COHU · Cohu, Inc. · Stock research

Current Thesis

The picks-and-shovels leg — AI/HPC test handling (Eclipse with Active Thermal Control) plus HBM inspection (Neon) — printed its strongest quarter of the cycle on 2026-07-30, and the stock is 28.5% below its 52-week high anyway. Q2 FY26 revenue was $149.0M against a $144.08M consensus, non-GAAP EPS $0.26 against a $0.14 estimate, and Q3 was guided to $163–177M against a $146.33M consensus. B. Riley ($69) and Needham ($65) raised targets the next morning. The tape absorbed none of it: the 2026-08-07 close was $52.88, beneath the ~$55 shelf that marked the June re-rate breakout, and the three-month return is +6.7% — most of the June–July advance has been handed back. The binding variable stopped being Cohu's order book around 2026-07-28, when the chip complex began de-rating on AI-capex-peaking and China-competition concerns (CNBC, 2026-07-29: more than $1T of chip market value shed in the selloff). Company results are compounding while the multiple contracts.

Bullish and bearish views on Cohu, Inc.

The model's bull view on Cohu, Inc. (COHU), in brief: Q2 FY26 (2026-07-30): revenue $149M, +38% YoY, versus $144.08M consensus; non-GAAP EPS $0.26 versus $0.14 estimate; non-GAAP gross margin 45.5%, above the ~44% guided on the Q1 call. The bear view: The structure broke while the fundamentals improved. Both cases follow in full.

Bull Case

• Q2 FY26 (2026-07-30): revenue $149M, +38% YoY, versus $144.08M consensus; non-GAAP EPS $0.26 versus $0.14 estimate; non-GAAP gross margin 45.5%, above the ~44% guided on the Q1 call. Recurring revenue was 53% of the total; adjusted EBITDA margin 12%.
• Q3 FY26 guide ~$170M ±$7M (+14% sequential, +35% YoY) against a $146.33M consensus, with gross margin ~45% and opex ~$54M — the guide, not the print, is the larger surprise.
• HPC pipeline raised to ~$850M annually on the Q2 call, from ~$750M given on 2026-04-30: $190M qualified across 4 customers, $250M in active qualification across 5, $445M early-stage across 10.
• FY26 HPC revenue outlook lifted to $100–110M from the prior $80–100M range.
• Order mix rotated hard into compute: computing was 46% of system orders and +150% YoY in Q2; industrial orders +87% YoY; software analytics orders +140% YoY, with the analytics line hitting its first $1M revenue quarter.
• A $26M Eclipse systems order booked in early July 2026, plus semiconductor test utilization at 80% at quarter end — management framed that utilization level as a turning point for customer capex.
• Capacity is being built ahead of the ramp: Malaysia manufacturing capacity targeted to double by year-end 2026.

Bear Case

• The structure broke while the fundamentals improved. The 2026-08-07 close of $52.88 is 28.5% under the $73.91 52-week high and below the June breakout shelf near $55 that prior coverage flagged on 2026-07-12 as the level defining the re-rate.
• A beat-and-raise of this size failing to hold a bid is the market repricing the theme rather than the company. The late-July chip rout (Bloomberg 2026-07-28 on AI fatigue and China competition; CNBC 2026-07-29 on >$1T of chip value shed) overwhelmed a guide well above consensus at the midpoint.
• Every published target now sits above spot and none has been cut: TD Cowen $80 (2026-06-16), Baird $65 (2026-07-09), Stifel $70 (2026-07-10), Needham $65 (2026-07-31), B. Riley $69 (2026-07-31). A stack struck at $65–80 against a $52.88 close is a source of downgrade risk rather than support.
• Legacy end-markets are still shrinking: automotive orders −24% YoY in Q2 FY26, offsetting part of the compute ramp.
• Input-cost pressure flagged on the Q2 call — longer lead times and higher prices for memory and specialty components, with no shortages yet reported — against a Q3 gross-margin guide of ~45%.
• GAAP profitability is not established. Q1 FY26 (2026-04-30) printed a net loss of $12.1M, −$0.26/share; the $0.26 Q2 figure is non-GAAP.

Setup & Price Structure

• Last completed daily close $52.88 (2026-08-07); 52-week range $17.71–$73.91; −28.5% from the high; three-month return +6.7%; RSI(14) 51.8.
• Prior coverage on 2026-07-12 recorded a then-record $70.47 set 2026-06-18. The $73.91 in the adjusted daily bars therefore printed after 2026-07-12 — an inference, but it places the peak inside the same three-week window that contains the sector rout and the 07-30 report.
• RSI at 51.8 after a 28.5% drawdown indicates the decline was quick and the recovery off the early-August low was quick too. Alphastreet reported shares at $47.26 during the sector-wide selling in that stretch. This is a mid-range chart sitting under broken support, not a washed-out one.
• The $55 area has flipped from the shelf the June breakout built into overhead supply. The first structural observable that would matter is a weekly close back above it.
• Crowding/positioning observables: five targets between $65 and $80 struck between 2026-06-16 and 2026-07-31, all above the 08-07 close, none cut yet; no reporting event inside 30 days; no insider transactions or issuance in the recent filing record reviewed. The $260M of 1.50% convertible senior notes due 2031 priced in September 2025 remain in the capital structure.
• Life-cycle: SATURATED. Dating it — the sell-side re-rate completed on 2026-07-31 with the fifth target raise in seven weeks, and within the same week the strongest guidance of the company's AI cycle could not hold a bid. Mainstream coverage, best-case news flow, thin marginal buyer.

Catalyst Calendar (next 30 days)

• 2026-08-19 — Needham Virtual Semiconductor Conference (company IR event calendar). First scheduled management appearance since the 07-30 guide; Eclipse follow-on cadence and whether the $26M July order has been repeated is the datapoint that carries.
• 2026-08-24 — Jefferies Semiconductor Conference, Chicago (company IR event calendar).
• No scheduled reporting event falls inside the window. Q3 FY26 results ~2026-10-29 (est.), based on the 2026-04-30 and 2026-07-30 cadence.
• Beyond the window and dated: CEO Summit 2026-10-13 (San Francisco) and investor day 2026-11-10 (New York) — the latter is the next scheduled multi-year model update.

What Would Change Our Mind

The structural break has already occurred: the June re-rate shelf near $55 gave way and the 2026-08-07 close of $52.88 sits under it, which is why the fundamental acceleration and the price now point in opposite directions. What settles the argument in each direction is observable.

• Downside, gradeable: a weekly close below $47 completes the round-trip of the June–July re-rate and takes out the level Alphastreet reported during the early-August sector selling.
• Fundamental break: a Q3 FY26 revenue print beneath the $163M low end of the 2026-07-30 guide, or FY26 HPC revenue walked back below the $100M floor set on that call.
• Sell-side break: the first price-target cut out of the $65–80 cluster, which would remove the last external support for the re-rate case.
• Upside flip: a weekly close back above $55 while the back-end test peers make higher lows would re-establish the trend and move the label off SATURATED.
• Silence is also information — an August conference appearance that adds no new Eclipse order or qualification conversion would mean the $850M pipeline stopped producing bookings headlines within a month of being raised.

Correlation Notes

• Trades with the back-end test and inspection cluster (TER, ONTO, FORM, ACLS) and with front-end semicap (KLAC, AMAT, LRCX) on sector flow days; the 2026-07-28/29 rout that hit SK Hynix, Samsung and Micron is what moved this name, absent any company-specific disclosure.
• Neon HBM inspection ties revenue directly to memory capex, so DRAM/HBM headlines are a live input; the same memory tightness shows up on the cost line, with management flagging longer lead times for memory components on the Q2 call.
• The residual automotive and industrial test business (auto orders −24% YoY in Q2) keeps a second, uncorrelated cycle inside the P&L that tracks the analog/MCU complex rather than AI compute.
• Small-cap semicap beta: with a 52-week range of $17.71–$73.91, index-level moves in the semiconductor complex are amplified here, and the June-to-August round trip happened without any change to company guidance.