ON Semiconductor (ON): Cyclical Recovery Play with Selective Upside

Thesis

ON Semiconductor(ON) is a cyclical semiconductor name trying to turn a weak earnings base into a stronger mix story. The near-term numbers are messy: 2025 revenue fell 11.2% YoY to $6.0B, earnings growth fell 48.7%, GAAP net margin collapsed to 2.0%, and trailing P/E looks absurd at 249.8x because the denominator almost disappeared. But that headline ugliness hides a more useful fact pattern for a moderate-risk investor with a medium-term horizon. ON still produced $1.76B of operating cash flow, roughly $1.42B of annual free cash flow, a 23.7% free cash flow margin, and a 7.36% free cash flow yield while keeping cash near $2.5B and net debt under $1B.

The core case rests on three points. First, ON is not a broken business. It is a cyclical power and sensing supplier coming out of an automotive and industrial digestion period with signs of stabilization. Second, management is actively improving the quality of revenue by exiting roughly $300M of non-core business, pushing differentiated products like Treo, silicon carbide, SiC JFET, and vertical GaN, and using FabRite actions to reduce underutilization drag and lower depreciation by $45M to $50M in 2026. Third, the company has credible exposure to faster-growth markets, especially AI data center power, where it generated more than $250M in 2025 revenue and expects continued growth in 2026.

The catch is valuation discipline. ON has a real recovery path, but the stock also reflects that hope. Analyst consensus target sits at $69.31, well above the DCF intrinsic value estimate of $28.19 provided here. That gap is the market's argument in one line: investors are paying for normalized earnings power and future mix improvement, not current GAAP earnings. For a balanced investor, ON looks more like a selective Buy on pullbacks than a stock to chase at any price. The opportunity is real, but so is the execution risk.

Company Overview

ON Semiconductor(ON), headquartered in Scottsdale, Arizona, designs and sells intelligent power and sensing semiconductors across automotive, industrial, AI data center, energy infrastructure, factory automation, medical, aerospace, and defense markets. The company operates globally, employs 22,600 people, and trades on the NASDAQ. Its strategic identity is narrower and more focused than a broad analog giant like Texas Instruments(TXN). ON is built around power conversion, power management, connectivity, and sensing, with a particular emphasis on applications where efficiency, thermal performance, and reliability matter more than marketing slogans.

Management under CEO Hassane El-Khoury has spent the last several years reshaping ON from a more commodity-like manufacturing-heavy supplier into a product-centric semiconductor company. That shift matters. In semis, the difference between a commodity part and a qualified, system-critical design win is the difference between price taker and price setter. ON is trying to move up that ladder through automotive power, advanced sensing, silicon carbide, GaN, and AI power delivery.

The company reports through Power Solutions Group, Analog and Mixed-Signal Group, and Intelligent Sensing Group in its business description, though some segment disclosures have shifted in recent data sets. That is not unusual during portfolio realignment, but it does mean investors should focus less on label changes and more on economic substance. ON's business is still driven by power semiconductors and sensing content sold into long-cycle, qualification-heavy markets.

At a market cap of about $28.5B, ON sits in an awkward but potentially attractive middle ground. It is large enough to matter to major customers and invest in new platforms, but small enough that a successful mix shift into AI power, wide-bandgap semis, and zonal automotive architecture can still move the needle. That is often where the best medium-term semiconductor setups live, right between mature utility-like analog names and speculative science projects.

Business Segment Deep Dive

Power remains the center of gravity. In the annual segment data, Power Solutions Group generated $2.81B in 2025 revenue. In quarterly commentary, PSG delivered $724M in Q4 2025, down 2% sequentially and 11% YoY. This segment includes discrete, module, and integrated semiconductor devices for power conversion, switching, signal conditioning, and protection. It is ON's most direct lever to EVs, energy infrastructure, industrial power, and AI data center power delivery.

Analog and Mixed-Signal is the connective tissue between raw power devices and complete system functionality. In Q4 2025, AMG revenue was $556M, down 5% sequentially and 9% YoY. These products include power management, sensor interface, connectivity, and standard products. This segment matters because it broadens ON's content per system. A company that can sell both the muscle and the nervous system of a platform usually gets a better seat at the design table.

Intelligent Sensing Group generated $250M in Q4 2025, up 9% sequentially but down 17% YoY. Annual segment data shows ISG revenue of $928.4M in 2025. The business includes CMOS image sensors, image signal processors, short-wave infrared sensors, photon-counting technologies, and silicon photomultipliers. This is not just a camera business. It is a sensing franchise aimed at ADAS, factory automation, robotics, medical devices, and industrial machine vision.

By end market, automotive and industrial remain the anchors. Q4 2025 automotive revenue was $798M, up about 1% sequentially, while industrial was $442M, up about 4% sequentially. Management said automotive inventory digestion is largely behind the company and industrial showed its first YoY growth after eight quarters of decline. That does not mean a straight-line recovery. It means the floor may be in place, which is often enough to rerate a cyclical stock before the income statement fully recovers.

The smaller but faster-moving opportunity is AI data center power, classified in the other category. ON generated more than $250M of AI data center revenue in 2025. For a $6B revenue company, that is still modest, but it is large enough to prove product-market fit and small enough to grow rapidly. In semiconductors, early revenue in a new architecture wave is like seeing the first current run through a new circuit. It does not guarantee a full ramp, but it proves the line is live.

Flagship Product Analysis

The most important flagship product family today is not a single chip but ON's AI data center power portfolio across the power tree. Management highlighted silicon, SiC, SiC JFET, GaN, and VCORE assets spanning 0.8V to 800V. That breadth matters because AI infrastructure is not just about the accelerator. It is about moving large amounts of power efficiently from the grid to the rack to the board to the processor. Every conversion step wastes heat if done poorly, and heat is the tax collector of modern compute.

ON's SiC JFET and power module offerings appear especially relevant in high-voltage, high-efficiency stages such as power supplies, battery backup disconnects, hot swaps, and UPS systems. Management cited a design win with a leading U.S. power supply supplier that cuts system footprint by about 50% using ON's SiC power module. That is exactly the kind of application-level benefit investors should watch. Semiconductor vendors rarely win on transistor romance alone. They win when the customer can shrink the box, lower heat, reduce failure points, or save rack space.

In automotive and industrial sensing, ON's eight-megapixel image sensor and advanced ultrasonic sensors stand out as flagship offerings tied to ADAS and zonal vehicle architecture. The company also highlighted more than 600M automotive sensors installed base and leadership positions in industrial image sensors and ultrasonic and inductive sensing. Those are not vanity metrics. Installed base and qualification history create inertia, and inertia in automotive is often another word for moat.

The Treo platform deserves mention as a flagship architecture rather than a single product. Management said Treo doubled the number of products sampling in 2025 and now supports a design funnel above $1B. It is already being designed into automotive zonal architecture, ultrasonic sensors, LED drivers, HVAC, energy storage systems, medical devices, and continuous glucose monitors. A platform that can travel across multiple end markets is more valuable than a one-hit wonder, because it spreads R&D across more sockets.

Innovation & Competitive Advantage

ON's competitive edge comes from a combination of application-specific power expertise, automotive qualification depth, manufacturing control, and a growing wide-bandgap roadmap. This is not a software business with zero marginal cost and magical network effects. The moat is more industrial than digital: hard-to-qualify parts, sticky design wins, and system-level know-how in power efficiency and sensing.

The strongest innovation angle is wide-bandgap semiconductors. ON is pushing both silicon carbide and GaN, including vertical GaN. Management said the company is preparing to sample more than 30 new GaN devices spanning 40V to 1,200V, with lateral GaN revenue beginning in 2026 and first VGaN revenue expected in 2027. The investor presentation says ON is first to scale vertical GaN, with 700V and 1,200V devices already sampling and volume production targeted for late 2026.

That claim should be treated as strategic positioning, not gospel. Competitors will not quietly clap from the sidelines. Still, ON's broad voltage coverage from high-voltage conversion down to point-of-load devices is a real advantage in AI power architectures. Management argued that future architectures will collapse the conversion tree, requiring expertise in both high-voltage and low-voltage stages. If that view is right, ON's cross-stack portfolio could matter more than any single device family.

The second innovation pillar is product mix improvement. CFO Thad Trent said ON is reshaping revenue toward differentiated products and away from volatile non-core businesses. That sounds dry, but in plain English it means management is trying to stop renting low-quality revenue just to keep fabs busy. That is usually painful in the short term and healthy in the long term.

The third advantage is customer stickiness in automotive and industrial. ON's parts are sold into systems where reliability, qualification, and redesign costs are high. Once a power device or sensor is designed into an EV platform, ADAS stack, industrial controller, or medical device, replacement is not as simple as swapping a phone accessory. That gives ON a degree of pricing support and demand durability that commodity chip suppliers lack.

Operations & Supply Chain

ON's operations story is central to the investment case because margins are currently being distorted by underutilization. In Q4 2025, non-GAAP gross margin was 38.2%, but management said the quarter included about 700 basis points of underutilization charges. Manufacturing utilization was 68% in Q4 and is expected to rise to the low 70% range in Q1 2026, with mid-70% levels possible later in the year depending on demand.

FabRite is the operating lever here. ON reduced fab capacity by 12% in 2025 and announced additional actions to rationalize the manufacturing footprint. Management expects these actions to lower 2026 depreciation by about $45M to $50M, with gross margin benefits showing up in the second half of the year. This is the mechanics-and-gears part of the story. When a cyclical semiconductor company cuts excess capacity and demand stabilizes, the margin model can improve quickly.

CapEx has also normalized. 2025 capital expenditures were $341.2M, down sharply from $694.0M in 2024 and $1.54B in 2023. That matters because ON's large capacity investment cycle appears to be behind it. Lower CapEx supports stronger free cash flow conversion and gives management more room for buybacks or selective strategic investment.

Inventory remains elevated but manageable. Q4 inventory was 192 days, down from 194 days in Q3. Of that, 76 days was strategic inventory, down from 82 days. Excluding strategic builds, base inventory was 117 days, which management described as healthy. Distribution inventory of 10.8 weeks sits within the target range of 9 to 11 weeks. In other words, the warehouse is still fuller than ideal, but not flashing distress.

Supply chain flexibility is improving through foundry partnerships, especially for lateral GaN. ON highlighted new foundry relationships to broaden regional supply options. That is important in a world where customers increasingly care about resilience, regional sourcing, and geopolitical exposure. The old semiconductor model of optimizing only for cost now shares the stage with a newer model: optimize for cost, resilience, and politics, in that order depending on the week.

Market Analysis

ON is exposed to several attractive semiconductor markets, but they are not all moving at the same speed. Automotive and industrial remain soft relative to peak conditions, while AI infrastructure is the clear growth engine. That split explains both the frustration and the opportunity in the stock. Investors are waiting for legacy markets to stop dragging while new markets become large enough to matter.

Management's 2026 investor materials point to a $64B TAM by 2029 across key growth areas: $26.3B in EVs, $12.2B in AI data center, $11.3B in factory automation, $9.6B in energy infrastructure, $3.7B in advanced safety, and $0.7B in EV charging. The most eye-catching figure is AI data center, where ON sees 40% CAGR. That is plausible given the power density demands of AI racks and the increasing semiconductor content in power delivery.

The AI rack opportunity is especially striking. ON says content per rack could rise from about $9.5K today to about $105K by 2030 as power density increases. Even if that estimate proves optimistic, the direction is clear. AI compute growth does not just benefit GPU vendors. It also benefits the companies that keep those systems fed, cooled, and electrically stable. That is where ON wants to live.

Automotive remains a large but slower market near term. Industry demand has been affected by inventory digestion, uneven EV adoption, and macro caution. ON's management believes the digestion is largely behind it, and the company is positioning for content growth through zonal architecture, ADAS, ultrasonic sensing, Ethernet transceivers, and silicon carbide power devices. The key distinction is between unit growth and content growth. Vehicle units can wobble, but semiconductor content per vehicle can still rise.

Industrial is the other swing factor. Management said Q4 2025 marked the first quarter of YoY industrial growth after eight quarters of decline and pointed to early PMI improvement. If industrial recovery broadens, ON gets both volume support and better fab utilization. If it stalls again, AI can help, but it may not be enough by itself to carry the whole company in the next 12 months.

Customer Profile

ON's customer base is concentrated in long-cycle, engineering-heavy markets. Automotive OEMs and Tier 1 suppliers are core customers, especially for EV powertrains, ADAS, zonal architecture, and in-vehicle networking. Industrial customers span factory automation, machine vision, robotics, HVAC, energy storage, and medical devices. AI data center customers include power supply makers, rack-level system suppliers, and board-level power architecture customers.

Management named Delta, Lite-On, and Great Wall as rack-level design partners for AI power systems, and cited SunGro as a win for next-generation SiC MOSFET hybrid power modules in energy infrastructure. Dexcom was mentioned as a customer using ON's low-power analog front end in continuous glucose monitors. GM was cited as a development collaborator for vertical GaN electric drive systems. These references matter because they show ON is not pitching only to abstract markets. It is inside real customer programs.

The customer profile also shapes risk. Automotive and industrial customers are sticky once qualified, but they are also conservative and slow-moving. That supports durability but can delay ramps. AI infrastructure customers move faster and can scale quickly, but they are demanding and often price-conscious once competition increases. ON's challenge is to balance these worlds: the slow, sticky cash flows of automotive and industrial with the fast, high-growth but more competitive economics of AI power.

Ownership data suggests the shareholder base is heavily institutional, with institutional ownership above 110% and insider ownership at just 0.35%. Short interest at 11.96% of float is not extreme, but it is high enough to show skepticism. Recent institutional activity was mixed, with 7 tracked holders increasing positions and 13 decreasing. That is a fair picture of the stock itself: respected, owned, but not universally trusted.

Competitive Landscape

ON competes most directly with Infineon(IFNNY), STMicroelectronics(STM), Texas Instruments(TXN), NXP Semiconductors(NXPI), Wolfspeed(WOLF), ROHM, and Nexperia. In power semiconductors and automotive power, Infineon and ST are probably the cleanest large-cap comparables. Texas Instruments is a broader analog benchmark with stronger diversification and a richer quality premium. Wolfspeed is the more focused SiC pure-play, with higher technology leverage and higher financial drama. ON sits between these camps.

Relative to TI, ON is less diversified and more cyclical. Relative to Wolfspeed, ON is far more financially stable and commercially diversified. Relative to Infineon and ST, ON is smaller but more tightly focused on intelligent power and sensing. That position can work well if management keeps winning in high-content applications, but it also means ON cannot afford sloppy execution. In a market with strong competitors, being merely decent is often an expensive hobby.

The peer comparison data set failed here, so precise multiple comparisons are limited. Still, the broad market frame is clear. Mature analog leaders like TXN typically command premium multiples because of stability and margin durability. More cyclical auto-industrial names like STM, NXPI, and ON trade at lower normalized earnings multiples when demand is weak. Pure-play wide-bandgap names can trade on technology optionality rather than current profits. ON's valuation sits in the middle because it has both real cash flow and real cyclical baggage.

ON's strongest competitive angle is breadth across the power tree combined with sensing and automotive qualification. Its weakest angle is scale versus the biggest incumbents and exposure to cyclical end markets. The company does not need to beat everyone everywhere. It needs to win enough high-value sockets in AI power, EVs, industrial automation, and sensing to improve mix and margins faster than the market expects.

Macro & Geopolitical Landscape

The macro backdrop is mixed but improving. Gartner expects the semiconductor market to grow in 2025, driven largely by AI-related demand, while automotive and industrial demand remain weaker. That lines up almost perfectly with ON's current setup. The strongest tailwind is in the part of ON's portfolio that is still relatively small, while the largest parts of the company are only beginning to stabilize.

Interest rates and industrial activity matter because ON sells into capital-intensive, economically sensitive markets. A broad manufacturing recovery would help industrial automation, energy infrastructure, and general industrial demand. A softer macro environment would likely delay that rebound. ON's beta of 1.524 reflects this sensitivity. The stock tends to move like a semiconductor name with a caffeine habit, not a sleepy compounder.

Geopolitics also matter. Semiconductor supply chains are being regionalized, and export controls remain a constant risk, especially around China. ON may benefit from customers seeking U.S.-based or regionally resilient suppliers, particularly in industrial sensing and AI infrastructure. Management highlighted strong interest from customers seeking a U.S.-based supplier and positioned ON as a broad-based U.S. power semiconductor supplier aligned with resilient AI infrastructure priorities.

At the same time, regionalization raises costs and complicates manufacturing strategy. ON's foundry partnerships and internal manufacturing footprint can help, but they do not make the company immune. The macro lens here is straightforward: ON is well placed for a world that values power efficiency, electrification, and supply assurance, but it still has to navigate the usual semiconductor hazards of inventory cycles, pricing pressure, and policy shifts.

Balance Sheet Health

Income Statement Strength

Estimates Outlook

Valuation Assessment

Target Prices & Recommendation

Closing

ON Semiconductor(ON) is a credible recovery story with real technology substance. The company is not selling hope alone. It has strong free cash flow, manageable leverage, improving operational discipline, and genuine exposure to attractive markets like AI data center power, EV content, industrial automation, and advanced sensing. Management's comments on stabilization, margin expansion, and product traction are supported by enough hard data to take seriously.

Still, this is not a no-brainer. Revenue has been shrinking, GAAP margins have been hit hard, and valuation depends on recovery more than current earnings. The stock works best for investors willing to underwrite a 12 to 24 month normalization story rather than demand immediate proof in the rearview mirror. In that sense, ON is less a pristine compounder and more a well-built machine coming out of a rough patch in the road. The engine still runs. The question is how much to pay before the speedometer catches up.

For a balanced investor, the answer is constructive but disciplined. ON deserves a Buy rating, especially on weakness, because the combination of financial resilience, product leverage, and improving cycle conditions creates a reasonable path to higher earnings and a better stock. Just do not confuse a good company with a stock that is always cheap. Markets enjoy that confusion far more than investors do.