Cohu
Sells capital test/handling systems + inspection metrology, monetized further by a large recurring stream of consumables (contactors/kits), spares, services and software analytics (~60% of revenue is recurring); cyclical to semiconductor unit volumes, with an emerging AI/HPC system growth leg.
- 2026-08-04This market capitalisation previously read ~$2.8B (as of Jul 2026). Restated to ~$2.31B on this refresh, roughly 18% lower.
- 2026-08-04This share price previously read ~$59.27 (as of Jul 3 2026). Restated to $48.84 on this refresh, roughly 18% lower.
Sources — 12 figures with citations
- Q2 FY2026 net sales, gross margin, GAAP and non-GAAP resultsfiled2026-06-27 (released 2026-07-30)Net sales $149.002M (+38.4% YoY vs $107.680M); gross margin 45.4% GAAP / 45.5% non-GAAP; GAAP operating income $0.292M; GAAP net loss $(0.159)M, $(0.00)/share; non-GAAP net income $14.051M, $0.26/share; total cash and investments $498.2Msec.gov — Form 8-K Exhibit 99.1, Cohu Q2 FY2026 earnings release, including the Consolidated Statements of Operations and the GAAP-to-non-GAAP reconciliation (which shows the $14.8M of add-backs: stock comp $6.593M, intangible amortisation $7.277M, restructuring $0.633M).
- Balance sheet at 2026-06-27filed2026-06-27Cash and investments $498.167M; short-term borrowings $9.976M; current installments of long-term debt $1.206M; long-term debt $285.049M; stockholders' equity $774.588Msec.gov — Condensed Consolidated Balance Sheets in the same exhibit.
- Net cash positionderived2026-06-27~$201.9M net cashsec.gov — Arithmetic: $498.167M cash and investments - ($9.976M + $1.206M + $285.049M = $296.231M total debt) = $201.936M.
- 1H FY2026 cash flow and capexfiled2026-06-27Net cash provided by operating activities $20.754M; purchases of property, plant and equipment $4.328M; free cash flow $16.426M; depreciation and amortisation $20.839Msec.gov — Form 10-Q Condensed Consolidated Statements of Cash Flows, six months ended 2026-06-27. FCF derived as OCF less PP&E purchases: 20.754 - 4.328 = $16.426M (6.0% of 1H sales).
- Q3 FY2026 guidance and raised FY2026 HPC outlookfiled2026-07-30Q3 net sales $170M +/- $7M; FY2026 HPC revenue raised to $100-110M (from $80-100M); annual AI-driven compute opportunity pipeline raised to ~$850M; test cell utilization ~80% at end of Junesec.gov — Guidance paragraph, headline bullets and CEO commentary in the 8-K exhibit. The HPC increase is attributed to Eclipse handler adoption with T-Core active thermal control.
- Q3 margin/opex guidance, pipeline composition, recurring revenue, ordersfiled2026-07-30Q3 gross margin ~45%, operating expenses ~$54M; pipeline $190M qualified (4 customers) / $250M active qualification (5) / $445M early engagement (10); recurring revenue 53% of Q2 sales; computing segment orders +150% YoY; $26M single-customer Eclipse order shipping largely in Q4; Malaysia capacity to double by year-endfool.com — Q2 FY2026 earnings call transcript (management statements, same day as the 8-K). Management explicitly frames the $850M as ANNUAL SPEND by targeted customers, not a multiyear total or a company revenue forecast.
- FY2025 net sales (TTM base)filed2025-12-27$452.956M for fiscal 2025 (52 weeks ended 2025-12-27)data.sec.gov — SEC XBRL company-concept API, FY2025 value as tagged in Cohu's 10-K.
- TTM revenue (Q3 FY2025 - Q2 FY2026)derived2026-06-27$522.6Msec.gov — Arithmetic: FY2025 $452.956M - 1H FY2025 $204.477M + 1H FY2026 $274.121M = $522.600M.
- FY2026 revenue run-rate implied by guidancederived2026-07-30~$444.1M through three quarters; a flat Q4 at the Q3 guided midpoint would put FY2026 at ~$614M, ~+35% YoYsec.gov — Arithmetic: 1H actual $274.121M + Q3 guided midpoint $170M = $444.1M, versus FY2025 total of $452.956M. This implies the previously carried FY2026 '+20-25%' growth guide is stale; the release does not restate a full-year revenue guide (see notVerified).
- Price and market capitalisationmarket2026-08-03 (close)$48.84 close; market cap $2.31B; ~47.34M shares outstandingstockanalysis.com — Closing price, not intraday. Down from $59.27 on 2026-07-03.
- EV/TTM salesderived2026-08-03~4.0x (EV ~$2.11B on $522.6M TTM sales)stockanalysis.com — Arithmetic: $2.31B market cap - $201.9M net cash = $2.108B EV; $2.108B / $522.6M = 4.03x. Compares with ~5.7x EV/Sales on FY2025 revenue.
- Q2 FY2026 operating expense detailfiled2026-06-27R&D $24.943M; SG&A $34.445M; intangible amortisation $7.277M; restructuring $0.633M; interest income $3.868M vs interest expense $1.620Msec.gov — Consolidated Statements of Operations. Total costs and expenses $148.710M against $149.002M of sales — the margin of GAAP profitability is $0.3M wide.
Earnings, margins, COGS & capex
FY2025 revenue rose ~13% to $453.0M off a cyclical trough but the company remained GAAP loss-making (op margin -15.4%, net loss -$74.3M) on under-absorbed capacity, restructuring and intangible amortization; only ~$10.7M FCF. The 2026 story is a demand inflection: Q1'26 revenue +29% YoY with orders +57% YoY and gross margin recovering to 46.5%, driven by AI/HPC xPU test handlers (Eclipse w/ active thermal control) and HBM inspection (Neon metrology). Recurring consumables/services (~60% of sales) cushion the cyclicality.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~55¢ is cost of goods and ~45¢ operating expense, leaving ~0¢ of operating profit.
Revenue trend
Margins
recovering with volume + mix toward HPC systems
negative at trough; leverage returning as revenue scales
structurally sticky, margin-stabilizing
positive despite op loss; +237% YoY FCF
COGS structure
COGS is dominated by precision mechatronic/thermal hardware, purchased electronic components and semiconductors, and manufacturing labor/overhead in a largely outsourced/contract-manufactured supply chain (~90% of revenue is international). Under-absorption at low volumes is the swing factor on gross margin; the high-margin consumables/contactor kit stream is the structural offset.
Capex
Modest — capex runs roughly 2-3% of revenue; the model's real 'investment' is R&D and applications/qual engineering (opex was elevated to ~$55M in Q1'26 specifically to scale HPC/AI support ahead of revenue).
Latest earnings
Beat and raise on the operating line. Q2 sales of $149.0M landed above the ~$144M +/- $7M guidance midpoint set in April, gross margin held at 45.4%, GAAP operating income crossed into positive territory, and management raised the FY2026 high-performance-computing revenue estimate to $100-110M from $80-100M — the entire increase attributed to Eclipse handler sales. Specific sell-side consensus figures were not sourced, so the beat is stated against company guidance, not the Street.
Q3 FY2026 net sales $170M +/- $7M (~+14% QoQ at the midpoint), gross margin ~45%, operating expenses ~$54M. FY2026 HPC revenue raised to $100-110M (from $80-100M). Annual AI-driven compute opportunity pipeline raised to ~$850M. Malaysia manufacturing capacity expected to double by year-end.
- Q2 FY2026 net sales
- $149.0M (+38.4% YoY, +19.1% QoQ)
- GAAP vs non-GAAP operating income
- $0.3M (0.2% of sales) vs $15.1M (10.1%) — the $14.8M gap is stock comp $6.6M + intangible amortisation $7.3M + restructuring $0.6M
- Test cell utilization
- ~80% at end of June (sequential increase); management calls it 'a turning point for capex by core customers'
- Computing segment orders
- +150% YoY
- AI compute pipeline (annual addressable)
- ~$850M: $190M qualified (4 customers) + $250M in active qualification (5) + $445M early-stage engagement (10)
- Recurring revenue
- ~53% of Q2 sales — DOWN from the ~60% previously carried, because systems revenue is growing faster than consumables
- Largest disclosed order
- $26M single-customer Eclipse order, shipping largely in Q4 FY2026
- Cash and investments / net cash
- $498.2M / ~$201.9M net cash at 2026-06-27
Growth drivers
- AI/HPC xPU test handlers — Eclipse handler with Active Thermal Control; ~$650M SAM, FY2026 revenue outlook raised to ~$80-100M (computing segment), repeat multi-unit orders from US + Korean customers
- HBM inspection metrology — Neon platform (HBM3/HBM4, investing for HBM5); ~$100M SAM, ~$20M FY2026 revenue (~80% YoY growth, market ~68% CAGR)
- Total computing-segment opportunity pipeline of ~$750M ($650M handlers + $100M HBM inspection) across 12 customer engagements at various qual/engagement stages (customers undisclosed)
- Recurring consumables/contactors + software analytics — ~60% recurring base that grows with installed-base utilization
- Broad cyclical recovery in mobility, automotive, industrial and consumer semiconductor test volumes off the 2024 trough
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-17. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
The inflection stopped being a forecast and became a P&L event: revenue +38% YoY, GAAP operating income positive, non-GAAP EPS $0.26 against $0.02 a year ago, Q3 guided to $170M — and the stock is ~18% CHEAPER than when the story was still a pipeline.
- GAAP operating income crossed zero (+$0.3M) after -16.0% operating margin in Q2'25 — operating leverage is now demonstrated, not projected
- Non-GAAP net income $14.1M / $0.26 per share versus $0.7M / $0.02 a year ago, a ~19x step on 38% revenue growth
- Q3 guidance of $170M +/- $7M is ~+14% QoQ on top of +19% QoQ — sequential acceleration, and it puts 9-month FY2026 sales (~$444M) within 2% of ALL of FY2025
- Test cell utilization hit ~80%, which management frames as the trigger point for core-customer capex — the cyclical leg of the story is arriving on top of the AI leg
- HPC revenue estimate raised to $100-110M from $80-100M with the increase attributed entirely to Eclipse handler sales, plus a disclosed $26M single-customer Eclipse order for Q4
- The pipeline both grew and hardened: ~$850M annual addressable with $190M now QUALIFIED across 4 customers (previously the whole $750M was largely pre-qual), and computing orders +150% YoY
- Balance sheet still funds it without dilution: $498.2M cash and investments, ~$202M net cash, capex only ~1.6% of sales, and 1H FCF positive at $16.4M despite a GAAP loss
- Valuation reset in the buyer's favour: ~$2.11B EV on $522.6M TTM sales is ~4.0x, down from a punitive ~5.7x
Strip out $14.8M of stock comp and amortisation and the quarter earned $0.3M — this is a sub-scale niche vendor still loss-making over 1H, whose recurring-revenue cushion is thinning, whose margin ticked DOWN sequentially, and which is spending opex ahead of an order book it does not yet own.
- GAAP is still the truth and GAAP is still red: 1H FY2026 net loss $(12.2)M / $(0.26) per share and a $10.9M operating loss; the $0.26 of 'earnings' is entirely a non-GAAP construction (stock comp $6.6M + amortisation $7.3M added back in Q2 alone)
- Gross margin fell sequentially, 46.5% to 45.4%, and Q3 is guided to ~45% — the systems-heavy AI mix is DILUTIVE to margin, the opposite of the mix-up thesis
- The recurring-revenue cushion shrank from ~60% to ~53% of sales, so the downside protection that justified the multiple is structurally weaker as systems scale
- Opex is being pushed to ~$54M in Q3 ahead of the revenue; a single quarter's demand pause flips the freshly-positive operating line straight back negative
- Two-thirds of the ~$850M pipeline ($695M of $850M) is still in active qualification or early-stage engagement, not backlog — and it is an ANNUAL ADDRESSABLE spend estimate by targeted customers, not a company forecast
- Concentration risk is embedded in the good news: a single $26M Eclipse order is ~15% of a guided quarter, and HPC at $100-110M is a handful of accounts
- Advantest and Teradyne still own the great majority of test spend and are pushing into AI test; Cohu's differentiation is thermal handling, an adjacency they can attack
- The market de-rated the stock ~18% ($59.27 to $48.84) through exactly the period this beat landed — the tape is discounting the durability, not the quarter
What it is worth
Peer-relative (EV/Sales + through-cycle EV/EBIT vs semi-cap test/inspection peers TER, ATEYY, FORM, ONTO, CAMT, AEHR) cross-checked with a reverse-DCF read on what the ~$2.8B cap implies.
~$25-35
cycle stalls or pipeline conversion disappoints, GAAP losses persist, and the ~3x re-rating unwinds toward book/through-cycle sales multiples.
~$55-65
recovery + partial HPC conversion delivers growth and a return to profitability, roughly justifying the current price; multiple holds on execution.
~$85-95+
HPC/HBM pipeline converts, FY2026 lands at the high end (+25%) and margins normalize toward mid-teens+ operating; the name re-rates as a structural AI-test grower, not a cyclical niche.
At ~$2.8B market cap / ~$2.6B EV on ~$453M FY2025 revenue (~5.7x EV/Sales) with GAAP losses, the stock is not valued on trailing trough earnings — it prices a successful AI/HPC + HBM ramp toward mid-teens+ operating margins on a materially larger revenue base. The reverse-DCF implication: the market is underwriting roughly the FY2026 +20-25% growth continuing for several years plus a normalized double-digit operating margin. Reasonable only if the ~$750M pipeline converts; punitive if it slips.
SWOT
Strengths
- ~60% recurring, consumables-led revenue smooths a violently cyclical end-market
- Differentiated in thermal/handling for high-power AI xPU test (active thermal control) — a hard, hot problem as chip TDPs climb
- Net-cash balance sheet (~$184M net cash) funds the AI/HPC ramp without dilution pressure
- Diversified test/inspection portfolio (handlers, contactors, thermal, software analytics, HBM metrology) vs a single-product vendor
Weaknesses
- Sub-scale vs the ATE duopoly (Advantest + Teradyne ~80% of tester share) — Cohu is a niche/back-end player
- GAAP loss-making through FY2025 (-15.4% op margin, -$74.3M net loss); profitability depends on volume recovery
- High operating leverage cuts both ways — under-absorption crushes margins in down-cycles
- Customer + geographic concentration; ~90% international revenue adds FX and geopolitical exposure
Opportunities
- AI/HPC xPU test + HBM inspection — a ~$750M pipeline vs a ~$453M revenue base — a genuine growth leg if it converts
- HBM3/HBM4->HBM5 inspection roadmap in a market growing ~68% CAGR
- Attach more software/analytics and consumables to a growing installed base (higher-margin mix)
- Content growth per device as chiplets/advanced packaging raise test + inspection intensity
Threats
- Semiconductor cyclicality and capex air-pockets can stall the recovery
- Advantest/Teradyne moving deeper into AI test and adjacent handling could squeeze Cohu's niche
- Pipeline conversion risk — much of the $750M is still in qual/engagement, not backlog; timing could slip
- Geopolitics and export-control shifts on a 90%-international revenue base
- Rich valuation (~$2.8B cap on a loss-making trough year) prices in the HPC ramp — execution misses de-rate hard
Moats, dependencies & bottlenecks
Moats
Contactors/kits/spares/services (~60% of revenue) are specified into customer test cells and reorder with utilization — sticky, but tied to Cohu's installed-base share.
Active-thermal-control handling is genuinely hard and a current differentiator, but well-capitalized ATE leaders can invest to close it.
Once qualified into a device's test flow, displacement is costly mid-program — but the moat is per-socket, not company-wide.
Cohu is the small player vs the Advantest/Teradyne duopoly; no scale-cost advantage.
Dependencies
Back-end test demand tracks chip units + capex; a down-cycle under-absorbs the cost base fast.
The bull case rests on ~$750M of qual/engagement pipeline actually converting to backlog and revenue on schedule.
Purchased semiconductors, precision mechatronics and outsourced manufacturing; shortages hit lead times and margins.
Concentrated customer base (IDMs/OSATs + AI datacenter processor makers) Customer concentration ~90% international; a few large test-equipment buyers drive order lumpiness.
Customer/geography HBM inspection growth leans on Korean HBM makers' HBM3/4/5 ramps.
Advantages
- High-margin recurring consumables/contactor + software stream (~60% of revenue)
- Net-cash balance sheet (~$184M net cash) funds the AI ramp without dilution
- Differentiated active-thermal-control handling for AI xPU test
- Diversified test + inspection + analytics portfolio rather than a single product
- Order momentum already visible (+57% YoY) validating the demand thesis
Weaknesses
- Sub-scale vs the ATE duopoly; niche positioning
- GAAP-unprofitable through FY2025; recovery-dependent earnings
- High cyclicality + operating leverage amplifies down-cycles
- Rich valuation prices the HPC ramp — little error margin
- ~90% international revenue -> FX + geopolitical exposure
Bottlenecks
- Pipeline-to-backlog conversion timing (qual cycles gate revenue recognition)
- Operating leverage — fixed cost/opex absorption at low volumes
- Scale disadvantage vs Advantest/Teradyne in R&D and account reach
- Cyclicality of end-market capex limits visibility
Top signals & trends
Top signals
Orders +57% YoY (compute orders ~+211%) in Q1'26 — the leading indicator of the recovery; watch for sustained multi-unit AI-datacenter orders.
Track how much of the ~$450-500M engagement bucket converts and how fast — the crux of the thesis.
46.5% in Q1'26 vs 42.7% FY2025; continued climb signals volume + mix flow-through to profitability.
A quarterly scorecard against the raised outlook; a miss would puncture the re-rate.
Advantest/Teradyne/OSAT capex tone is a read-through on Cohu's back-end demand.
Trends
Higher-power, higher-value AI chips need more sophisticated thermal handling and test — expands Cohu's SAM.
HBM3->HBM4->HBM5 and stacked-die inspection is a fast-growing (~68% CAGR) metrology market Cohu is entering with Neon.
Structural growth but violent cycles; recovery timing dominates near-term results.
More die, more interconnects, more test insertions per package raise back-end intensity.
New back-end capacity is a tailwind, but export controls + concentration add friction on a 90%-international base.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Electronic component & semiconductor suppliers Cohu buys chips, precision mechatronics and thermal/motion components; specific vendors are not disclosed in filings.
Much of Cohu's system build is outsourced; supply lead times feed directly into gross margin. Specific CMs not disclosed.
Representative IDM back-end test customer (automotive/industrial/mobility).
IDM test-equipment buyer across automotive/industrial.
Large analog/embedded IDM with substantial back-end test capacity.
Analog IDM; representative test-handling/consumables customer.
OSAT that runs test/handling equipment — channel for Cohu systems + consumables.
World's largest OSAT; representative back-end test/assembly customer.
AI-datacenter processor makers + Korean HBM memory makers The HPC/HBM growth leg's customers (xPU designers and HBM3/4/5 makers) are not disclosed by name; repeat multi-unit orders confirmed from US + Korean customers.
Automated test equipment leader; co-owns ~80% of tester share with Advantest and pairs with/competes against Cohu handling; pushing into AI/SoC test.
The other ATE duopolist (also TSE:6857), a scale leader in AI/HBM test; scale + R&D dwarf Cohu.
Probe cards + test/measurement; adjacent back-end test-cell content, overlapping in advanced-packaging test.
Wafer-level burn-in/test; niche competitor/adjacent in reliability test, also chasing AI/HBM demand.
Inspection & metrology (incl. advanced packaging/HBM) — competes with Cohu's Neon inspection push.
Inspection & metrology for advanced packaging/HBM; direct competitor to Cohu's HBM-inspection growth leg.