COHU
Cohu
Summary
What they do:
Design and manufacture semiconductor test and inspection equipment — including the Eclipse test handler platform for large AI devices like GPUs, HBM inspection and metrology systems, and PMIC (power management IC) test equipment for data center power delivery.
Why they matter:
As AI chips get larger, hotter, and more expensive, the test equipment must keep pace. Cohu's Eclipse handler is designed specifically for high-power thermal testing of AI devices, and their HBM inspection systems address the quality control bottleneck in memory stacking. HBM revenue is projected to grow from $11M to $15–20M in 2026, and the Eclipse AI configuration is ramping.
Recent performance:
FY2025 revenue $453M, up 13% YoY. Q4 revenue $122M, up 30% YoY. Q1 2026 guided $122M (±$7M). HBM revenue $11M in 2025, guided $15–20M in 2026. Stock ~$28, market cap ~$1.3B.
Our Verdict
A semiconductor test equipment company with two emerging AI proof points — Eclipse handlers for GPU testing and HBM inspection systems for memory quality control — trading at ~3x trailing revenue with 30% Q4 growth and a recovery cycle just getting started.
Structural trends
Structural
70
/ 100
Moat
5/10
Eclipse high-power thermal handler + HBM inspection, but competitive vs TER/Advantest
AI Exp.AI Exposure
Embedded~20% AI
Play Type
EmergingAI Growth
~25%
Rel. Value
64
ATTRACTIVEEvery semiconductor chip must be tested before it ships. For AI chips costing thousands of dollars each, the test step is critical — one defective GPU in a $50M server cluster can bring down an entire training run. Cohu makes the equipment that performs these tests.
The company operates across three product areas. Test handlers (~40% of revenue) physically position chips for electrical testing — inserting them into test sockets, controlling temperature, and cycling through test patterns. The Eclipse platform is the flagship, designed for large, high-power AI devices that require precise thermal control during testing. Inspection and metrology (~25%) uses cameras, sensors, and algorithms to check semiconductor packages for visual defects, dimensional accuracy, and assembly quality — including HBM stack inspection. Interface products (~35% recurring) provides contactors, sockets, and probe cards that wear out with use and generate recurring replacement revenue.
Cohu competes with Teradyne (TER) in handlers, Advantest in test systems, and KLA (KLAC) in inspection/metrology. Cohu is smaller than all three but differentiated in specific niches: Eclipse's high-power thermal capability for AI devices and HBM inspection where few alternatives exist at production scale.
FY2025 revenue was $453M with the company recovering from the 2023–2024 semiconductor downturn. The 60/40 split between recurring revenue and systems provides stability through cycles — recurring revenue increased sequentially for four consecutive quarters through Q4, a reliable market recovery signal. Test utilization rates rose to 76% in December (up 1pt in a typically slow seasonal quarter), with compute the strongest segment at 78%. Management is streamlining product lines — discontinuing certain offerings to focus engineering and support resources on HPC, HBM, and AI-related opportunities. The company raised $287.5M via convertible notes (1.5% rate, 5-year term) in Q4, leaving it with $484M cash and $305M total debt, well-capitalized for the recovery ramp. Full-year 2025 orders were up 29% YoY, and management expects another growth year in 2026 driven by Eclipse handler ramp, HBM inspection, and recovering auto/industrial demand.
Supply Chain Dependencies
The Catch
Cohu is a $1.3B market cap test equipment company competing against Teradyne ($15B) and Advantest ($30B). The AI-specific products (Eclipse AI handler + HBM inspection) represent a small fraction of total revenue today — perhaps $25–30M combined on a $453M base. If these niches don't scale meaningfully, Cohu remains a generic semiconductor cycle play with limited differentiation and compressed margins.
If They Win
If AI chip testing becomes a recognized bottleneck — with GPU test times increasing, HBM inspection becoming mandatory for every stack, and test equipment content per chip growing — Cohu's early positioning in Eclipse handlers and HBM metrology pays off as the AI test equipment specialist. Revenue reaches $600M+ by 2028, AI-related revenue grows to 20%+ of the total, and the company earns a premium multiple as the only mid-cap pure play on AI semiconductor testing.
Others in Package the Chip
Not financial advice. All scores generated via AI algorithms using public data.