the guide has recently been conservative.
- Management says the computing handler line is capacity-bound through Q4, so the flat fourth quarter is a supply ceiling rather than a demand reading, and the guide carries no information about 2027.
- 2026-10-28editedthe guided quarter was reported at $173.0M and the fourth quarter is now guided up to $178M, not flat
C13management claimmedium confidence
Management estimated semiconductor test utilisation at about 80% at the end of Q2 2026, which it calls typically the turning point for its core customers' capital spending, and at 82% at the end of Q3 2026; it puts an ordinary upcycle at about six quarters. By market at the end of Q3, computing was about 84%, industrial 83% and automotive 80%.
sourceQ2 2026 earnings call - test utilisationprimaryQ2 2026 earnings call - the length of an upcycleprimaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primaryQ3 2026 earnings call transcript (FMP), held October 28, 2026primary
against this
- Utilisation is the company's estimate with no source given; no outside body publishes test-floor utilisation.
- The chief financial officer added that 2021 and 2022 ran above 80% for seven or eight quarters, so the last cycle ran longer than the norm he quotes.
- 2026-10-28editedthe Q3 release gives estimated test-cell utilisation at 82%, up 200 basis points from the second quarter
- 2026-10-29editedthe Q3 call gave utilisation by market and the refresh date was reset
C47facthigh confidence
The shares closed as high as $51.86 in 2021, fell to $24.06 in 2022 as orders turned, and hit $12.57 on 4 April 2025 before rising about 486% to $73.73 on 5 October 2026 and falling to $46.26 on 22 October.
sourceResearch pack, COHU, 2026-10-06secondaryDaily closes to 22 October 2026 (FMP end-of-day prices)secondary
- 2026-10-23editedthe price path now runs to the 22 October close
The betPriced for success
If it works
0.8×
from today's value, at the stated multiple
If it fails
keep 42%
of today's market value: the rest of the company, less its debt
How we read the multiple
Counting cash and debt, the whole company costs $2.28B. In the seven years since it bought Xcerra, Cohu averaged $62.8M a year of operating profit before charges for that purchase, with two loss years. Its best year, 2021, earned $236.9M.
So the buyer pays 36 times the average and about 10 times the record. For a more ordinary 15 times to hold on an average year, the average would have to rise to about $152M, in a business whose sales more than halved after the last record.
Against its own history. Far dearer than its own normal, at a time when profits have barely recovered, so the price leans on years that have not happened.
The evidence under this - 4 claims
C06estimatehigh confidence
At $46.26 the whole enterprise costs $2.28B: 36 times the $62.8M seven-year average operating income before amortisation, and 9.6 times the $236.9M Cohu earned in FY2021, its best year.
sourceResearch pack, COHU, 2026-10-06secondaryDaily closes to 22 October 2026 (FMP end-of-day prices)secondary
against this
- The vendor's enterprise value nets off only part of the $498.2M of cash and investments Cohu reports. On the filed balance sheet the net cash is about $194M and enterprise value nearer $1.99B, which takes the multiples to 32 and 8.4.
- If the computing handler business is a lasting step up, the seven-year average understates what Cohu now earns mid-cycle, and 36 times is a fact about the window rather than the business.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
C05estimatehigh confidence
Over FY2019 to FY2025 Cohu earned an average of $62.8M a year of operating income before amortisation of purchased intangibles. Its best year, FY2021, was $236.9M.
derivedC04: the seven filed operating income figures sum to $179.9M and the seven amortisation charges to $259.9M, a total of $439.8M over seven years; FY2021 is $201.5M plus $35.4M.
against this
- Seven years is a choice. It starts at a trough year, FY2019, and contains two loss years at the 2024-2025 bottom. Dropping FY2019 lifts the average to $81.1M, which moves the multiple from 57 to 44 rather than removing it.
- Adding back amortisation flatters the figure: the intangibles were bought with real money, mostly Xcerra in 2018, and the charge is a cost of the business as acquired. Struck on GAAP operating income the seven-year average is $25.7M and the multiple near 140.
C03facthigh confidence
Revenue peaked at $887.2M in FY2021, fell to $401.8M in FY2024, 54.7% lower, and recovered to $453.0M in FY2025. FY2019, the first full year after Cohu bought Xcerra in October 2018, was $583.3M.
C36estimatemedium confidence
At $46.26 the shares cost about 18 times the FY2028 per-share estimate and 25 times FY2027's, on adjusted figures that exclude amortisation and stock pay.
derived$46.26 over $2.56 and $1.85 in C35.
against this
If the computing ramp continues past FY2028 these are early-ramp earnings, and the multiple on a later year would be lower.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
Priced against its rivals
Only Teradyne 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, Hon Precision Inc.
The bet: what it pays if it works
This bet is on one business: Cohu's handler for AI processors. The rest of the company is counted separately at what the market has paid for it across a cycle, and covered well under half of the price this bet was struck at.
Even if the AI line reaches the yearly sales management sized its doubled factory for, the company is worth little more than it costs, short of the doubling a fair bet needs. If it stalls, shareholders keep an ordinary test-equipment business with spare cash.
| High-performance computing thermal handlers revenue in two to three years | Qualified computing customers needed | At the stated multiple |
|---|---|---|
| $105.00M | about 2 at 100% kept | 0.6× |
| $225.00M (the big win) | about 5 at 100% kept | 0.8× |
| $330.00M | about 7 at 100% kept | 0.9× |
To pay 2× (a Fair bet) it would need about $1.12B of High-performance computing thermal handlers revenue (about 25 qualified computing customers); to pay 5× (a Cheap bet), about $3.27B of High-performance computing thermal handlers revenue (about 73 qualified computing customers), at the same multiple.
Each row is High-performance computing thermal handlers alone, plus the rest of the company (Test handlers outside computing, interface, testers, inspection, software and services) counted at about $1.57B.
The steps it has to take
- Done Eclipse handlers qualified and shipping to computing customers
- In progress FY2026 computing revenue of $100M-$110M due 2027-02-28
- In progress Handler output up about 50% by the end of 2026 due 2026-12-31
- In progress A fifth computing customer qualified due 2027-03-31
- Not started Handler output doubled, supporting a $200M-$250M computing rate due 2027-07-31
- Not started Computing revenue above a $200M annual rate at a profit due 2027-12-31
The detail behind the label
Rating
Expensive
the big win's payoff against the 2× and 5× lines
Cheap bet at about
≈ $9.28
87% below the price (close, )
The option
0.6×
what High-performance computing thermal handlers is worth if it works, against what the price pays for it
The evidence under this - 6 claims
C11management claimmedium confidence
Management plans capacity to ship about 3 times as many Eclipse handlers in 2027 as in 2026, with output exiting 2026 at twice the June rate and still rising through June 2027. Its chief financial officer put 2027 computing revenue at the higher end of the $200M to $250M range, and the Q2 call said doubling output supports more than $200M, probably closer to $250M, a year.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ2 2026 earnings call - capacity and the HPC revenue it…primaryQ3 2026 earnings call transcript (FMP), held October 28, 2026primary
against this
- Management says the supply chain, not the factory, is what takes most of its attention, and that lead times and input costs for specialty components are rising: capacity is a plan with suppliers in it.
- Capacity is not demand. $230M of the pipeline is qualified, close to the $200M to $250M the capacity is being sized for, and the 2027 figure is a preliminary view that depends on qualifications landing on time.
- 2026-10-29editedthe Q3 call restated the plan as 3 times 2026 shipments and the chief financial officer put 2027 computing revenue at the higher end of $200M to $250M
C10management claimmedium confidence
Cohu puts its annual computing opportunity pipeline at about $1.0 billion, up 18% from about $850M a quarter earlier: $230M of annual revenue potential is qualified and another $440M is in qualification, across 13 ranked customer programs, and seven more programs worth about $340M a year reach qualification milestones between late November and the first quarter of 2027; several hundred Eclipse systems are forecast for 2027.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ2 2026 earnings call - the pipeline by stageprimaryQ2 2026 earnings call - capacity and the HPC revenue it…primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primaryQ3 2026 earnings call transcript (FMP), held October 28, 2026primary
against this
- A pipeline is the company's tally of what its targeted customers spend on this class of equipment, not what Cohu will win: management itself calls it a served market rather than a total one.
- Three months earlier the same measure was $750M including about $100M of memory inspection, so part of the rise is reclassification and better visibility rather than new demand.
- 2026-10-28editedthe release puts the annual computing pipeline at about $1.0B, up 18% from the quarter before; the split by stage is in slides and a call not yet read
- 2026-10-29editedthe Q3 call gave the split by stage ($230M qualified, $440M in qualification, 13 programs) the release left out; the count of qualified customers is still not stated
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
C54estimatemedium confidence
Computing handler revenue near $105M in FY2026 would be about a sixth of the roughly $612M year, leaving about $507M for everything else - below Cohu's FY2019 to FY2025 average revenue of $630M.
derivedC08's 35% growth on C34's $453.0M is about $612M; C09's midpoint is $105M; the seven filed revenue years in C03 average $630.1M.
against this
The core is mid-recovery: management puts test utilisation at 80% and an upcycle at six quarters, so the rest of the business should keep growing into 2027.
C27facthigh confidence
At June 27, 2026 Cohu held $498.2M of cash and investments against $304M of debt, $287.5M of it the 2031 convertible notes; shareholders' equity was $774.6M. Net cash was about $194M.
C06estimatehigh confidence
At $46.26 the whole enterprise costs $2.28B: 36 times the $62.8M seven-year average operating income before amortisation, and 9.6 times the $236.9M Cohu earned in FY2021, its best year.
sourceResearch pack, COHU, 2026-10-06secondaryDaily closes to 22 October 2026 (FMP end-of-day prices)secondary
against this
- The vendor's enterprise value nets off only part of the $498.2M of cash and investments Cohu reports. On the filed balance sheet the net cash is about $194M and enterprise value nearer $1.99B, which takes the multiples to 32 and 8.4.
- If the computing handler business is a lasting step up, the seven-year average understates what Cohu now earns mid-cycle, and 36 times is a fact about the window rather than the business.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
Where we disagree with the market
The market believes
Cohu is an AI test winner early in an upturn, which is why the shares nearly quadrupled in a year.
We believe
The AI handler demand is real, but the price already pays for it and for the upturn peaking at a record.
Analysts and the company lead with the AI handler, and the ramp is genuine. But it is about a sixth of this year's sales, the rest is an ordinary upturn, and the rival in that socket is bigger and far more profitable. Analysts already expect sales nearly to double by 2028.
We would be wrong if the AI line passes $200M a year at a healthy company profit margin while the rest of the business holds.
What analysts say, and the evidence
Analysts positive
100%
of 8, 1 Oct
Four months earlier
100%
of 7, 1 Jun
The evidence under this - 5 claims
C09management claimmedium confidence
Management raised its FY2026 high-performance computing revenue estimate to about $120M, from $100M-$110M a quarter earlier, $80M-$100M before that and $60M-$85M at the start of the year, almost all of it the Eclipse thermal handler.
sourceQ2 2026 earnings call transcript, held July 30, 2026primaryQ1 2026 earnings call transcript, held April 30, 2026primaryThird quarter 2026 results release (Form 8-K, Exhibit 99.1),…primary
against this
- The figure is the company's own and is not reported as a line in any filing: Cohu reports one segment and splits revenue only into systems and recurring, so nothing filed will confirm or refute it.
- The $26M single-customer Eclipse order taken early in Q3 is about a quarter of the whole year's estimate, so the estimate leans on few buyers.
- 2026-10-28editedthe Q3 release raised the fiscal-2026 computing revenue outlook to about $120M
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
C35factmedium confidence
Five analysts model FY2026 sales of $612.5M and $0.95 a share, and FY2027 $769.8M and $1.85; three model FY2028 at $887.4M and $2.56, almost double FY2025's sales.
C06estimatehigh confidence
At $46.26 the whole enterprise costs $2.28B: 36 times the $62.8M seven-year average operating income before amortisation, and 9.6 times the $236.9M Cohu earned in FY2021, its best year.
sourceResearch pack, COHU, 2026-10-06secondaryDaily closes to 22 October 2026 (FMP end-of-day prices)secondary
against this
- The vendor's enterprise value nets off only part of the $498.2M of cash and investments Cohu reports. On the filed balance sheet the net cash is about $194M and enterprise value nearer $1.99B, which takes the multiples to 32 and 8.4.
- If the computing handler business is a lasting step up, the seven-year average understates what Cohu now earns mid-cycle, and 36 times is a fact about the window rather than the business.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
C54estimatemedium confidence
Computing handler revenue near $105M in FY2026 would be about a sixth of the roughly $612M year, leaving about $507M for everything else - below Cohu's FY2019 to FY2025 average revenue of $630M.
derivedC08's 35% growth on C34's $453.0M is about $612M; C09's midpoint is $105M; the seven filed revenue years in C03 average $630.1M.
against this
The core is mid-recovery: management puts test utilisation at 80% and an upcycle at six quarters, so the rest of the business should keep growing into 2027.
What could go wrong, and how we would know
The risks
- A stronger rival. Hon Precision already supplies most AI test floors at about 50% operating margins, so Cohu may win only a few sockets.
- The cycle turns. Sales fell from $887.2M in 2021 to $401.8M in 2024 when customers stopped adding capacity, and the same fixed costs produced losses.
- Shares from the bonds. The convertible bonds hand about a tenth more shares to bondholders at today's price, so shareholders keep less of any success.
- Supply limits the ramp. Lead times and parts costs are rising, and management says suppliers, not its factory, are what take most of its attention.
The evidence under this - 6 claims
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
C17factmedium confidence
Hon Precision, the Taiwanese test-handler and thermal-control maker, more than doubled 2025 revenue to NT$30.27B with operating income of NT$15.00B, a margin of about 50%, and grew first-half 2026 revenue another 91.6% to NT$24.52B.
C03facthigh confidence
Revenue peaked at $887.2M in FY2021, fell to $401.8M in FY2024, 54.7% lower, and recovered to $453.0M in FY2025. FY2019, the first full year after Cohu bought Xcerra in October 2018, was $583.3M.
C29estimatemedium confidence
At $46.26 the notes would add about 1.2M net new shares, about 2.6% of the 47.0M outstanding, after the capped call, and pay out $287.5M of cash for principal.
derivedC28: 10.58M underlying shares times ($46.26 - $27.18)/$46.26 is 4.36M shares of excess value; the capped call returns ($41.02 - $27.18)/$46.26 of them, 3.17M; the net is 1.20M against 46,985,679 shares outstanding in C41.
against this
Nothing converts until holders choose to or the notes mature in 2031; an early conversion in the July to September window has not been reported, and the Q3 10-Q will say.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
C48management claimmedium confidence
Lead times are 13 to 14 weeks, memory and specialty component costs are rising, and conversations about passing those costs to customers have only just started.
C38facthigh confidence
Short interest was 8,628,062 shares, 18.7% of the free float and 8.62 days of average volume, at the 15 September 2026 settlement.
What we still do not know
- Whether any bondholders turned their bonds into shares and cash in the July to September window. The September-quarter filing will say.
- How many AI test sockets Hon Precision holds and whether Cohu is taking any. Nobody publishes the split.
- Who placed the $26M order, and whether it was a test house or a chip designer buying directly.
The evidence under this - 3 claims
C29estimatemedium confidence
At $46.26 the notes would add about 1.2M net new shares, about 2.6% of the 47.0M outstanding, after the capped call, and pay out $287.5M of cash for principal.
derivedC28: 10.58M underlying shares times ($46.26 - $27.18)/$46.26 is 4.36M shares of excess value; the capped call returns ($41.02 - $27.18)/$46.26 of them, 3.17M; the net is 1.20M against 46,985,679 shares outstanding in C41.
against this
Nothing converts until holders choose to or the notes mature in 2031; an early conversion in the July to September window has not been reported, and the Q3 10-Q will say.
- 2026-10-23editedre-struck at the 22 October close of $46.26, 37.3% below the 5 October close it was struck at
C21inferencemedium confidence
The computing handler socket already has a profitable incumbent: Hon Precision earns about 50% operating margins on twice Cohu's revenue while Cohu lost money in 2025, so Cohu is the challenger, not the leader, in the business carrying its case.
derivedC17 and C18 against C04 and C34: NT$30.27B is about twice Cohu's $453.0M at any recent exchange rate, Hon's operating margin was about 50% while Cohu's GAAP operating margin was negative.
against this
- Management says hyperscalers and chip designers are looking for a second source as processor power rises, which is exactly how an incumbent loses share; four qualified customers is evidence that some are switching.
- Hon's margins may reflect a capacity-short market rather than a lasting advantage; a shortage rewards whoever has the tools.
C14management claimmedium confidence
Early in Q3 2026 a single customer ordered $26M of Eclipse systems, mostly to ship in Q4.
against this
- One order is about a quarter of the year's computing estimate, which shows how few buyers the line rests on.
- Orders in this industry can be cancelled or rescheduled with limited or no penalty, by the 10-K's own account.
What would change our view
- Unattractive→AttractiveThe computing line proves itself; the price is already below the record year
It needs all of:
- ○The AI handler line reaches $200M of sales a year in 2027.checked at the next results, due by 10 Nov
- ○Cohu's profit margin passes 15% of sales for a full quarter, under ordinary accounting.checked at the next results, due by 10 Nov
- Reopens the researchCohu takes sockets from the incumbent
It needs:
- ○Cohu reports more than eight AI customers qualified, or names a test house moving from Hon Precision to its handler.watched as it happens
We would also be wrong if
- Sales stay above $170M every quarter through 2027 while car-chip demand is still weak.
What is coming
- 29 October 2026 Q3 2026 results: Whether sales land near the $170M guide, the computing estimate holds at $100M-$110M, the fourth quarter is still guided flat, and whether any notes converted.
Every dated event ahead
- 29 October 2026 Q3 2026 results: Whether sales land near the $170M guide, the computing estimate holds at $100M-$110M, the fourth quarter is still guided flat, and whether any notes converted.
- 10 November 2026 Investor day in New York: The long-term targets for computing revenue and margins, and the capacity plan's cost.
- 28 February 2027 Q4 2026 results and FY2026 10-K: Whether computing revenue reached $100M, the first backlog since the ramp, and whether a fifth computing customer qualified as planned in early 2027.
- July 2027 Malaysia handler capacity doubled: Whether output reaches the $200M-$250M computing revenue rate management sized it for.
Since we wrote this
Written
6 Oct
Price the thesis read
$73.73
5 Oct
- 29 OctEarningsCall put 2027 computing at $200M-plus; margin and customer count unprovenManagement stated a larger 2027 line and tripled capacity, but not the count of qualified customers or a 15% quarterly margin.
- 28 OctEarningsComputing outlook raised to $120M and fourth quarter guided upResults leaned against us, but our routes need $200M of computing and a 15% margin, which the quarter did not show.
- 23 OctPrice movePrice re-struck: 10 times the record year, view heldThe shares fell 37% with no company news; the computing line and margin tests come with the 28 October results.
- 6 OctThesis writtenUnattractive, medium convictionThe AI handler ramp is real, but an incumbent holds most sockets at 50% margins and the price pays for a record year.
Also checked 4 times with nothing new to act on.
Sources 26
Where this came from
Sources - 26
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 filing2026-02-17Item 1 products, the installed base, the named competitors and the $3 billion served-market estimate; sales split between systems and recurring revenue for three years; backlog at both year ends and its cancellation terms; customer concentration; the convertible notes, capped call and buybacks; executive officersbacksC01C03C04C22C24C25C26C28C31C43C44C45C46C52
- S02primarycompany filing2026-07-31systems and non-systems revenue for the quarter, sales by ship-to country, the balance sheet, the convertible notes' conversion window and terms, and customer concentrationbacksC02C07C25C26C27C28C33C53
- S03primarycompany release2026-07-30the quarter's income statement on GAAP and non-GAAP bases, cash and investments, and the Q3 2026 outlookbacksC02C08C27C31C49C50
- S04primarytranscript2026-07-30figures with no filed home: the computing pipeline by stage, the raised HPC revenue estimate, system orders by end market, test utilisation, the capacity plan, the $26M order; and the question section, where the chief executive named Hon Precision as the incumbent and the chief financial officer gave the typical length of an upcycle
Overall, this momentum reinforces an expanding high performance computing customer pipeline that we now estimate at approximately $850 million annually.
backsC08C09C10C11C25C27C30C48C50 - S04bprimarytranscript2026-07-30the chief executive's own account of who holds the computing handler sockets
I would say there is primarily a single competitor, which has been the, you know, forever supplier at the, test subcontractor.
backsC18 - S04cprimarytranscript2026-07-30what the capacity plan is sized to deliver, and the per-customer range of $30 million to $60 million a year
we should have the capacity to do more than $200 million, probably closer to $250 million by the beginning of next year
backsC10C11 - S04dprimarytranscript2026-07-30the chief financial officer's typical length of a utilisation upcycle
Yeah, we would say about 6 quarters is probably the typical average or call it a cycle.
backsC13 - S04eprimarytranscript2026-07-30the computing pipeline by stage
This includes about $190 million in qualified annual opportunity across 4 customers $250 million in active qualification across 5 customers, and approximately $445 million in early stage engagement across 10 additional customers.
backsC10 - S04fprimarytranscript2026-07-30system orders by end market
computing led with 46% of total system orders, representing an impressive 150% increase year over year driven by Eclipse growth in high performance computing.
backsC12 - S04gprimarytranscript2026-07-30management's estimate of test-floor utilisation
Estimated semiconductor test utilization and improved sequentially to 80% at the end of the second quarter.
backsC13 - S04hprimarytranscript2026-07-30the single-customer Eclipse order
We did get here in early Q3 a single customer order for $26 million again for our Eclipse 6 systems, for the HP market
backsC14 - S05primarycompany release2026-02-12FY2025 and Q4 2025 sales, gross margin and GAAP lossbacksC03C34
- S06primarytranscript2026-04-30the guidance a quarter earlier - the Q2 outlook, full-year growth and the HPC revenue range - so the Q2 call can be read as a change
We're increasing our full year 2026 revenue outlook for growth over last year of 20% to 25%.
backsC09C50 - S07primarycompany filing2026-04-02security ownership of directors, officers and 5% holders; the bonus metrics and payout; the equity mix and its performance condition; the chief executive's salarybacksC40C41C42C43
- S08primaryindustry data2026-07-14the size and growth of the test equipment market Cohu sells into, measured by the industry's trade association
After surging 55.3% in 2025, semiconductor test equipment sales are projected to increase 31.0% to $15.3 billion in 2026, marking a strong upward revision from SEMI's 2025 year-end forecast.
backsC15 - S08bprimaryindustry data2026-07-14the 2028 test equipment forecast
Growth is expected to continue through 2028, with test equipment reaching $20.8 billion and assembly and packaging equipment reaching $8.6 billion
backsC15 - S09secondarynews2026-07-13Hon Precision's first-half 2026 revenue growth, and that it sells chip handlers with active thermal control systems
Second-quarter revenue was up 100.36% from a year earlier to NT$13.79 billion. First-half revenue rose 91.6% year-on-year to NT$24.52 billion.
backsC17C20 - S10secondarynews2025-05-24Taipei Times - Hon Precision's products and buyers Hon Precisionthat Hon Precision's handlers sit in the advanced-packaging AI chip flow - the sockets Cohu is now trying to win
Hon Precision primarily supplies test handlers used in advanced packaging of chip-on-wafer-on-substrate (CoWoS) equipment.
backsC19 - S11secondaryvendor data2026-10-06Research pack, COHU, 2026-10-06content/COHU/research-pack.jsonprice, market and enterprise value, returns and their decomposition, the analyst estimate panel and rating split, earnings surprises, the insider-filing roster with Rule 10b5-1 flags read from Form 4, FINRA short interest, and the daily price historybacksC06C07C20C35C37C38C39C47C49
- S12primarycompany filing2026-02-19that the largest US test equipment maker names Cohu among its Semiconductor Test competitorsbacksC51
- S13primarycompany filing2026-10-06the filed annual revenue, operating income, amortisation of purchased intangibles, net income and diluted share series for FY2016 to FY2025, and the cost lines in content/COHU/financials.jsonbacksC03C04C07C20C31C32C33
- S14secondaryvendor data2026-10-06FMP profile and income statement, Hon. Precision, Inc. (7769.TW) Financial Modeling PrepHon Precision's market value and its 2025 revenue and operating income
Hon. Precision, Inc. marketCap 1062486650000 TWD; 2025-12-31 revenue 30270625000, operatingIncome 14996909000
backsC17C20 - S15secondaryvendor data2026-10-22the 22 October 2026 close of $46.26backsC06C07C47
- S16primarycompany filing2026-10-28third-quarter sales, margins, bookings, the computing pipeline and the raised fiscal-2026 computing outlook, test-cell utilisation, cash and the fourth-quarter guidebacksC08C09C10C13C56
- S17primarycompany filing2026-10-28the Delaware Section 204 ratifications of two charter amendments, two equity plans and nine director elections, and the record-date timing that prompted thembacksC57
- S18primarytranscript2026-10-28the pipeline by stage ($230M qualified, $440M in qualification, 13 ranked programs), the 3x Eclipse output goal for 2027, the 2027 guide (sales up 30% to 40%, gross margin 46% to 48%, about half of added sales to operating income), utilisation by market, Q3 orders by end market and the analysts' questions
We now have programs representing $230 million in qualified annual revenue potential and another $440 million in active qualification.
backsC10C11C12C13C58