ON Semiconductor Corporation (NASDAQ: ON) — Bought the Trough at $49; the Market Already Paid $110 for the Recovery

Report date: 2026-06-11 · Price: $110.17 (2026-06-10 close) · Market cap: ~$43–47B · Enterprise value: ~$44B · 52-week range: $44.56 – $134.92 Fiscal year: December · Segments: Power Solutions (PSG), Analog & Mixed-Signal (AMG), Intelligent Sensing (ISG)

⚡ Claude’s Take

This block is the author’s own subjective opinion. It is not investment advice and is general information only. The analysis that follows takes no position and carries no price target — it discusses valuation only as embedded expectations.

Verdict: HOLD / accumulate-on-weakness. A genuinely good cyclical at a full-but-not-absurd price. Not a short. Entry zone ~$80–90 (≈18–20x mid-cycle non-GAAP EPS of ~$4.50); at $110 the risk/reward is roughly balanced after a ~150% rally off the trough. Conviction: Medium.

onsemi is one of the better-run cyclical semiconductor franchises in the market, and the evidence is in how it behaved at the bottom. Through a brutal downcycle — revenue fell 27% peak-to-trough, GAAP operating income collapsed from $2.5B to $84M — the company still printed a record 23.7% free-cash-flow margin, kept a near-net-cash balance sheet, ran unusually clean non-GAAP accounting (it expenses stock comp; the add-backs are genuine one-time restructuring), and trebled its buyback to repurchase 28M shares at ~$49 while peers froze. Management bought low. That is the rare counter-cyclical capital allocation Marathon would applaud. Layer on three real secular content stories — silicon carbide for EV traction (≈55% of new China EV models at the 2026 Beijing show), the “only broad-based US power-semi supplier” position into the AI-data-center power tree (revenue doubling in 2026), and the high-margin Treo analog platform (60–70% gross margin, $1B target by 2030) — and you have a credible path to mid-cycle non-GAAP EPS of ~$4.50–5.00 over 2027–2028.

The problem is that the stock has already discounted most of this. At $110 it has roughly doubled off the $44 low, sits in the 95th percentile of its own ten-year valuation history (P/S 99th), and trades at ~26x forward and ~22–24x my mid-cycle EPS — a fair multiple for a business whose moat, honestly assessed, is narrower than TI’s or ADI’s. Only the AMG segment (51% gross margin, customer captivity) is a true Greenwald-grade moat; PSG (power discretes/SiC) is a contestable cost-and-scale position in a fragmenting, China-subsidized, oversupplied price-war industry, and the ISG image-sensor lead is eroding (a $230M inventory write-off in 2025 told that story). The framing is quality-cyclical-recovery-already-priced, not deep value and not a falling knife. The market has the recovery roughly right; it is not handing you the asymmetry that existed at $49. Single bullish trigger: an H2-2026 automotive replenishment cycle (not just the current “shipping-to-natural-demand” stabilization) layering on top of the AI-data-center ramp, stair-stepping gross margin back toward the mid-40s% and confirming the $5+ EPS path. Single bearish trigger: SiC oversupply and China competition keep PSG margins pinned in the low-20s% while the auto recovery stalls, leaving EPS stuck near ~$3 and exposing a ~26x multiple on under-earning. Tag: they bought the bottom; you’re being asked to pay for the top of the recovery.

1. Executive Summary

ON Semiconductor (onsemi) is the world’s #2 power semiconductor and discrete maker, the #1 automotive image-sensor supplier, and a top-tier silicon-carbide (SiC) franchise, selling into a portfolio that is 51% automotive, 28% industrial, and 21% “other” (AI data center, computing, consumer). Under CEO Hassane El-Khoury (since December 2020), the company executed a deliberate, multi-year strategic pivot — exiting low-margin commodity product lines, shrinking its fab footprint under the “Fab Right” program, and concentrating R&D and capital on intelligent power (SiC, IGBT, GaN, the Treo analog platform) and intelligent sensing. That transformation lifted through-cycle margins and returns materially: at the 2022–2023 peak the business earned a ~49% non-GAAP gross margin, ~31% ROE, and ~20% ROIC.

The company is now emerging from one of the deepest downturns in its history. Revenue fell from $8,253M (FY2023) to $5,995M (FY2025), a 27% decline driven by automotive and industrial inventory destocking and a sharp EV-demand air-pocket. GAAP results were further crushed by ~$980M of one-time charges (a $496M equipment impairment, $268M of excess-and-obsolete inventory write-offs concentrated in image sensors, and restructuring), collapsing GAAP operating income to $84M and GAAP EPS to $0.29. But this GAAP collapse is largely optical. Non-GAAP EPS was $2.35, free cash flow was a record $1.42B (23.7% margin) as the capex super-cycle ended, and the balance sheet finished the year near net-cash. The Q1-2026 call marked what management called a “clear inflection point”: automotive grew year-on-year for the first time in seven quarters, AI-data-center revenue grew 30% sequentially (set to double in 2026), gross margin expanded for a third consecutive quarter to 38.5%, and management guided to sequential margin expansion through 2026.

The investment tension is entirely about price, not business quality. The stock has rallied from a $44 low to $110 — management itself was buying at ~$49–60 — and now sits at ~26x forward earnings, in the 95th percentile of its own valuation history, pricing a substantial cyclical-plus-secular recovery. The bull case (AI-data-center power, SiC content, Treo/GaN, a margin march toward the 53% long-term target) is real but largely consensus and largely in the price. The bear case is that onsemi’s moat outside the AMG segment is contestable, the SiC market is oversupplied and China-subsidized, and the automotive “recovery” is so far only stabilization, not replenishment — leaving the multiple exposed if mid-cycle earnings land closer to $3 than $5. The 12.4%-of-float short interest signals that this debate is live and crowded. This report takes no position; it lays out what the price requires to be true.

2. Business Overview

What onsemi does. ON Semiconductor Corporation, operating under the onsemi brand, designs and manufactures intelligent power and intelligent sensing semiconductors — chips that convert, switch, condition, and protect electrical power, and that capture and process images and other real-world signals. Incorporated in Delaware in 1992 (spun out of Motorola’s semiconductor components group in 1999, IPO April 2000) and headquartered in Scottsdale, Arizona, the company today has ~22,600 employees and ~$6.0B of trailing revenue. Its stated strategic focus is the “high-growth megatrends” of vehicle electrification, energy-efficient power, industrial automation, and — most recently — AI-data-center power delivery (Fact — FY2025 10-K, filed 2026-02-09).

Three reportable segments (FY2025 revenue / segment gross margin):

(Source: FY2025 10-K, segment Note 3. onsemi reports segment gross profit, not operating income — the CODM evaluates segments on gross profit.)

End markets (FY2025): Automotive 51% (EV traction inverters, ADAS image sensors, in-vehicle networking, body/interior), Industrial 28% (energy generation/storage, EV charging, factory automation, aerospace/defense, medical), and Other 21% (AI data center, computing, consumer, networking). Automotive and industrial together are ~79% of revenue — the two longest-design-cycle, highest-content-growth, but also most cyclically-sensitive markets in the semiconductor complex.

How it makes money — the design-in/content model. Like its analog peers, onsemi sells physical components, not licenses, so revenue is transactional in form. But the economics behave annuity-like in the automotive and industrial core: power and sensing parts are designed into multi-year vehicle and equipment platforms after lengthy qualification (automotive-grade reliability, functional-safety certification), and once designed in they ship for the life of the platform — often 5–10+ years. The competitive event is the design win; the revenue is the multi-year tail. This is most true in AMG (the highest-captivity, highest-margin segment) and least true in commodity PSG discretes, where the part is more substitutable and pricing is more contestable. About half of distribution revenue is “fulfillment” (onsemi owns the end customer; distribution is just logistics), with the other half genuine distributor demand-creation into the mass market (Fact — Q1-2026 call).

The strategic transformation. The El-Khoury-era onsemi is a deliberately smaller, higher-margin company than the revenue-maximizing onsemi of the 2010s. Management has been exiting ~$300M of low-margin “non-core” revenue (roughly complete by end-2026), divesting subscale fabs, and concentrating capital on differentiated power (SiC/GaN/Treo) and sensing. The “Fab Right” strategy — fewer, larger, higher-utilization fabs, anchored by the East Fishkill 300mm site acquired from GlobalFoundries (closed end-2022) — is the manufacturing spine of the margin story. The thesis management is selling: a focused portfolio plus a lean cost structure equals structural operating leverage, so incremental revenue in the recovery drops through to expanding margins, earnings, and free cash flow.

Verdict. onsemi is a focused, vertically-integrated power-and-sensing franchise with one genuinely high-quality segment (AMG), a large but cyclically-and-competitively-pressured power segment (PSG), and a #1-but-eroding sensing segment (ISG). The model is structurally better than it was five years ago, but it remains concentrated in cyclical automotive/industrial end markets and exposed to the SiC capital cycle. It is a good business — not a great one in the TI/ADI sense.

3. Industry Dynamics

Where onsemi sits in the semiconductor complex. Power and analog/mixed-signal semiconductors occupy the structurally more attractive end of the industry — far from the boom-bust commoditization of memory and the winner-take-all capital intensity of leading-edge logic. Parts are designed on trailing-edge nodes (no $20B EUV fabs), product lifecycles are long, and the customer base is diversified across tens of thousands of industrial and automotive accounts. The broad analog TAM is ~$93–94B (2025); the power-semiconductor sub-segment onsemi anchors is ~$50B and growing with electrification. This is a good industry to own the cost-or-content leader in across a cycle (Fact — industry aggregators; the author TXN/ADI prior work, 2026-06-10).

But onsemi sits in the more contestable corner of that good industry. Unlike TI (cost-leader in broad catalog analog) or ADI (performance-leader in high-end signal chain), onsemi’s largest segment is power discretes and SiC — a segment with real but narrower differentiation, more exposure to commodity pricing, and a live capital-cycle problem.

The silicon-carbide capital cycle — the central industry dynamic. SiC is the swing factor for onsemi’s power business and a textbook Marathon capital cycle. The 2021–2023 EV boom drove the entire industry — onsemi, STMicroelectronics, Infineon, Wolfspeed, ROHM, Bosch — to announce SiC capacity simultaneously, underwriting a hockey-stick EV-adoption forecast. That capacity is now arriving into a sharply slower EV market (2024–2025 EV demand decelerated, especially in the West), producing a classic supply overshoot:

• SiC substrate prices fell 60%+ in 2024 as capacity outran demand (Fact — Yole/industry data, 2024–2025).
• Wolfspeed — the pure-play SiC leader — filed Chapter 11 in July 2025, emerging end-September 2025 after wiping out most of its equity. The marginal Western SiC producer effectively failed (Fact — public filings, 2025).
• Yole and others see SiC overcapacity persisting through 2027–2028, with Chinese producers (subsidized) adding capacity that distorts the normal capital-cycle clearing mechanism.

onsemi cut capex hard in response ($1.5B → $0.34B) and idled SiC capacity — the correct supply-side discipline — but the industry as a whole is still digesting. This is late-bust / early-trough in Marathon terms: capacity is starting to come out (onsemi’s restraint, Wolfspeed’s failure), which is constructive for survivors, but China’s subsidized supply may keep the recovery shallow and the pricing soft longer than a normal cycle.

Automotive — content growth offsetting a flat unit market. Global light-vehicle production (SAAR) is roughly flat-to-down, but semiconductor content per vehicle is rising structurally with electrification (EV powertrains carry 3–5x the power-semi content of an ICE), ADAS (more image sensors), and the shift to zonal/software-defined architectures (more networking and power-management content). onsemi’s Q1-2026 evidence: China passenger-vehicle units fell 6% year-on-year, yet onsemi’s China auto revenue grew — a content-and-share story, not a unit story (Fact — Q1-2026 call). The risk: EV adoption in the West has stalled, and the content tailwind is concentrated in China, where onsemi competes against subsidized domestic players.

AI data center power — the new growth vector. As AI racks scale from ~10–15kW toward 100kW+ and the industry transitions from 12V/48V to 800V HVDC architectures, the power-delivery content per rack explodes. onsemi positions itself as “the only broad-based US power-semiconductor supplier” able to address the full power tree from grid to processor, and cites content rising from ~$9,500/rack (120kW) toward ~$115,000/rack at 800V (Fact — Q1-2026 call; treat the rack-content figure as management TAM narrative, not booked revenue). This is real (an NVIDIA 800VDC collaboration, 30+ Flex Power programs) but it is a land-grab against entrenched incumbents — Monolithic Power (MPS) owns point-of-load, Infineon owns much of the power stage, and Vicor/Navitas/Power Integrations all compete. onsemi’s AI-DC revenue is ~$250M in 2025 (small), doubling to ~$500M in 2026.

Industry-structure verdict: a good industry, but onsemi sits in its more contestable, more cyclical, more China-exposed corner. The power/SiC segment is structurally attractive over a full cycle (electrification is a multi-decade tailwind) but is currently oversupplied, price-competitive, and distorted by Chinese subsidies. This is a better industry than memory and a worse one than high-end signal chain. Structurally good, cyclically and competitively pressured.

4. Competitive Position

The honest assessment: onsemi has one Greenwald-grade moat (AMG), a contestable cost-and-scale position (PSG/SiC), and an eroding intangibles lead (ISG). It is a materially lower-quality competitive franchise than Texas Instruments or Analog Devices — and the cycle just proved it.

The empirical moat test. The cleanest test of a moat is what happens to pricing and margin when demand falls. A true moat holds price through the cycle. onsemi’s gross margin fell from 49% to 33% (GAAP) / 38% (non-GAAP) and operating income fell 97% in two years. By contrast, ADI held a 73% gross margin and TI held ~34% operating margin at their troughs. onsemi’s through-cycle margin volatility is the tell: outside AMG, it does not have durable pricing power. The 27% revenue collapse and the one-year margin evisceration are the disconfirming evidence against a wide-moat thesis (Interpretation, from segment data above).

Segment-by-segment moat analysis (Greenwald taxonomy):

• AMG (Analog & Mixed-Signal) — genuine customer-captivity moat. This is the highest-quality leg: 51.1% gross margin (and rising through the downturn), built on long-lifecycle analog design-ins and the proprietary Treo platform (a configurable analog/mixed-signal platform onsemi builds application-specific products on, at 60–70% gross margin). Treo revenue rose 2.5x sequentially in Q1-2026 and targets $1B by 2030. This passes the share-stability and ROIC tests. It is onsemi’s best claim to a TI/ADI-style moat — but it is the smallest-growth path to scale and only ~38% of revenue.

• PSG (Power / SiC) — contestable cost-and-scale position, not a moat. onsemi is #2 globally in power discretes and a top-2/3 SiC player with a rare advantage: vertical integration into SiC substrate (via the 2021 GTAT acquisition; Pocatello, Idaho and Hudson, New Hampshire boule growth). That integration is a genuine cost/supply-security edge versus device-only rivals. But it does not confer pricing power: SiC is oversupplied, substrate prices fell 60%+, and onsemi holds ~22% device share behind ST (~29%) in a field that includes Infineon, ROHM, Bosch, and subsidized Chinese entrants. In commodity silicon discretes (IGBTs, MOSFETs) the position is even more contestable, competing against Infineon, Nexperia, ST, and a long tail. This is a good cost-advantaged business, not a moat — the kind of position that earns 47% gross margin at the top of the cycle and 24% at the bottom.

• ISG (Intelligent Sensing) — eroding intangibles lead. onsemi is the genuine #1 in automotive image sensors (~40% share), an intangibles/know-how position built over a decade. But it is the structurally weakest leg: the $230M 2025 inventory write-off (concentrated here) is direct evidence of fragility, OmniVision/Will Semi (China) and Sony are taking share, and ADAS sensor competition is intensifying. Management is pivoting ISG toward robotics, industrial autonomy, and multimodal sensing (SWIR, depth/time-of-flight) — a sensible reposition, but it is defending and redirecting a lead, not extending an unassailable one.

Direct competitive comparison:

Verdict: a narrow, segment-concentrated competitive advantage. onsemi has a real moat in AMG, a defensible-but-contestable cost position in PSG/SiC, and an eroding lead in ISG. Its vertical SiC integration is a genuine and underappreciated structural asset (supply security + cost), and its breadth across the AI-DC power tree is a real differentiator versus point-product rivals. But this is not a wide-moat compounder; it is a well-run cyclical with pockets of genuine durability. Durable advantage in AMG; contestable everywhere else.

5. Growth History and Forward Opportunities

Historical growth — a transformation that traded revenue for margin, then hit a cyclical wall. onsemi’s revenue history is not a smooth compounding story; it is a deliberate high-grading overlaid on a violent cycle:

The 2021–2022 surge was the post-COVID electrification/shortage boom plus the margin-accretive strategic repricing and mix-shift under El-Khoury (the company nearly doubled non-GAAP EPS from ~$3.10 to $5.33 in one year). The 2023–2025 decline was the mirror image: automotive and industrial customers, having over-ordered through the shortage, destocked aggressively, while EV demand decelerated and onsemi simultaneously chose to walk away from ~$300M of low-margin non-core revenue. So the revenue decline overstates the demand decline — part of it is self-inflicted high-grading that lifts margin.

The growth is high-quality where it is real, but it is cyclically and geographically concentrated. The content-growth thesis is genuine: onsemi grew China auto revenue while China auto units fell, proving content+share gains. But the growth is concentrated in (a) China EVs — where the competition is subsidized domestics — and (b) a still-small AI-data-center business.

Forward opportunities (ranked by credibility):

1. AI data center power (highest-conviction near-term). ~$250M in 2025, doubling to ~$500M in 2026, growing >30% sequentially in Q1-2026 (nearly double the expected rate). onsemi addresses the full power tree (high-voltage conversion, intelligent power stages, point-of-load) with silicon, SiC, SiC-JFET, GaN, and Treo/Vcore. The 800V HVDC transition (NVIDIA collaboration, Flex Power’s 30+ programs) is a multi-year content-expansion vector. This is the cleanest growth story and the one most responsible for the stock’s re-rating.

2. Silicon carbide content in EVs (high-conviction long-term, soft near-term). ~55% SiC share of new EV models at the 2026 Beijing Auto Show; expanded Geely/NIO collaborations on 900V platforms; the front-axle SiC + cost-optimized IGBT pairing. The secular content tailwind is real, but near-term pricing is soft (oversupply) and Western EV adoption has stalled. Growth is real; margin is hostage to the SiC cycle.

3. Treo analog platform (high-conviction margin story). 60–70% gross margin, 2.5x sequential revenue growth, $1B-by-2030 target, broadening across automotive (zonal/10BASE-T1S), industrial, and AI. This is the single best margin-accretive growth vector — it pulls the AMG mix richer.

4. GaN (gallium nitride) — emerging. A >$1.5B design funnel (incl. vertical GaN from the NexGen tuck-in), the new GaNEXUS portfolio (40V–650V, launched June 2026), 10 products sampling now + 20 more in 2H2026, revenue starting 2027. Optionality, not yet revenue.

5. Energy storage / “AI halo” industrial (recovering). SiC hybrid modules for utility-scale solar/ESS, ~40% revenue growth expected in 2026 with share approaching 60%, riding the AI-power-demand halo. A genuine industrial recovery vector.

6. Robotics/industrial sensing (ISG pivot, longer-dated). Multimodal sensing (image + depth/iToF + SWIR) design wins with a “leading global robotics platform” — repositioning the weakest segment toward a higher-growth adjacency.

Verdict: high-quality but concentrated and partly-priced growth. The forward opportunity set is genuinely good — content-driven, margin-accretive (Treo), and levered to two of the best secular themes (AI power, electrification). But it is concentrated in China EVs and a still-small AI-DC business, and the market has already capitalized much of it into a 95th-percentile multiple. High-quality growth, already substantially in the price.

6. Financial Quality

This is onsemi’s strongest chapter, and the one most misread by the headline P/E. The trough financials are far better than they look.

The GAAP collapse is optical; the cash generation is real. FY2025 GAAP operating income of $84.2M and GAAP EPS of $0.29 (which produce the absurd ~78–81x trailing P/E) are the product of ~$980M of one-time charges, almost all non-cash:

Strip these and normalized operating income was ~$1.1B (non-GAAP), non-GAAP EPS was $2.35, and — critically — operating cash flow was $1,759.8M against GAAP net income of just $121.0M. That ~$1.64B gap is the non-cash impairment/E&O charges plus D&A and working-capital release. The cash did not lie; the GAAP P&L did.

Free cash flow rose through the downturn — the harvest phase. Because the 2021–2023 capex super-cycle (SiC capacity + the East Fishkill build-out) ended just as the downturn hit, capital intensity collapsed and FCF expanded even as revenue fell 27%:

A 23.7% FCF margin at the trough is a remarkable quality signal — and management guides to mid-single-digit capital intensity “for the foreseeable future” (fully-utilized capacity needs ~25% more revenue before new capex is required). The capacity is built; the bill (depreciation) is being absorbed; the cash is being harvested.

Unusually clean non-GAAP accounting — a genuine positive. Three QoE checks that onsemi passes where many semis fail:

1. Stock-based comp is expensed inside non-GAAP. SBC is only $144M (2.4% of revenue) and is not added back — every gross-margin and EPS guide explicitly “includes share-based compensation.” The non-GAAP adjustments are restructuring, impairment, E&O, and acquisition-intangible amortization only — genuine one-timers. This is the analytically honest treatment and rare for the sector.
2. Goodwill is tiny ($1.68B, ~22% of equity). No roll-up balance sheet, no near-zero tangible book (contrast ADI’s $27B goodwill). Tangible book value is real.
3. The GAAP→non-GAAP gap, while wide ($0.29 vs $2.35), is fully explained by legitimate one-time items, not by scrubbing recurring costs.

Margins — peak-to-trough, and the recovery lever. Non-GAAP gross margin fell from 49.2% (2022) to 38.4% (2025); management attributes ~700bps of that gap to underutilization alone (Q4-2025 utilization 68%; Q1-2026 up to 77%). Every point of utilization is worth ~25–30bps of gross margin, and reaching low-90s utilization requires ~25% more revenue. The longer-term 53% gross-margin / 40% operating-margin target stacks utilization recovery (~700bps) + further Fab Right (~200bps) + absorbing divested-fab fixed costs (~200bps) + richer new-product mix (Treo/GaN, ~200bps+). This is achievable directionally but utilization-gated and multi-year — it is not yet earned.

Balance sheet — a fortress. Cash + short-term investments $2.55B against gross debt of $3.0B (0% converts due 2027 $805M; 0.50% converts due 2029 $1,500M; 3.875% notes due 2028 $700M), revolver undrawn. Net debt is only ~$457M — effectively net-cash on normalized EBITDA (~0.3x). The converts are cheaply financed (issued at prior-cycle low rates); the dilution risk is real only if the stock runs well above conversion prices (~40M potential shares across the two converts, partially hedged). Leverage is not a thesis risk.

Returns. At the 2023 peak: ROE ~31%, ROIC ~21% — genuinely high-return, evidence the transformation created real economics. At the 2025 trough: GAAP ROE ~1.5% / ROIC <1% (charge-distorted), but on normalized non-GAAP earnings ~11% ROE / ~8% ROIC — i.e., even at the worst of a brutal downcycle the business stays modestly above its cost of capital. That trough floor is the core quality signal.

One watch item: inventory. Total inventory is still 192 days (Q4-2025), including 76 days of “strategic” SiC inventory to be depleted over two years. Base inventory (117 days) is healthy, but the multi-year strategic SiC overhang means a soft SiC ramp could force additional marks. Inventory actually rose again in Q1-2026 (die-bank builds for the recovery).

Verdict: high financial quality, do economics improve with scale — yes. Clean accounting, record trough FCF, fortress balance sheet, and clear operating leverage into a recovery. The economics demonstrably improve with utilization. The only caveats are the utilization-gated margin recovery (not yet earned) and the SiC inventory overhang. The financial quality is the strongest pillar of the thesis — and it is genuine, not engineered.

7. Capital Allocation

Verdict up front: above-average, genuinely counter-cyclical capital allocation — the strongest behavioral signal in the thesis — tempered by a comp structure that doesn’t pay for per-share value and an absence of insider buying.

Counter-cyclical buybacks — management bought the bottom. This is the standout. onsemi’s stated policy is to return 100% of free cash flow to shareholders via buyback (no dividend). What matters is the timing, and onsemi got it right:

Management trebled repurchases into the 2025 trough at ~$49 average — versus ~$110 today — and bought another $346M at $60.54 in Q1-2026 at ~160% of free cash flow. Share count fell from 426.4M to 396.7M (~7%). This is textbook Marathon counter-cyclical behavior: flexing up buybacks exactly when the stock was cheapest and peers were frozen. A new $6.0B authorization (Nov 2025, running 2026–2028) gives ample runway, with a self-imposed guardrail (unlimited repurchase only while net leverage ≤2.75x). Buying back ~7% of the company at half today’s price is real, demonstrated value creation.

M&A discipline — no empire-building. Under El-Khoury, onsemi has deliberately avoided large dilutive M&A, doing only small, IP-focused, often earnout-protected tuck-ins:

(Fairchild $2.4B in 2016 and Quantenna $1.07B in 2019 predate El-Khoury; Quantenna’s Wi-Fi assets were largely de-emphasized as onsemi refocused on power/sensing.) Goodwill of just $1.68B corroborates the restraint. This is high-quality, capital-light portfolio shaping — building captive SiC/GaN/analog IP rather than overpaying for revenue.

Capex discipline. Capex was slashed from $1.54B to $341M as the SiC/EFK build-out completed, and management commits to mid-single-digit capital intensity going forward — the harvest phase. The earlier super-cycle investment (East Fishkill 300mm, SiC capacity) is what is now ramping Treo, SiC-JFET, and AI-DC products without new capital.

The caveats (kept honest):

1. Incentive metrics do not pay for per-share value. Comp is tied to correlated revenue + non-GAAP operating margin (STI) and new-product revenue + FCF + relative TSR (LTI). There is no EPS, no ROIC, and no absolute per-share/share-price metric in the formula. So the very buyback-driven per-share accretion that is management’s best capital-allocation achievement is not what they are paid on. Relative TSR is the only market link, and only as a modifier. This is standard semi-peer design, but it is “moderately,” not “strongly,” aligned.
2. Soft say-on-pay. The 2025 say-on-pay vote passed with only ~86% support (vs. 90%+ typical for peers), an irritant tied to an outsized $31M 2024 CEO grant. Management responded with design changes and outreach. A watch-it signal, not a crisis. (To his credit, CEO El-Khoury voluntarily forwent his entire 2025 cash bonus — a genuine alignment gesture.)
3. No insider buying; CFO sold into strength. There were no open-market purchases (code P) by insiders in the reviewed window — a neutral-to-mildly-cautionary signal. The CFO sold 90,000 shares in April 2026 at $80/$93/$100 (into the recovery; 10b5-1 plan status unconfirmed — an Open Question). The people closest to the numbers are not adding at these prices.

Verdict: strong, above-average capital allocation. The counter-cyclical buyback and M&A discipline are real, demonstrated, and value-creating — the behavioral evidence of a management team that understands cycles. The comp-design and insider-selling caveats temper the grade from “exemplary” to “strong,” but the core verdict is clear: management has allocated capital intelligently.

8. Changes and Headwinds — Last Two Years

Strategic / portfolio changes:

• The “Fab Right” manufacturing realignment culminated in a 2025 program cutting ~2,400 employees and ~12% of fab capacity, with ~$980M of charges — the most aggressive footprint reset in the company’s history, aimed at structurally higher utilization and margin.
• Ongoing exit of ~$300M low-margin non-core revenue (roughly complete by end-2026) — depresses headline revenue but lifts mix/margin.
• Three on-strategy tuck-ins (Qorvo SiC JFET, Aura/Vcore, NexGen/vGaN) building out the AI-DC power and GaN portfolios.
• New $6.0B buyback authorization (Nov 2025).
• GaNEXUS GaN portfolio launch (June 2026) — entering the GaN power market.

Leadership / governance changes:

• Board Chair Alan Campbell announced retirement (effective 2026 annual meeting).
• PSG Group President Simon Keeton resigned (announced March 2026, effective June 30, 2026) — a senior departure from the largest, most-pressured segment, worth flagging as a potential signal or simply succession.
• CEO El-Khoury (since Dec 2020) and CFO Trent remain stable at the top — continuity through the cycle.

The cyclical inflection (the dominant recent change): The Q1-2026 call marked management’s declared “clear inflection point” — automotive up year-on-year for the first time in seven quarters, AI-DC accelerating, three consecutive quarters of gross-margin expansion, and lead times stretching from 23 to 26 weeks with some technologies on allocation. Management frames this as “shipping to natural demand” (stabilization), explicitly not yet a replenishment/restocking cycle (which would be incremental upside).

Headwinds:

• SiC oversupply and pricing pressure (substrate prices −60%+, China subsidized capacity, overcapacity to 2027–2028).
• Western EV demand stall — the content tailwind is concentrated in China, where competition is subsidized.
• Automotive recovery is stabilization, not replenishment — the “real” recovery is not yet visible in order replenishment.
• Image-sensor (ISG) competitive erosion — the $230M write-off and share pressure from OmniVision/Sony.
• Input-cost inflation (energy/commodities) ahead of the pricing pass-through, a near-term margin headwind that reverses in 2H2026.
• A 95th-percentile valuation that has already priced much of the recovery.

Verdict: The strategic changes (Fab Right, portfolio high-grading, AI-DC/GaN build-out, counter-cyclical buyback) strengthen the long-term thesis and the margin model. The headwinds (SiC oversupply, China concentration, valuation) temper the near-term risk/reward. Net: a stronger business at a fuller price.

9. Risk Analysis

The dominant risks are valuation and the SiC/automotive cycle, not the balance sheet or accounting. This is a high-quality balance sheet with clean accounting facing a fully-priced stock in a still-uncertain cyclical/competitive environment. The catastrophic-loss risk is low (net-cash, diversified, real cash generation); the de-rating risk is high (you are paying a peak multiple for a recovery not yet fully delivered).

10. Valuation Discussion (Embedded Expectations)

No price target and no recommendation here — this section frames what the $110 price requires to be true.

The headline multiples are trough-distorted and must be discarded. Trailing P/E of ~78–81x and trailing EV/EBITDA of ~21.5x are meaningless — they divide today’s price by charge-crushed trough earnings. The only honest valuation approach for a deep cyclical is to value normalized/mid-cycle earnings power.

Earnings-power ladder:

(Mid-cycle assumes a return toward 2023-level revenue and margin over 2027–2028, ~390M shares net of continued buyback. Forward consensus P/E is ~26x — between the recovery and mid-cycle rungs.)

What the $110 price embeds. At ~26x forward and ~22–24x mid-cycle, the market is underwriting a substantial and durable recovery toward ~$8B revenue and mid-40s% gross margin — essentially a full cyclical normalization plus credit for the secular AI-DC/SiC/Treo content layer. The stock has already moved from pricing a trough ($44, ~18x trough non-GAAP EPS) to pricing a recovery ($110). The embedded expectation is no longer “will it survive the downturn” but “will it deliver the mid-cycle.” That is a much less asymmetric bet than the one management took at $49.

Peer comparison (forward P/E, the cycle-neutral lens):

(Source: yfinance, 2026-06-10; reconcile to filings.) onsemi at ~26x forward sits above the cheaper auto/industrial peer (NXPI ~16x) and in line with the premium franchises (ADI/TXN) — despite a narrower moat. On P/S (7.1x) it is cheaper than the premium names purely because its margins are trough-depressed (the same mechanical artifact flagged for KO/LIN — low margin inflates the apparent P/S discount). The relative case is not compelling: you are paying a premium-franchise forward multiple for a contestable-moat business mid-recovery.

Own-history valuation. onsemi’s composite valuation percentile is the 95th versus its own trailing ~10-year history (P/E 93rd, P/B 92nd, P/S 99th). Even adjusting for trough-distorted P/E, the P/B (5.9x, 92nd pct) and P/S (7.4x, 99th pct) confirm the stock is expensive against its own past — though that partly reflects the genuinely higher-margin, higher-quality business it has become.

Scenario analysis (illustrative, not a target):

• Bear (~$70-80): SiC oversupply + China competition pin PSG margin in the low-20s%, auto recovery stalls, mid-cycle EPS lands ~$3.00–3.50, multiple compresses to ~20–22x → ~$65–80. A ~25–40% downside with no business failure required — just a shallower, later recovery.
• Base (~$100-120): Cyclical recovery toward ~$8B revenue and mid-40s% GM by 2027–2028, non-GAAP EPS ~$4.50–5.00, ~22–24x → roughly today’s price. The base case validates $110 — which is precisely the issue: limited upside if base plays out.
• Bull (~$150-180): AI-DC and SiC content drive revenue past prior peak (~$9B+), margins march toward the 53% target, EPS reaches ~$6–7, the market pays a premium content multiple (~26–28x) → ~$150–180. Requires the secular overlay to genuinely re-rate onsemi from “cyclical” to “content compounder.”

Embedded-expectations verdict: At $110 the market is pricing the base-case recovery roughly correctly. The asymmetry that existed at $49 (pricing a trough) is gone; what remains is a balanced-to-slightly-unfavorable risk/reward — meaningful downside if the recovery disappoints, modest upside if it merely meets expectations, and real upside only if the secular content story re-rates the whole franchise. The price requires the mid-cycle to be delivered; it no longer offers a margin of safety for it failing to.

11. Variant Perception

Consensus view. onsemi is a quality cyclical at a cyclical inflection — the worst is behind it, AI-data-center power and SiC content are secular growth vectors, margins are inflecting, management is excellent at capital allocation, and the stock deserves a premium multiple as it transitions from “cyclical” toward “secular-content compounder.” Analyst rating ~3.97/5 (modestly bullish), with recent price-target raises (Wells Fargo to $140).

The strongest bull case. The transformation is real and the operating leverage is enormous. A focused, lean onsemi printing 38% trough gross margin and 24% trough FCF margin will, on a real recovery to ~$8B+ revenue and 90%+ utilization, deliver the 53% gross-margin / 40% operating-margin model — implying $6–7 of normalized EPS. Layer on three secular content stories (AI-DC power doubling, SiC EV content at 55% China share, Treo at 60–70% GM scaling to $1B) that justify a structural re-rating, and the stock works toward $150–180. Management is buying, the balance sheet is a fortress, and the short interest (12.4%) is fuel for a squeeze.

The strongest bear case. You are paying a premium-franchise multiple (~26x forward, 95th-percentile valuation) for a business whose moat — proven by a 27% revenue collapse and a margin evisceration — is contestable outside the small AMG segment. The largest segment (PSG/SiC) is in a China-subsidized, oversupplied price war that may keep margins depressed through 2027–2028; the auto “recovery” is only stabilization; the content growth is concentrated in China against subsidized domestics; the AI-DC story is a small land-grab against entrenched MPS/Infineon; and the CFO is selling into strength. If mid-cycle EPS lands at $3.00–3.50 rather than $5+, the stock is worth $70–80. The market has extrapolated a clean V-recovery that the cycle and the competition may not deliver.

The 3–5 assumptions that matter most:

1. Does the automotive market move from stabilization to genuine replenishment in 2H2026? (Determines whether utilization — and thus margin — actually recovers.)
2. Does SiC pricing stabilize, or does China-subsidized oversupply persist through 2027–2028? (Determines PSG margin and the whole power thesis.)
3. Does AI-data-center power scale into a multi-hundred-million-to-billion-dollar franchise, or stall as a land-grab against incumbents?
4. Is mid-cycle gross margin ~45% (recovery) or ~50%+ (the 53% target)? (The difference between $4.50 and $6+ EPS.)
5. Does the market keep paying a premium-content multiple, or re-rate onsemi back toward a cyclical multiple?

What would falsify each side. Bull falsified: a 2025-style 2H deceleration, SiC marks/further inventory write-offs, or AI-DC growth stalling — any of which exposes “normalized” earnings well below the $5 the price assumes. Bear falsified: auto replenishment + sustained AI-DC doubling + a gross-margin stair-step into the mid-40s%, confirming the operating-leverage model and the $5+ EPS path is real, not a hope.

Variant takeaway: The market and the bulls are probably directionally right on the business (it is a good, recovering franchise) but are paying a price that already embeds the win. The variant perception is not “this is a bad company” — it is “this good company is priced for a recovery it has not yet earned, against a competitive/cyclical backdrop more contestable than the multiple implies.”

12. Fact vs. Interpretation

13. Open Questions

1. Is the automotive turn replenishment or just stabilization? Management says “shipping to natural demand” — the incremental restock has not happened. The single biggest swing factor for 2H2026 margin.
2. When does SiC pricing stabilize? Substrate oversupply and China subsidies could persist through 2027–2028; onsemi’s 76 days of strategic SiC inventory is exposed to further marks if the ramp disappoints.
3. What is the true mid-cycle gross margin? 45% (recovery) vs 50%+ (the 53% target) is the difference between $4.50 and $6+ EPS — and it is utilization- and mix-dependent.
4. Can AI-DC power scale past a land-grab? Is onsemi taking durable sockets from MPS/Infineon, or winning early designs that incumbents reclaim?
5. Why is the CFO selling into strength (90,000 shares at $80–100 in April 2026), and were those sales 10b5-1-planned? (Form 4 bodies not mirrored locally — unresolved.)
6. Does the PSG President’s departure (Simon Keeton, eff. June 2026) signal anything about the most-pressured segment, or is it routine succession?
7. What are the exact convertible conversion prices, and what is the 2027/2028 refinancing/dilution path at $110?
8. Does ISG’s robotics/industrial pivot restore a defensible position, or is the auto-CIS lead structurally eroding?

14. What Must Be True

Bull case — what must be true:

1. Automotive moves from stabilization to genuine replenishment in 2H2026, lifting utilization toward the low-90s%.
2. SiC pricing stabilizes and onsemi’s vertical integration protects its margin as Western EV demand recovers.
3. AI-data-center power scales into a durable multi-hundred-million-dollar (toward $1B+) franchise.
4. Gross margin stair-steps toward the mid-40s% and onward to the 53% target; mid-cycle EPS reaches $5–7.
5. The market re-rates onsemi from “cyclical” toward “secular content compounder.”

Falsification test: A 2025-style second-half deceleration, any fresh SiC/inventory write-off, or AI-DC growth stalling below the doubling guide — any one falsifies the clean-recovery thesis and exposes mid-cycle EPS below $5.

Bear case — what must be true:

1. SiC oversupply and China-subsidized competition keep PSG gross margin pinned in the low-20s% through 2027.
2. The automotive recovery stalls at stabilization; no replenishment cycle materializes.
3. AI-DC remains a sub-$1B land-grab that incumbents (MPS/Infineon) defend.
4. Mid-cycle EPS lands at $3.00–3.50, not $5+, and the ~26x multiple compresses toward 20x.

Falsification test: Auto replenishment + sustained AI-DC doubling + a gross-margin march into the mid-40s% across 2026–2027 — confirming the operating-leverage model is real — falsifies the bear and validates a $5+ EPS path.

The synthesis: Both cases agree the business is good and the balance sheet is a fortress. They disagree on how much of the recovery is already in the $110 price and how durable the margin recovery is against a contestable competitive backdrop. At $49 the bull had a wide margin of safety; at $110 the question is whether the mid-cycle gets delivered on schedule — a coin-flip the bear can win without any business failure.

15. Source Appendix

See Appendix B — Source Appendix (separate deliverable) for the full cited source list. Primary sources: ON Semiconductor FY2025 Form 10-K (filed 2026-02-09), FY2023 Form 10-K, Q1-2026 Form 10-Q, FY2025 DEF 14A proxy (2026-04-02), Form 4 filings; Q1-2026 (2026-05-04) and Q4-2025 (2026-02-09) earnings-call transcripts; SEC EDGAR XBRL financial data; market-data aggregators; yfinance market data; Yole/industry SiC and image-sensor data; prior Texas Instruments and Analog Devices analyses.

APPENDIX A — Standard Diligence Questionnaire

ON Semiconductor Corporation (NASDAQ: ON) · As of 2026-06-11 · Fact / Interpretation / Assumption labeled where it matters.

General

What thoughtful questions have other investors asked about this company? The central debates: (1) Is the cyclical recovery real or a false start (auto stabilization vs replenishment)? (2) How much SiC margin damage from China-subsidized oversupply, and when does pricing stabilize? (3) Can AI-data-center power become a needle-moving, durable franchise against MPS/Infineon? (4) Is the 53% long-term gross-margin target credible, and what utilization does it require? (5) After a double off the low, is the stock pricing too much recovery? The 12.4%-of-float short interest shows the bear case is actively held.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Decisively a low (Fact). FY2025 non-GAAP EPS $2.35 vs a $5.33 prior peak; GAAP EPS $0.29 (charge-distorted). Q1-2026 marked management’s declared cyclical inflection. Driven by external environment or internal actions? Both: the downturn was external (auto/industrial destock, EV-demand air-pocket), amplified internally by a deliberate ~$300M low-margin revenue exit and the Fab Right restructuring (Interpretation). How stable are revenues? Cyclical and volatile — revenue swung +24% (2022) then −1%/−14%/−15% (2023–25). Long automotive/industrial design-in tails provide a base but not stability through a destock. Outlook for products/services? Improving: AI-DC doubling in 2026, auto turning +YoY, Treo scaling, GaN launching. Recovery underway but utilization-gated. How big will this market be? Power semis ~$50B growing with electrification; broad analog ~$93–94B; AI-DC power and SiC are secular multi-year growth vectors. Growing, global, with China the swing region.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More, in onsemi’s core: SiC is oversupplied with subsidized Chinese entrants; image sensors face OmniVision/Sony; AI-DC power is a contested land-grab (Interpretation). How profitable is the business (ROIC/ROE)? Peak (2023): ~31% ROE, ~21% ROIC — genuinely high. Trough (2025): ~11% ROE / ~8% ROIC on normalized non-GAAP earnings (GAAP near-zero on charges). Stays above cost of capital even at trough (Fact/calc). How profitable is the industry / barriers to entry? Moderately profitable; barriers are real (automotive qualification, functional-safety certification, SiC substrate vertical integration, capital intensity) but lower than high-end signal chain. Many competitors per segment. Can the business be easily understood? Yes — it sells power and sensing chips into autos, industrial, and data centers. The complexity is in the cycle and the segment-mix, not the model. Undermined by foreign low-cost labor? The threat is subsidized Chinese semiconductor competition (SiC, image sensors), not labor arbitrage — a real structural risk in the contestable segments. Do brands matter? Modestly — “onsemi” carries automotive-grade reliability credibility, but the buying decision is design-win/performance/price, not brand. Nature of competition? Design-in driven in auto/industrial (multi-year, sticky); more price-competitive in commodity discretes and oversupplied SiC. Customers’ switching costs? High once designed into a vehicle/equipment platform (re-qualification cost/risk); low at the design-decision point, especially for substitutable discretes.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The SiC substrate vertical-integration capability (GTAT IP, Pocatello/Hudson capacity) and the Treo platform IP are worth more than book; the East Fishkill 300mm capacity is a strategic asset (Interpretation). Off-balance-sheet liabilities? None material flagged; convertible-note dilution (~40M potential shares) is disclosed, not off-balance-sheet. How conservative is the accounting? Unusually conservative for a semi (Fact): SBC fully expensed in non-GAAP, tiny goodwill ($1.68B), one-time charges genuinely one-time, OCF >> GAAP NI. A QoE positive. How CapEx-hungry? Was very (18.6% of revenue in 2023); now harvesting at ~5.7% with mid-single-digit guidance — the super-cycle is built and paid for.

Capital Allocation & Management

How much FCF, and how is it used? Record $1.42B FCF (2025); policy returns 100% to shareholders via buyback (no dividend). Trebled buyback to $1.38B at the ~$49 trough — counter-cyclical (Fact). Significant acquisitions recently? Only small on-strategy tuck-ins (Qorvo SiC JFET $118.8M, Aura/Vcore ≤$144M mostly earnout, NexGen/vGaN). No large dilutive M&A under El-Khoury. Goodwill $1.68B confirms discipline. Buying back shares? Yes, aggressively and counter-cyclically — share count down ~7% (426M→397M). New $6.0B authorization (2026–2028). Issuing shares to insiders? SBC modest (2.4% of revenue), expensed; not a dilution machine. Compensation policy? STI on revenue + non-GAAP op margin; LTI on new-product revenue + FCF + relative TSR (70% PBRSU). No EPS/ROIC/per-share metric — moderately aligned. Say-on-pay only ~86% (soft flag). CEO forwent his 2025 cash bonus (alignment gesture). Motivations of management? Building a focused, higher-margin franchise; demonstrably willing to sacrifice revenue for margin and to buy stock low. Caveat: no open-market insider buying; CFO sold into strength (Apr-2026, $80–100).

Valuation & Market Data

ADR, MLP, or K-1? No — US-domiciled C-corp common stock, standard 1099. Dividend policy? None — 100% of capital return via buyback. How profitable? Trough-depressed now (GAAP), but ~49% peak non-GAAP gross margin and 31% peak ROE show the underlying quality. Net income diverging from cash flow? Yes, dramatically and favorably in 2025 — OCF $1.76B vs GAAP NI $121M (the gap is non-cash impairment/E&O). Cash earnings power is intact (Fact).

Risks & Downside

What would cause the stock to decline? A valuation de-rate from the 95th percentile (the highest-probability risk); SiC pricing war persisting; the auto recovery stalling; AI-DC disappointing; a fresh inventory write-off. Risk of catastrophic loss? Low — net-cash balance sheet, $4B liquidity, diversified, real cash generation, clean accounting. No solvency risk. Chance of total loss? Negligible. The risk is de-rating and under-earning, not impairment of capital.

Recent News & Events

Has the business environment changed recently? Yes — Q1-2026 marked a cyclical inflection (auto +YoY first time in 7 quarters; AI-DC +30% QoQ; 3rd consecutive quarter of gross-margin expansion; lead times stretching, some technologies on allocation). Wells Fargo raised its target to $140 (Jun-2026); GaNEXUS GaN portfolio launched (Jun-2026). Significant acquisitions? Only tuck-ins (above). Change in accounting policies? None material; the 2025 charges are restructuring/realignment, not policy changes. Recent changes — markets, facilities, management? Fab Right footprint reduction (~12% capacity, ~2,400 jobs); East Fishkill ramping new products; Board Chair retiring (2026 AGM); PSG President Simon Keeton resigning (eff. Jun-2026); new $6.0B buyback.

APPENDIX B — Source Appendix

ON Semiconductor Corporation (NASDAQ: ON) · Report date 2026-06-11 · Primary sources prioritized; all financial figures reconciled to SEC filings.

Primary — SEC filings (EDGAR, CIK 0001097864)

Primary — Earnings-call & event transcripts (company earnings-call & event transcripts)

Secondary — Market & industry data

Related prior coverage

Notes on data reliability

• Trailing P/E (~78–81x) is a trough artifact — discarded in favor of forward (~26x) and mid-cycle (~22–24x) earnings-power analysis.
• Own-history P/E percentile (93rd) is trough-distorted; P/B (92nd) and P/S (99th) percentiles corroborate an expensive own-history valuation more reliably.
• Non-GAAP figures are unusually clean (SBC expensed; only one-time restructuring/impairment/amortization excluded) — relied upon for earnings power.
• Form 4 bodies not mirrored locally — insider read is a representative sample, not an exhaustive 5-year sweep (Open Question).
• Mid-cycle EPS (~$4.50–5.00) and scenario price ranges are the author’s model assumptions, not company guidance or price targets.