Skip to content

Dossier · ON · Dormant

ON · ON Semiconductor Corporation

Last analysed ·

Current thesis

Company narrative intact AI data-center power re-rate (NVIDIA 800VDC/MGX content step-up, Q2 guide: AI revenue to double YoY in 2026) but the tape broke: -11.05% to $117.26 on 2026-06-05 in the worst SOX session since March 2020 (Broadcom soft AI guide + dead rate-cut odds). ACCELERATING narrative, tightening regime; buyable only if the ~$113 20-week EMA holds.

Invalidation trigger

Weekly close below the rising 20-week EMA (~$113); or Q2 print (~late July) shows AI data-center revenue not tracking the guided "double YoY in 2026"; or NVIDIA awards the 800V MGX power sockets to Infineon/Navitas over onsemi.

Thesis status

Open commitment scored if the trigger above fires How this is scored →

Current Thesis

The dead EV-SiC name re-rated into an AI-datacenter-power story, and that company-specific narrative is still intact but as of 2026-06-05 the tape carrying it broke. ON closed $117.26, -11.05% on the day, inside the worst Philadelphia Semiconductor Index session since March 2020 ($1T+ of market value erased). The trigger was first-order: Broadcom's soft fiscal-Q3 AI guide (AI chip sales $16B vs $17.2B est; AVGO -13/14% on 6/04), compounded by a hot May jobs print (172k vs ~80k est) that killed near-term rate-cut odds (10yr >4.5%) and Hormuz tension (Brent >$84). The narrative leg worth owning is the 800VDC rack transition onsemi's silicon content per NVIDIA rack steps up sharply as racks move from 54V toward 800V/1MW+, and management has guided AI data-center revenue to double YoY in 2026. That leg is ACCELERATING; the regime around it is now actively tightening.

Bull Case

  • Q1 2026 print (2026-05-04) was the inflection: revenue $1,513M beat the guide midpoint; AI data-center revenue +>30% QoQ; non-GAAP gross margin 38.5%; non-GAAP EPS $0.64; non-GAAP operating margin 19.1%.
  • Q2 2026 guide (issued 5/04) extends the ramp: revenue $1,535–$1,635M, GAAP EPS $0.60–$0.72 / non-GAAP $0.65–$0.77, gross margin 38–40%, and management reiterated AI data-center revenue doubling YoY in 2026.
  • NVIDIA design-in is real and widening: onsemi formally collaborated on 800VDC power (5/29) and expanded its MGX ecosystem role (~6/01) across power supply units, battery backup units and power-distribution boards power FETs, multi-phase controllers, SiC JFETs and GaN already shipping in MGX.
  • Content-per-rack is the mispriced variable: the 2026-04-20 "Wolf Pick" framed each 800V rack NVIDIA ships in 2027 as ~11x the onsemi silicon of today's 54V racks, with consensus not yet modeling it.
  • High-end sell-side still sits above market: Mizuho Outperform PT $150 (5/28), BofA Buy PT $138 (5/26) both above the 6/05 close of $117.26.
  • Power-basket cluster confirmation: Navitas +346% YTD (6/04) and itself joined the NVIDIA MGX ecosystem (6/03); new leveraged ETFs built around ON/NXP/Microchip launched 6/01–6/03, adding structural flow.

Bear Case

  • The tape regime flipped on 6/05 (-11.05%) inside the worst SOX day since March 2020. Broadcom's soft AI guide is a first-order AI-capex crack, and ON is a second-order derivative of that same spend more exposed to AI-capex doubt, not less.
  • Consensus already prices downside from here: 12-month average analyst PT ~$104, below the $117.26 close; ratings split 12 buy / 12 hold / 0 sell (Moderate Buy). Barclays is Equal Weight with a PT of $100 (raised from $75 in May) a hold rating that puts fair value under the current quote.
  • Still richly valued into a derating: P/E ~83, market cap ~$46B, after a 52-week high of $134.92 (low $44.56). A multiple that rich needs an expanding tape, and the tape just contracted.
  • Core remains in a trough: Q1 GAAP operating margin (3.5)% and GAAP diluted loss ($0.08). Roughly 50% auto / 50% Asia-ex-Japan exposure (Barclays' standing overhang) means China-EV softness still drags the blended story; AI data-center is fast-growing but a small slice of revenue.
  • Crowded socket: Infineon joined MGX within two days of onsemi (late May / early June) and Navitas followed (6/03) 800V power-FET share is contested, not locked, with commoditization risk on the power layer.
  • No near-term fuel: the next hard catalyst (Q2 print) is ~7 weeks out, leaving nothing scheduled in 30 days to defend the multiple while rates, oil and geopolitics drive sector beta.

Setup & Price Structure

  • Last $117.26 (2026-06-05 close), -11.05% on the day; after-hours $115.30. That is ~12% below the 6/03 area (~$133) and ~13% off the 52-week high of $134.92.
  • The plunge dragged price toward the rising 20-week EMA (est. ~$113), which is now the decision level. A weekly hold above it keeps the breakout structure alive and frames this as the macro-driven pullback an accelerating name rarely offers; a weekly close below it says the AI-power re-rate is unwinding with the sector.
  • Price now sits below both high-end PTs ($138/$150) and below the $104 consensus average, so sell-side is split around the tape rather than uniformly chasing the "below fresh PTs = headroom" read from the prior week no longer holds cleanly.
  • RSI has reset hard off the 6/05 flush; this is no longer a stretched-above-MA, peak-momentum chart. It is a violent mean-reversion in progress with the 20-week EMA as the line in the sand.
  • Regime read: macro is actively tightening (rate-cut odds gone, 10yr >4.5%, Brent >$84). That is a hard timing blocker strength was the setup three sessions ago; the question now is whether support holds.

Catalyst Calendar (next 30 days)

  • No company-specific hard catalyst inside 30 days. Q2 2026 print is the next binary, est. ~2026-07-28 (Q1 was 5/04) confirm the exact date mid-July before sizing into it. No earnings blackout right now.
  • Macro overhang: the next CPI print and Fed commentary, plus any Hormuz/Brent escalation, drive sector beta near-term. With ~50% Asia-ex-Japan revenue, a China-EV data point or tariff headline moves ON independent of the AI story.
  • Peer prints/guides in the 800V basket (Infineon, Navitas, NXP, Microchip, Vertiv) any soft AI-power commentary reads straight through to ON ahead of its own print.

What Would Change Our Mind

  • Bull confirmation: a weekly close holding above the ~$113 20-week EMA with a higher low forming over the next 1–2 weeks → the macro flush rather than the narrative drove it, and the re-rate resumes toward $138/$150.
  • Bear confirmation / thesis break: a weekly close below the 20-week EMA (~$113); or the Q2 print (~late July) showing AI data-center revenue not tracking the guided "double YoY in 2026"; or NVIDIA steering the 800V MGX power sockets to Infineon/Navitas over onsemi.
  • Macro all-clear: rate-cut odds rebuilding (10yr back below ~4.3%) and the SOX reclaiming its pre-6/04 level would remove the regime blocker and turn a held-support chart back into an add candidate.

Correlation Notes

  • First-order tether: ON trades as a second-order NVIDIA/AI-capex derivative AVGO and NVDA guidance set the ceiling. Broadcom's 6/03 soft AI guide (-13/14%) was the proximate trigger for ON's -11% on 6/05.
  • Power-basket cluster: Navitas (GaN pure-play, +346% YTD), Infineon, Vertiv, NXP, Microchip pure-plays break first in both directions; ON follows the 800V pack.
  • Macro factors: rate-sensitive (10yr >4.5% on the 6/05 jobs shock), oil/geopolitics (Brent >$84 on Hormuz), and ~50% Asia-ex-Japan revenue make China-EV and tariff headlines independent drivers.
  • ETF flow: the new leveraged ON/NXP/Microchip ETFs (6/01–6/03) add two-way amplification they magnified the 6/05 downside and would magnify any bounce.

Notes

  • Q2 2026 print est. ~2026-07-28 (Q1 was 5/4) OUTSIDE 30d window as of 6/4; confirm exact date mid-July before sizing into the print (no earnings blackout right now = clean entry).
  • Stock at ATH ~$133 (6/3), +184% YoY, P/E ~85 extension is confirmation per momentum playbook, NOT a fade; do NOT wait for a 'cheap' pullback an accelerating name won't give, and do not anchor to a re-entry level.
  • Two stories in one ticker: AI data-center (small, +30% QoQ, the bull leg) vs legacy auto/industrial majority still at GAAP op loss in Q1 blended thesis depends on legacy not deteriorating further.
  • Cluster tells to watch: Navitas (NVTS, +346% YTD), VRT, Infineon, NXP onsemi follows the 800V-power pack; pure-plays break first in both directions.
  • Price ~$132.95 sits BELOW fresh PTs (BofA $138 / Mizuho $150) treat a move above $150 with no PT follow-through as a saturation watch.
  • Q2 2026 print est. ~2026-07-28 (Q1 was 5/04) confirm exact date mid-July; no earnings blackout in the next 30d.
  • 2026-06-05: -11.05% to $117.26 inside the worst SOX day since March 2020 (Broadcom soft AI guide, 172k jobs print killed rate cuts, Hormuz/Brent >$84). Macro/sector-driven, not an ON-specific narrative break but it flips the regime from 'extension = confirmation' to 'does the 20-week EMA hold.'
  • 20-week EMA (~$113) is THE decision level post-crash: weekly hold = buyable macro pullback; weekly loss = re-rate unwinding with the sector. Do not anchor to the old ~$133 ATH.
  • Second-order AI play: ceiling set by NVDA/AVGO guidance. AVGO's soft AI guide read straight through more exposed to AI-capex doubt than first-order names, not less.
  • Sell-side now split: 12mo consensus PT ~$104 and Barclays Equal Weight PT $100 both BELOW the $117 close; only Mizuho ($150) and BofA ($138) sit above. P/E ~83, market cap ~$46B.
  • Two stories in one ticker: AI data-center (small, doubling) vs legacy auto/industrial (~50% auto, ~50% Asia-ex-Japan, still GAAP op loss in Q1). Blended thesis needs legacy not to deteriorate further.

Related · shared themes