COHU

Current Thesis

COHU is a small-cap semiconductor back-end test/handling/inspection vendor that has just re-rated from a cyclical handset+auto laggard into an AI-compute test picks-&-shovels play. The narrative leg we'd be buying: the AI xPU test-handler ramp (Eclipse, Active Thermal Control) + HBM inspection (Neon) inflecting in H2 2026, with sell-side only now catching up (multiple $33→$50-60 target hikes in the four weeks after the 2026-04-30 Q1 print). Theme status: ACCELERATING — orders +57% YoY, FY26 guide raised to +20-25%, AI-compute SAM lifted to ~$750M. The catch: the tape is stretched at 52-week highs (~$57, +192% over 12 months) with no fresh catalyst inside 30 days — the next binary is the 2026-07-30 Q2 print. Fresh entries here are HIGH-conviction-but-extended, not a SUPREME fat-pitch.

Bull Case

• Q1 FY26 (reported 2026-04-30): revenue $125.1M vs consensus ~$122.1M; orders +57% YoY; gross margin 46.5% (above guide); recurring revenue 60% of total — beat-and-raise with a high-quality (consumables-led) mix.
• Q2 FY26 guide: $144M ±$7M = +15% QoQ and +34% YoY — sequential acceleration, not a one-quarter pop.
• FY26 revenue-growth guide RAISED to +20-25%; AI-driven compute SAM raised to ~$750M on the Q1 call.
• Eclipse AI xPU test handler: $650M SAM, 2026 revenue $60-80M, $30M follow-on orders already booked; supports 3kW power dissipation with a 10kW roadmap into a GPU market growing ~23% CAGR.
• Neon HBM inspection/metrology: $100M SAM, ~$20M 2026 revenue, +80% YoY — direct leverage to the HBM build-out.
• HPC outlook raised to ~$80-100M for FY26; new DiamondX GaN power-device orders (~$5M) add AI-data-center-power exposure.
• Sell-side cluster upgrade (orthogonal confirmation): Stifel $34→$50, Needham $33→$54, plus TD Cowen / Jefferies / Evercore ISI / B. Riley to $53-60 — narrative still being "discovered" (Motley Fool framed it 2026-05-27 as "under-the-radar"), i.e. not yet peak-retail.

Bear Case

• Stretched and late on the chart: ~$57 sits at/near the 52-week high ($57.50 / range top $58.47) after a +192% YoY run from a $16.95 low — buying the top of the range with the next catalyst 8 weeks out.
• Gross margin already guided DOWN to ~44% for Q2 (from 46.5%) as systems mix rises and HPC scales — the easy-margin part of the beat is behind it.
• Binary H2 ramp: the whole re-rate prices an H2-2026 HPC production ramp that hasn't shipped; a 2026-07-30 order-stall or pushout snaps the multiple.
• Small-cap semicap beta: ~$2.5B mcap trades violently with the semicap cluster; a sector de-rate (TER/ONTO/FORM/KLAC rolling over) drags it regardless of company news.
• Customer concentration / lumpy orders: handler orders are chunky; one large AI customer slipping a PO is a double-digit revenue swing.
• Valuation now premium vs its own cyclical history — fine while accelerating, a value trap the moment orders decelerate.

Setup & Price Structure

• Price: ~$54.50 (2026-06-01) → ~$57.50 (2026-06-02/03), pressing the 52-week high; range $16.95–$58.47.
• Trend: cleanly ACCELERATING — post-Q1 gap (late April, ~$33→$50s) held and extended; price above all major MAs.
• Estimated structure (small-cap, verify live): post-earnings breakout base ~$48-52 (gap/consolidation = first re-entry zone on any pullback); 20-week EMA roughly ~[entry redacted] = the line that defines the momentum leg.
• Beginner-trap read: stretched above MA = YES; earnings <3 trading days = NO (2026-07-30); peak-retail mania = NOT yet (still sell-side-led, "under-the-radar"); averaging-down = N/A (not held).
• Action frame: ACCELERATING + cluster-confirmed → default is to participate, but with no catalyst inside 30 days and price at range-top, a fresh buy is a HIGH-conviction starter, not max size. Cleanest add is a higher-low / breakout-retest into ~$48-52, NOT a chase at the literal high. Do not average down below the 20-week EMA.

Catalyst Calendar (next 30 days)

• 2026-06-02 — Stifel investor conference (PASSED) — fireside, no print.
• 2026-06-03 — Evercore Global TMT conference (PASSED) — fireside, no print.
• ~Mid/late June 2026 — possible additional sell-side semicap conferences (est., unconfirmed) — watch for incremental Eclipse/Neon order commentary; these have been moving the stock.
• No hard binary inside the 30-day window → catalyst_date = null.
• 2026-07-30 (after close) — Q2 FY26 earnings (THE binary, ~8 weeks out): guide is $144M ±$7M; the print must confirm the H2 HPC ramp and fresh xPU handler wins. This is where the trade is made or broken — flag the 3-day blackout when it approaches.

What Would Change Our Mind

• Bullish escalation → upgrade toward SUPREME: a pre-Q2 announcement of new Eclipse/xPU handler design wins or another $20M+ follow-on order; semicap cluster breaking to new highs in sympathy; a clean pullback-and-reclaim of $48-52 that resets the setup with a tight stop.
• Invalidation → exit/skip: weekly close below the 20-week EMA (~[entry redacted]); OR the 2026-07-30 Q2 print missing the $144M ±$7M guide, guiding GM <43%, or cutting FY26 growth below +20%; OR HPC/Eclipse orders visibly stalling (no new handler wins on the Q2 call). Any of these breaks the acceleration thesis — cut, do not "wait for it to come back."
• Saturation watch: when COHU flips from "under-the-radar" (current) to CNBC/retail-mania framing with the stock 50%+ above the 50-day, the easy money is gone — trim into strength, don't add.

Correlation Notes

• Theme cluster: trades with semicap-equipment / back-end test & inspection — Teradyne (TER), Onto Innovation (ONTO), FormFactor (FORM), Advantest, KLAC. Use the cluster as confirmation: COHU breaking out alone = suspect; the whole group ripping = real (orthogonal cluster confirmation present now).
• Upstream drivers: levered to AI GPU/accelerator unit growth (NVDA/AMD xPU volumes) and the HBM build-out (HBM inspection demand) — same demand curve as the AI-chip-infra-memory theme.
• Beta caution: as a ~$2.5B small cap it amplifies semicap moves both ways; a broad semis de-rate or rate-shock will hit it harder than the large-cap names. Don't treat its move as idiosyncratic alpha when the whole group is moving.
• Macro: sensitive to the semiconductor capex cycle and any AI-capex-digestion narrative; a "hyperscaler capex peaking" headline is a sector-wide risk-off trigger for this name.