Executive Summary / Key Takeaways
- Cohu has transformed from a cyclical equipment supplier into a more resilient, AI-exposed business model where recurring revenues now represent 60% of sales, providing a stable foundation that historically never existed in semiconductor test equipment.
- The company's strategic pivot toward high-growth AI applications—High-Bandwidth Memory inspection and high-power processor test—positions it to capture value from the AI infrastructure buildout, with AI-related system revenue reaching $40 million in 2025 and HBM revenue forecast to grow significantly in 2026.
- Management's aggressive restructuring and manufacturing consolidation to Asia, combined with the Tignis acquisition, signal a focus on margin expansion and operational leverage that should become visible in 2026 as one-time charges roll off and utilization rates continue climbing from current 76% levels.
- Trading at $29.98 with an enterprise value of $1.25 billion (2.8x sales), Cohu trades at a discount to AI-exposed peers like Teradyne (TER) (14.5x sales) while offering comparable gross margins and balance sheet strength, suggesting the market hasn't recognized the strategic transformation.
- The investment thesis hinges on two variables: whether Cohu can sustain its design win momentum in AI processors and HBM against larger competitors, and whether the 60% recurring revenue mix provides enough cyclical insulation as the semiconductor industry navigates its next downturn.
Setting the Scene: The Quiet Transformation of a 78-Year-Old Test Equipment Supplier
Cohu, Inc., founded in 1947 and headquartered in Poway, California, operates in a critical corner of the semiconductor ecosystem: the equipment that ensures chips actually work before they ship. The company builds the test handlers , inspection systems, and interface products that semiconductor manufacturers use to verify that every device—from the AI processor in a data center to the radar chip in a car—meets performance and quality standards. This is a business where precision matters: a single defect in a high-performance AI processor can render a $10,000 server board useless.
The semiconductor test equipment industry is cyclical, seasonal, and concentrated. Three players—Advantest (ATEYY), Teradyne, and Cohu—control roughly 55% of the market, with Cohu holding an estimated 10-15% share primarily in test handlers and interface solutions. The industry structure creates a challenging dynamic: equipment sales surge during capacity expansions and decline during downturns, while customers wield pricing power. This cyclicality has historically made pure-play equipment suppliers volatile investments, as revenue can fall significantly in a single year when capital spending freezes.
Cohu's current positioning has evolved to mitigate these structural challenges. Through strategic acquisitions—Xcerra in 2018, MCT and EQT in 2023, and Tignis in 2025—Cohu has expanded beyond capital equipment into a recurring revenue model that now represents 60% of sales. This is a fundamental change in the company's risk profile. While competitors like Teradyne and Advantest remain weighted toward cyclical tester sales, Cohu has built a base of interface products, services, spares, and software that customers purchase to keep existing equipment running.
The industry is now experiencing a structural tailwind: the AI semiconductor boom. High-Bandwidth Memory (HBM) devices, which stack memory chips vertically to feed data-hungry AI processors, require 100% inspection because any defect ruins the entire stack. High-performance AI processors from companies like NVIDIA (NVDA) and AMD (AMD) now dissipate up to 3,000 watts of power, requiring sophisticated thermal management during test that Cohu's Eclipse handler was specifically designed to address. These represent a step-function increase in test complexity that favors specialized equipment providers.
History with a Purpose: How Acquisitions Created an AI-Ready Platform
Cohu's transformation was driven by deliberate expansion. The 2018 acquisition of Xcerra for $350 million brought critical test handler and contactor capabilities. The integration involved closing facilities in Malaysia and California and consolidating operations, creating the scale necessary to compete globally. By December 2022, these integration efforts were largely complete.
The 2023 acquisitions of MCT and EQT were surgical. MCT brought strip test handlers and film frame handlers that expanded Cohu's addressable market in advanced packaging , while EQT added test contactors and consumables that boosted recurring revenue. The EQT deal brought the Ultra-S contactor, which contributes to Cohu's revenue and innovation reputation. These deals filled product portfolio gaps that limited Cohu's ability to capture more of each customer's test cell spending.
The January 2025 acquisition of Tignis, an AI process control and analytics software provider, represents the final piece of the transformation. It gave Cohu an AI-driven digital twin platform that complements its hardware. This positions Cohu to capture software-like recurring revenue from customers who want to optimize yield and equipment utilization. In Q1 2025, Cohu signed three new demonstration opportunities with front-end equipment companies, material suppliers, and a U.S. Defense Military Research Group.
The operational restructuring announced in February 2025—consolidating remaining volume manufacturing from Poway, California to Asian facilities—will reduce quarterly operating expenses to approximately $49 million when revenue reaches $130 million per quarter. This reduction in OpEx intensity will flow to operating leverage. The decision to end manufacturing of certain legacy products created a $3.3 million one-time inventory charge in Q4 2025, a choice to prune low-margin business and focus resources on AI-exposed products.
Technology, Products, and Strategic Differentiation: The AI Test Advantage
Cohu's competitive moat in the AI era rests on thermal management expertise, inspection capabilities for advanced packaging, and software analytics.
The Eclipse handler with Active Thermal Control (ATC) represents the company's most important product for AI applications. The platform can handle power dissipation up to 3,000 watts with fast temperature ramp rates and tight thermal guard bands. Next-generation AI processors generate more heat than previous semiconductor devices. Without precise thermal control during test, these chips throttle performance or produce false failures. Cohu's solution ensures optimal device temperature control and test repeatability, leading a U.S.-based semiconductor manufacturer and foundry services company to place a multi-unit order in March 2026 for testing next-generation AI datacenter processors.
The Eclipse platform's configurability allows it to cover passive testing for RF discrete components, ATC mobile applications, tri-temp automotive testing, and high-performance computing—all within one system with field upgrades. This addresses the requirement for a "future-proof platform" that can span multiple cycles. For capital-intensive test floors, this flexibility reduces obsolescence risk. The $28 million order in Q2 2025 for mobile and automotive end markets demonstrates customer demand for this versatility.
The Neon HBM Inspection System addresses a bottleneck in AI memory production. HBM devices require 100% inspection because defects in any layer of the 3D stack render the entire device useless. As HBM generations advance from HBM3 to HBM4, the complexity and inspection time increase. Cohu shipped its first HBM4-configured system in Q3 2025. The company exited 2025 with $11 million in HBM revenue and forecasts $15-20 million for 2026, representing 36-82% growth. HBM is a fast-growing segment, with AI servers requiring significantly more HBM than traditional servers.
Power Probe Cards for silicon carbide (SiC) testing represent another differentiated product. SiC power devices are essential for electric vehicles and AI data center power supplies, but testing them at high voltage creates a risk of arcing between probe tips. Cohu's patented technology enables multi-site testing of high-voltage die applications. Each customer win represents approximately $2 million per year in revenue, and the company qualified its first orders from a leading European semiconductor company in Q1 2025.
The DI-Core and Tignis software suite transforms Cohu into a solutions provider. The AI-driven digital twin platform provides real-time performance monitoring, prescriptive maintenance, and machine learning process control. While small today, software analytics generate higher gross margins than hardware and create switching costs. The demonstration opportunities signed in Q1 2025 suggest the addressable market extends beyond Cohu's installed base.
Financial Performance & Segment Dynamics: Evidence of Strategic Execution
Cohu's 2025 financial results show strategic repositioning. Consolidated net sales increased 12.7% year-over-year to $453 million, driven by demand for AI-based computing applications that offset declines in mobile and weakness in automotive and industrial segments. This demonstrates the company's ability to pivot toward growth markets while managing cyclical headwinds in traditional segments.
The gross margin decline from 44.9% in 2024 to 42.7% in 2025 was influenced by a $3.3 million one-time inventory charge related to discontinuing legacy products. Without this charge, gross margin would have been approximately 43.4%. Management is making choices to exit low-margin business and focus on higher-value AI applications. The Q1 2026 guidance for gross margin to return to approximately 45% suggests this was a temporary impact.
The segment mix shift is a key financial trend. Recurring revenue grew from 49% of sales in 2023 to 60% in 2025. Recurring revenue exhibits less cyclicality than capital equipment sales. In Q4 2025, recurring bookings were up 34% sequentially. The recurring revenue base provides a $272 million floor that supports R&D investment and operational expenses even when equipment sales soften.
Operating leverage is beginning to materialize. SG&A expense decreased from 31.9% of net sales in 2024 to 27.3% in 2025, reflecting cost control and restructuring savings. R&D expense increased to 20.4% of sales, including costs from the Tignis integration. Management is maintaining investment in AI-related product development while rightsizing the cost structure. The restructuring program will reduce quarterly operating expenses to approximately $49 million at a $130 million quarterly revenue run rate.
The balance sheet provides strategic flexibility. Cash increased to $484 million, funded by $287.5 million in convertible notes issued in September 2025 and $31.7 million in operating cash flow. The company holds $151 million of cash overseas, providing natural hedging for Asian operations. Total indebtedness of $296 million is manageable, with the convertible notes carrying a 1.5% coupon, creating financing that minimizes dilution through capped call transactions.
Outlook, Management Guidance, and Execution Risk
Management's guidance for Q1 2026—revenue of approximately $122 million with recurring revenue representing 60%—signals confidence in the business model's stability. Q1 is typically a seasonally weak quarter, yet Cohu expects systems revenue to offset this headwind. The forecast for gross margin to return to 45% indicates the Q4 inventory charges were one-time, while flat operating expenses indicate restructuring benefits are flowing through.
Utilization data suggests industry recovery. Estimated test cell utilization increased to 76% in December 2025, up from 72% in March. Historically, when utilization crosses 75%, it signals the beginning of a recovery phase. Management noted that mobile utilization is expected to cross 75% in Q1 2026, which would be a leading indicator for subsequent systems demand.
The AI revenue trajectory provides upside optionality. Management estimated AI-related system revenue at approximately $40 million in 2025 and expects growth in 2026. The HBM inspection revenue forecast of $15-20 million for 2026 represents growth in a market where Cohu has secured additional systems for Q1 2026. The Eclipse handler's multi-unit order for AI datacenter processors in March 2026 demonstrates that Cohu is winning functional test positions in next-generation AI chips.
Execution risk centers on capacity and competitive response. Management confirmed they have the capacity to meet demand for the Eclipse ramp. However, competitors like Teradyne and Advantest are also targeting AI test with larger R&D budgets. The key question is whether Cohu's specialized thermal and inspection capabilities can defend pricing power. The Eclipse's configurability addresses the need for "future-proof platforms" better than some rigid competitive offerings.
Competitive Context and Positioning
Cohu competes against companies with superior scale. Teradyne commands 20-25% market share in SoC testing with significant quarterly semiconductor test revenue. Advantest holds 30-35% share. Tokyo Electron (TOELY) and FormFactor (FORM) bring additional pressure with integrated solutions and specialized probe cards.
Cohu competes in specialized applications where its technology creates value. The Eclipse handler's thermal management capability addresses problems that generic handlers cannot solve, giving Cohu pricing power in AI processor test. The Neon HBM inspection system's ability to handle HBM4 devices creates a switching cost—once qualified, changing vendors requires months of requalification. This locks in recurring revenue.
Cohu's 42.7% gross margin trails Teradyne's and Advantest's, reflecting scale and mix. However, Cohu's 60% recurring revenue mix is higher than Teradyne's service revenue percentage, suggesting a resilient business model. The company's valuation multiple represents a discount to Teradyne and Advantest, despite growth in AI-exposed segments. This valuation gap implies the market views Cohu as a cyclical laggard, creating potential upside if the company executes on AI targets.
The competitive moat is strengthened by Cohu's global service network and installed base of over 25,000 systems across 108 customers. Semiconductor test is a mission-critical application where downtime is costly. Cohu's ability to provide support creates customer loyalty, while relationships with OSATs in Taiwan, Korea, and China provide visibility into capacity expansions.
Risks and Asymmetries
A material risk is customer concentration. In 2025, 60% of revenue came from the ten largest customers. If a major customer delays expansion or switches platforms, Cohu's revenue could decline. This amplifies the impact of competitive losses and creates volatility that the recurring revenue base may not fully offset.
Supply chain dependencies pose a threat. In 2025, supply constraints for critical test socket components impacted Cohu's ability to supply finished goods. With manufacturing concentrated in Asia, the company is exposed to geopolitical instability and trade barriers. Management's comment that exposure to China remains limited provides some insulation, but the concentration of suppliers in Asia creates a bottleneck.
The cyclical nature of the semiconductor industry remains a risk. While 60% recurring revenue provides a floor, the 40% systems revenue can swing. Mobile market sales and automotive system sales have shown significant fluctuations in recent years. End markets can deteriorate quickly, and the recurring revenue base may not grow fast enough to offset a severe equipment downturn.
AI regulation presents an emerging risk. The EU AI Act and U.S. executive orders could require modifications to products and processes. Compliance costs could impact the margin expansion story, while regulatory delays could slow the adoption of AI test equipment.
On the positive side, asymmetry works in Cohu's favor if AI semiconductor demand exceeds expectations. The HBM inspection market is growing, with Cohu raising its 2025 forecast during the year. If HBM4 adoption accelerates, Cohu's 2026 forecast could prove conservative. Similarly, if the Eclipse handler wins additional design wins, systems revenue could grow faster than implied by guidance.
Valuation Context
At $29.98 per share, Cohu trades at a market capitalization of $1.41 billion and an enterprise value of $1.25 billion (2.8x trailing revenue). This valuation multiple represents a discount to direct competitors Teradyne, Advantest, and FormFactor. The discount reflects Cohu's smaller scale and recent results, but it also creates upside if the strategic transformation delivers margin expansion.
The balance sheet strength supports the valuation. With $484 million in cash and short-term investments and $641 million in working capital, Cohu has liquidity to fund operations and strategic investments. The debt-to-equity ratio of 0.42 is manageable, with convertible notes carrying a 1.5% coupon. The $31.4 million spent on capped call transactions to reduce dilution shows a focus on protecting shareholder value.
Cash flow metrics are improving. Annual operating cash flow of $31.7 million and free cash flow of $10.7 million are positive, and quarterly free cash flow in Q4 suggests the business is generating cash as utilization improves. The price-to-operating cash flow ratio reflects a company in recovery mode, while the price-to-free cash flow ratio reflects the impact of restructuring costs and inventory charges that should normalize in 2026.
The absence of a dividend and minimal share repurchases indicates management is prioritizing reinvestment in growth. The remaining repurchase authorization provides a cushion if the stock remains depressed, but the primary value driver will be revenue growth and margin expansion.
Conclusion
Cohu has completed a strategic transformation that positions it to capture value from the AI semiconductor infrastructure buildout through a more resilient, higher-margin business model. The shift to 60% recurring revenue provides a stable foundation that changes the company's risk profile, while design wins in HBM inspection and high-power AI processor test create paths to revenue growth in 2026. The market's focus on recent results and cyclical headwinds has created a valuation discount that doesn't reflect the quality of the recurring revenue base or the AI exposure.
The investment thesis will be decided by two variables. First, can Cohu sustain its design win momentum against larger competitors, particularly for the Eclipse handler in AI datacenter applications? The March 2026 multi-unit order suggests progress, but additional wins are needed to drive systems growth. Second, will the 60% recurring revenue base provide sufficient cyclical insulation when the next semiconductor downturn arrives? Recurring revenue growth and improving utilization rates suggest the floor is rising.
Trading at 2.8x sales with a net cash position and drivers for margin expansion, Cohu offers risk/reward for investors focusing on the strategic inflection. If management executes on AI revenue targets and the restructuring delivers the projected operating leverage, operating margins could reach the mid-teens by 2027—a level that would command a higher valuation multiple. The story is about capturing AI-driven growth with a more durable business model.
