ON Semiconductor ON

Reference table

1. What it is

ON Semiconductor (Nasdaq: ON), branded onsemi, is a Scottsdale, AZ-based global supplier of intelligent power and sensing semiconductors — power discretes and modules, analog/mixed-signal ICs, and image sensors handling power switching, energy conversion, signal conditioning, and imaging. Founded 1992; ~22,600 employees per the current FMP profile (the feed listed ~26,400 on 7/3 — consistent with post-restructuring headcount, though we treat the delta as a data-vintage change, not a verified fact); CEO Hassane El-Khoury. End-markets: automotive (EV powertrains via EliteSiC silicon carbide, ADAS sensing), industrial (energy storage, solar, automation), and the fast-growing AI data center power line.

Segments. Three groups — Power Solutions Group (PSG), Analog & Mixed-Signal Group (AMG), Intelligent Sensing Group (ISG). FMP's FY25 segment feed remains incomplete (PSG $2,805M and ISG $928M captured; AMG dropped — same gap flagged on 7/3). The cleaner current mix, from the Q1'26 8-K cited in the prior dive: PSG $736.6M (49%, +14% YoY) — the SiC/power engine; AMG $540.4M (36%, −5%); ISG $236.3M (16%, +1%); total $1,513.3M, +4.7% YoY. For scale on the cycle: PSG peaked at $4,449M in FY23 (FMP segment feed) — the engine segment is still ~35%+ below its own peak.

Geography: FMP's geographic feed reports ship-to hubs (Hong Kong/Singapore/UK) and collapses FY25 to a single "Other" line — unusable; the honest read is global, Asia-weighted demand with auto/industrial as the swing.

The new strategic layer since the last dive: onsemi is on NVIDIA's named partner list (May 2025) for the 800 VDC data-center power architecture, with both a SiC device franchise and a power-stage catalog — one of the few partners with both. Full 800V production is timed to NVIDIA's Kyber rack generation in 2027 (SemiAnalysis, via CNBC 7/6, reports a slip to 2028; NVIDIA denies). (External context, verified via deep research 2026-07-19 — not from the FMP dataset.)

2. The expert thesis — why the panel is bullish (traceable)

The KB picture has changed — barely — since 7/3, when this name had zero coverage. There is now exactly one tagged claim:

• forward_guidance · 2025-05-16 · bullish · conviction 66 — "Analog/robotics-supply-chain semis are calling the cyclical bottom — Microchip's CEO said the March-quarter decline marks the bottom, echoed by Infineon, Regal Rexnord, and [ON Semiconductor]."

Two honest caveats. First, it is 14 months old — a dated bottom-call, not live coverage. Second, it has been validated by the subsequent data: ON's revenue did bottom in Q1'25 ($1.446B) and turned YoY-positive by Q1'26 — which raises our trust in the cycle read but adds nothing about today's price. Breadth is 1 claim from 1 voice; that is not a conviction panel.

Thematic (untagged) context only: a KB voice (no_priors, 2026-06-18, bullish/66) frames GaN/SiC/InP as next-generation power materials — a tailwind to onsemi's category, but it does not name ON, so it is not counted. On the cautionary side, Dylan Patel (KB, Feb 2026) warns China power-chip suppliers are matching or undercutting Western firms on trailing edge, with China SiC/BYD insourcing the structural risk — again thematic, not ON-tagged, and it appears in §8 and §11 with its date.

Net: the verdict below is predominantly fundamentals- and quant-driven. Conviction rating: Low. Street context (not our anchor): consensus target $110.13, median $116, high $150, low $62; 24 Buy / 22 Hold / 1 Sell.

3. Synthos scores & the Bull / Base / Bear cases

The one-glance judgment — three scores, 0–10, each anchored to real metrics (not probabilities we can't honestly calibrate):

The three cases (our own scenario model — assumptions shown; each target is a ~12–18-month fair value). We deliberately do not attach probabilities: the base case is the expected path; the cases bound the range.

Testing the asymmetry frame (the editorial question for this refresh): the bull argument runs (a) 2.4× EPS torque at 15.5× the FY28 number, (b) ~10 points of mechanical GM recovery, (c) an unpriced 800V option, (d) a −34% correction already done, giving ~$73–78 downside vs $115–130 upside ≈ 2.5–3:1. Our verdict on the frame, with numbers: right shape, wrong price. (a) is real but leans on a 6-analyst FY28 estimate that embeds peak net margins and a ~380M share count. (b) is mostly already delivered — quarterly GM went 20.3% → 38.5%; from here to a realistic low-40s mid-cycle is ~2–4 points, not 10 (the 10-point framing uses the FY25 annual 32.3% base; 2022's 49% peak was shortage-priced). (c) is genuine and the strongest new element. (d) conflates a technical floor (rising 200-DMA ~$73, ~17× FY27E) with the fundamental bear ($53): from $88, upside to $130 is +$42 vs −$35 to the bear — roughly 1.2:1, not 3:1. Buy the same stock at $75–78 and the math becomes bear −$23 / bull +$53 ≈ 2.3–2.8:1 with the base case alone worth +22–27%. That is why the zone, not the current price, is the trade.

Synthos fair value = the base case, ~$95 (+8%), range $53–$130. Our base sits below the Street's $110 because we haircut the multiple for beta-2 cyclicality and refuse to capitalize an unquantified 800V option. This is a tracked call — the Forecaster Scorecard grades it once it matures.

4. Exponential Potential

Synthos separates compounders from exponentials. onsemi remains a recovering cyclical with exponential product lines bolted on — but the option leg is better-defined than it was on 7/3:

• Forward growth: consensus revenue $6.48B (FY26E) → $7.21B (FY27E) → $8.17B (FY28E), a ~12.3% CAGR; EPS $3.09 → $5.70, ~36% CAGR — trough-leverage plus buybacks, not a secular ramp. FY28E revenue only returns to the FY23 level ($8.25B).
• The 800V option (new, external — verified 2026-07-19): onsemi is a named NVIDIA 800 VDC silicon partner (May 2025 list), one of few with both a SiC device franchise and a power-stage catalog. Full production is timed to Kyber 2027 (reported slip to 2028 — SemiAnalysis via CNBC 7/6 — denied by NVIDIA). No public source quantifies per-rack dollar content for any partner, and the consensus model above is an auto/industrial recovery model with no DC-SiC revenue line — so any 800V win is upside to numbers built without it. We deliberately leave it out of the base FV; at $88 you are not paying for it.
• Acceleration: positive at the line level (AI-DC >2× YoY, +30% sequential in Q1'26; PSG +14% YoY — prior dive/8-K), still negative at the franchise level (total revenue −28% below the 2022 peak of $8.33B).
• The counter-option (China): the same SiC/power market is the target of Chinese insourcing (BYD) and trailing-edge price competition (KB voice, Feb 2026). The option leg and the threat leg point at the same P&L line.
• Reinvestment: capex has collapsed to $341M FY25 (5.7% of revenue, vs $1.54B in FY23) — great for FCF, but it also says the heavy-investment growth phase is behind, and capacity leadership in SiC is now something rivals can buy.

Exponential Potential: Moderate (5/10) — unchanged in score, upgraded in specificity: the option is now named, dated, and verifiably unpriced; the offsets (timing, slip risk, China) are equally concrete.

5. Financials (real numbers — FMP annual/quarterly, pulled 2026-07-20)

• Revenue: FY25 $5.995B, −15.4% (FY24 $7.08B, FY23 $8.25B, FY22 peak $8.33B — cumulative −28% peak-to-trough). Quarterly: Q1'25 $1,445.7M → Q2 $1,468.7M → Q3 $1,550.9M → Q4 $1,530.1M → Q1'26 $1,513.3M (+4.7% YoY — the first YoY growth after two down years).
• Gross margin (the recovery spine): GAAP quarterly 20.3% (Q1'25) → 37.6% → 37.9% → 34.9% (Q4'25) → 38.5% (Q1'26). FY25 annual 32.3% vs the 49.0% FY22 peak. Note Q4'25's dip to 34.9% — the rebuild is not monotonic. Management's Q2'26 guide: non-GAAP GM 38–40%.
• The Q1 charges (investigated honestly): GAAP operating income was −$573.7M in Q1'25 and −$53.4M in Q1'26 — negative in the seasonally first quarter two years running even as Q1'26 gross margin hit 38.5%. The arithmetic: Q1'26 gross profit $583.1M less a ~$295M/qtr R&D+SG&A run-rate implies roughly $340M of other operating charges (restructuring/impairment/amortization) that the FMP dataset does not itemize. What the file does show: FY25 cash flow carries $653.5M of "other non-cash items" added back, income quality (OCF/net income) runs 2.4×, and FY25 equity fell $8.80B → $7.67B — all consistent with large, predominantly non-cash write-offs rather than cash restructuring. Non-GAAP EPS was positive both quarters ($0.55 Q1'25, $0.64 Q1'26, both beats). We cannot fully explain the Q1'26 charge from this dataset — that is a flagged, must-answer item for the 2026-08-03 call (§10). Two consecutive years of "one-time" Q1 charges is a pattern, and a third would move our risk score.
• Earnings: FY25 GAAP net income $121M / EPS $0.29 (charge-distorted); FY25 consensus-basis EPS $2.33; TTM net income/share $1.46. TTM margins: gross 37.2%, EBITDA 20.1%, net 9.5%.
• Cash flow (the trough-proof): FY25 operating CF $1.76B, capex −$341M, FCF $1.42B — positive straight through the worst revenue year, FCF yield now 4.2%. Buybacks $1,377.6M in FY25 (~97% of FCF) plus $377M of debt paydown; $346M repurchased in Q1'26 alone (~160% of that quarter's FCF — prior dive/8-K).
• Balance sheet: cash & ST investments $2.55B, total debt $3.47B, net debt $1.32B, net-debt/EBITDA 0.8×, current ratio 4.9×. Leverage is not the risk.
• Data caveat: FMP's FY25 annual operating income ($747.6M) does not reconcile with the sum of quarters ($137.1M) — the feed classifies the Q1'25 impairment differently at annual vs quarterly level. We rely on the quarterly statements and cash flow, and flag the feed inconsistency here.

6. Valuation — priced in or room?

At $88.06 the multiples are: ~60× trailing GAAP EPS (trough-distorted — treat as a flag, not a valuation), 28.5× FY26E ($3.09) → 20.5× FY27E ($4.30) → 15.5× FY28E ($5.70), EV/S 5.8×, EV/EBITDA 29× TTM, price/FCF ~24×.

What 17 days changed: price −3.5% ($91.22 → $88.06) while estimates held essentially flat (FY26E $3.08→$3.09, FY27E $4.28→$4.30, FY28E $5.79→$5.70) — so the stock got modestly cheaper against an unchanged earnings path. The reverse read moved accordingly: at a mid-cycle ~22×, today's price implies ~$4.00 of EPS — now slightly below the FY27 consensus of $4.30, versus 7/3 when the price demanded essentially the full number. The market has gone from pricing the recovery fully to pricing ~93% of it. Progress, not a bargain.

Honest cross-checks, both directions. For the bulls: 15.5× FY28E is genuinely cheap if $5.70 lands — but that estimate rests on only 6 EPS analysts (vs 20 for FY26), embeds a 26.5% net margin equal to the FY23 peak, and assumes buybacks shrink the count to ~380M shares. For the bears: on EV/EBITDA the stock screens fuller than on P/E — ~15.9× FY27E EBITDA ($2.22B) against a 12–14× mid-cycle band for analog/power semis implies $63–74/share, one reason we resist rounding our FV up despite the 800V option. The P/E path flatters via an 11% effective tax rate and the buyback.

Street (context): consensus $110.13 (was $109), median $116, high $150 (raised from $140), low $62; 24 Buy / 22 Hold / 1 Sell (one Buy became a Hold since 7/3). FMP letter rating B−, P/E sub-score 1/5. Bottom line: fairly valued at $88 — cheap only in the $73–81 zone or on an estimate-raising print.

7. Technicals (from the tech block)

• Trend: still unconfirmed. $88.06 sits below the 50-DMA ($109.17, falling — it was $110.69 on 7/3) and above the 200-DMA ($73.34, rising — from $71.04). The averages are converging on the price from both sides; no trend signal yet.
• Location: −34.2% off the 52-week high ($133.93), +96.2% off the low ($44.90) — the correction has deepened slightly since 7/3 (−31.9% then). Max drawdown from peak equals the current drawdown: the stock is at its correction low zone, not bouncing from above it.
• Momentum: RSI(14) 49.4 — repaired from 37.6 on 7/3, the oversold condition is worked off without a price rally (constructive divergence). MACD −6.0, still negative and slightly worse than 7/3's −4.9.
• Relative strength: +48.2% 12-mo vs SPY +18.7% / QQQ +24.8% (the 12-mo figure compressed from +70% on 7/3 as strong base months roll off); 3-mo +6.1% vs SPY +5.0% — the price action has gone from big outperformance to market-line.
• Read: a coiling consolidation between a rising 200-DMA (~$73) and a falling 50-DMA (~$109), momentum neutral. Supports Hold-and-stalk: the $73–81 zone is where technical support (rising 200-DMA) and valuation support (~17–19× FY27E) coincide; a reclaim of the 50-DMA on a clean print is the momentum-confirmation alternative.

8. Moat & competitive position

Moat: moderate, not wide — unchanged. Edges: (1) the EliteSiC silicon-carbide franchise (EV powertrains, now 800V datacenter power) with capacity and automotive qualification barriers; (2) breadth across power discretes, analog, sensing with long auto design-in cycles (Geely, NIO, a North American OEM — prior dive/8-K); (3) the dual SiC + power-stage catalog position on NVIDIA's 800 VDC list, which few partners have (external, 2026-07-19). Against it: power/analog semis are competitive and partly commoditized, ROIC is 5.4% at this point in the cycle, and — the sharpest new framing since 7/3 — the China vector: a KB voice (Dylan Patel, Feb 2026) argues China power-chip suppliers are matching or undercutting on trailing edge, and China SiC/BYD insourcing is the biggest structural risk to margins in exactly the franchise the bull case leans on. A moat that is real in qualification and capacity, and under active siege on price.

Peer set: FMP's supplied list remains a poor comp set (CDW, Check Point, SS&C, Tyler, Toast, Jacobs are not power-semi peers). Usable names in it: STMicroelectronics ($55.7B) — the closest direct comp — plus GLOBALFOUNDRIES ($31.6B), UMC ($51.4B), Entegris ($20.8B). The true frame is STM, Infineon, Texas Instruments, and the Chinese SiC entrants; ON's forward premium to STM is justified only if the SiC/AI-power mix wins share at defended prices.

9. Management, capital allocation & guidance

• Capital allocation: buyback-heavy — $1,377.6M repurchased in FY25 (~97% of FCF) plus $376.7M net debt paydown; $346M more in Q1'26 (~160% of quarterly FCF — prior dive/8-K). No dividend. Sustained repurchase near/above 100% of FCF into a cyclical recovery is defensible only while the recovery holds — it remains on our watch list.
• Insider activity (through 2026-07-02): routine. The window shows director annual equity awards (2026-05-14/18), a small tax-withholding disposition by the AMG & ISG group president at $127 (May), a Form 3 for the newly appointed PSG group president (Achyut Shah, May — a leadership change in the key segment worth knowing), and a 116-share director award at $91.22 (2026-07-02). No discretionary open-market buying or selling of note.
• Guidance (management's book, half-weighted): from the Q1'26 8-K (2026-05-04, carried from the prior dive): Q2'26 revenue $1,535–1,635M, non-GAAP GM 38–40%, non-GAAP EPS $0.65–0.77; "moved beyond the cyclical trough." The Street's $0.71 / $1,588M sits mid-guide. Track record check from the earnings calendar: four consecutive non-GAAP EPS results at or above estimates (Q2'25 $0.53 in-line; Q3'25 $0.63 vs $0.59; Q4'25 $0.64 vs $0.62; Q1'26 $0.64 vs $0.61) — modest, consistent beats.

10. Catalysts & what to watch

Next earnings: 2026-08-03 (Q2'26). The five things to listen for, in order:

1. Gross-margin trajectory — Q2 landing in the 38–40% guide and, more important, a Q3 guide pointing at 40%+; utilization commentary tells you how much "mechanical" recovery is left.

2. First quantification of DC/AI revenue — any dollar disclosure, per-rack content color, or Kyber timing detail. This is the swing item: it converts the free option into an underwritable line (and tests the reported 2028 slip).

3. Do the charges stop? A clean GAAP quarter — no new restructuring/impairment — after two charge-laden Q1s. Ask what the Q1'26 ~$340M of below-gross-line operating charges was; the filing data we hold cannot itemize it.

4. Auto/industrial demand breadth — AMG turning positive YoY would broaden a recovery currently carried by PSG; China EV order commentary (Geely, NIO) and any mention of SiC pricing pressure.

5. Buyback pace vs FCF and the balance-sheet absorption of it.

Standing catalysts: NVIDIA Kyber/800V timeline news (either direction); China SiC pricing/insourcing datapoints; monthly auto production/EV incentive news.

Thesis tripwires (unchanged in spirit from 7/3, now pre-registered as flip/kill conditions in §12): two consecutive sequential revenue declines; GM back below ~35%; a third consecutive Q1-style charge; AI-DC growth stalling; a decisive close below the 200-DMA (~$73) on fundamentals.

11. Key risks

• Cyclicality + beta 2.01 (dominant): revenue fell 28% peak-to-trough and the stock traded $44.90–$133.93 within one year. If auto/industrial demand takes a second leg down, the $53 bear is the path, and technical floors will not hold it.
• China SiC / trailing-edge pricing (structural, new emphasis): KB voice (Feb 2026) — Chinese suppliers matching/undercutting on trailing edge; BYD-style SiC insourcing attacks the EliteSiC margin structure directly. This is the risk that outlives the cycle.
• The recurring Q1 charges: two consecutive years of large "one-time" operating charges (−$574M, then −$53M op income) that our dataset cannot fully itemize. Mostly non-cash so far (FCF stayed positive), but a pattern that erodes GAAP credibility until it stops.
• 800V timing/slip: the option is timed to Kyber 2027; SemiAnalysis (via CNBC 7/6) reports a 2028 slip, NVIDIA denies. A confirmed slip delays the only unpriced upside leg. And no partner's per-rack content is quantified anywhere — the option cannot yet be underwritten, only owned cheaply.
• Estimate fragility at the far end: FY28E $5.70 rests on 6 analysts, peak-equal net margins, and a ~380M share count. The 15.5× headline multiple is only as good as that number.
• Valuation on EV metrics: 15.9× FY27E EV/EBITDA vs a 12–14× mid-cycle band — the P/E lens is the friendliest one; others are fuller.
• Buybacks above FCF into a cyclical recovery, and near-zero live expert coverage (one 14-month-old tagged claim) — the thesis rests on fundamentals and quant.

12. Verdict, position sizing & monitoring

Hold — unchanged from 2026-07-03, re-underwritten rather than re-anchored. The fresh evidence moved toward the bull: first YoY revenue growth, GM at 38.5% with a 38–40% guide, estimates flat while price fell 3.5%, RSI repaired without a rally, a validated (if stale) KB bottom-call, and a verified, consensus-unpriced NVIDIA 800V partner slot. We considered Buy — Tactical and declined it on three numbers: +8% to our $95 base is not tactical-buy compensation for a beta-2.01 name whose fundamental bear is −40%; the celebrated GM recovery is ~80% delivered at the quarterly level (20.3% → 38.5%, vs a realistic low-40s mid-cycle); and GAAP operating income was negative last quarter on charges our data cannot itemize. The editor's 2.5–3:1 asymmetry frame is right about the shape of this trade and wrong about the price — at $88 the risk/reward vs the fundamental bear is ~1.2:1; at $75–78 it is 2.3–2.8:1 (§3).

Entry guidance (explicit zones):

• Buy zone: $73–81. Bottom = rising 200-DMA ($73.3) ≈ 17× FY27E; top = 19× FY27E ($81.7), the level below which the prior dive said it "gets interesting." In this zone a 2–3% tactical position is warranted; base case +17–30%, bull +59–78%, bear −27–35%.
• At $88 (current): hold existing ~1–2%; new money limited to a starter (≤1%) at most, with adds reserved for the zone or the trigger below.
• Momentum alternative: a reclaim of the 50-DMA (~$109) on a clean, estimate-raising print — valid but pays up ~24%; we prefer the zone.

Pre-registered flip conditions (upgrade to Buy — Tactical): EITHER (1) price enters $73–81 with the 200-DMA still rising and no tripwire tripped; OR (2) the 2026-08-03 print clears all three bars: Q2 GM ≥39% with Q3 guided toward 40%+, a first explicit DC/AI revenue quantification (or firm Kyber-2027 confirmation), and a clean GAAP quarter (no new charges) with Q3 revenue guided ≥ ~$1.63B — in which case our base moves toward ~$105 (22× an FY27/28 blend) and the upgrade prints even at high-$80s.

Pre-registered kill conditions (downgrade to Avoid/Sell): two consecutive sequential revenue declines; GM back below ~35%; a third consecutive quarter-one-style charge or any new impairment; a decisive weekly close below the 200-DMA (~$73) accompanied by estimate cuts; or confirmed Kyber slip to 2028 combined with softening auto demand. Any of these points the stock toward the $53 bear and takes precedence over the buy zone.

• Monitoring: formal re-score at the 2026-08-03 print; §10 tripwires continuously. This verdict is logged as a tracked Synthos call as of 2026-07-20 at $88.06.
• Single biggest risk: the cyclical recovery stalls — with the structural China-SiC threat as the risk that would keep us out even at technically attractive prices.

Provenance & disclosures

• Traceability: 1 tagged KB claim, breadth 1 — forward_guidance, 2025-05-16, bullish/66 (cyclical-bottom call naming ON among Microchip/Infineon/Regal Rexnord), cited in §2 with its age (14 months) stated plainly. Thematic, untagged KB context (GaN/SiC materials 2026-06-18; China power-semi risk, Feb 2026) is labeled as such and not counted. The verdict is predominantly fundamentals- and quant-driven. Fabricated conviction is structurally impossible (claim-ID reconciliation).
• Data as-of: fundamentals 2026-04-03 (Q1'26) · estimates & prices 2026-07-20 (FMP pull, scripts/deepdive/v800_data/ON_data.json) · management guidance from the Q1'26 8-K (2026-05-04), carried from the 2026-07-03 dive. Forward figures are analyst consensus (FMP) or management guidance, labeled as estimates. 800V/NVIDIA partner facts and the Kyber timing dispute are external, verified via deep research 2026-07-19 and labeled inline; the SemiAnalysis 2028-slip report (via CNBC 7/6) is contested by NVIDIA.
• Data caveats: FMP's FY25 annual operating income does not reconcile with the quarterly sum (charge classification differs; we rely on quarterlies + cash flow); the Q1'26 ~$340M of operating charges cannot be itemized from this dataset and is flagged as an open question for the 2026-08-03 call; segment (FY25 AMG dropped) and geographic feeds remain incomplete/misleading; the supplied peer list mixes in non-semiconductor names; trailing P/E is trough-distorted. All flagged inline where used.
• Management caveat: onsemi's guidance is management's own book, half-weighted by design.
• Prior version note: this refresh re-underwrites the 2026-07-03 dive (Hold, FV ~$95 at $91.22). Verdict and FV are retained on fresh evidence, not carried forward by inertia; the changes (price, estimates, KB, 800V verification, technicals) are itemized in §6, §7 and §12.
• Not investment advice. Independent research, educational and informational only, never personalized. Hypothetical/forward figures are labeled; the only performance numbers Synthos will headline are the live, real-money Flagship's.
• Version: 2026-07-20. Prior versions available via the deep-dive version dropdown ("based on the info at the time").