ON · ON Semiconductor Corporation · Stock research

Current Thesis

Three weeks after the 2026-06-25 Synaptics announcement, the tape has answered the question the sell-side is still debating: ON has not based. The stock gapped roughly 20% to ~$92.89 on 6/26 and has continued to bleed, closing $87.37 on 2026-07-17 below the deal-gap floor, with market cap down to $34.0B from ~$46B in early June. The 800VDC power-content leg into NVIDIA racks is unchanged as a business fact, but the equity is now trading as a merger-arb short leg and an integration-risk story rather than an AI-power re-rate. Management has spent July reinforcing the cost-structure narrative (two fab divestitures on 7/7) instead of the growth one. With Q2 results due 2026-08-03 after the close, this is a name in an unresolved downtrend into a print, and price sits ~23% below the $113.52 consensus target a gap that reflects disbelief in the estimates, not obvious value.

Bullish and bearish views on ON Semiconductor Corporation

The model's bull view on ON Semiconductor Corporation (ON), in brief: The AI data-center leg reported on 2026-05-04 was up >30% QoQ, with management reiterating AI data-center revenue doubling YoY in 2026; the Q2 guide set revenue at $1,535–$1,635M and non-GAAP gross margin 38–40%. The bear view: Price is the cleanest evidence: $87.37 on 7/17 is below both the 6/26 gap close (~$92.89) and the level that would have marked the post-deal shelf holding. Both cases follow in full.

Bull Case

• The AI data-center leg reported on 2026-05-04 was up >30% QoQ, with management reiterating AI data-center revenue doubling YoY in 2026; the Q2 guide set revenue at $1,535–$1,635M and non-GAAP gross margin 38–40%.
• The NVIDIA 800VDC/MGX design-in (collaboration announced ~2026-05-29) is untouched by the merger power FETs, SiC JFETs, GaN and multi-phase controllers ship into PSUs, battery-backup units and power-distribution boards, and each 800V rack carries roughly 11x the ON silicon of a 54V rack.
• Susquehanna raised its target to $150 on 2026-07-09, the highest on the street, against a $87.37 quote; Evercore ISI held Outperform at $137 (6/26) and B. Riley moved to $135 (6/26) on the larger addressable market and $200M of targeted synergies.
• The 2026-07-07 Fab Right divestitures Tarlac, Philippines to Greatek Electronics (closing in 3–6 months) and Mountain Top, Pennsylvania to Silex Microsystems (closing January 2028) target ~$35M of annual savings, beginning in 2027 and fully realized in 2028, with a long-term supply agreement covering the transition.
• Forward P/E of 25.8 against trailing 61.8 implies consensus expects a sharp 2026–27 earnings recovery off the 2025 trough, when revenue fell 15.4% to $6.00B and net income collapsed 92% to $121M.

Bear Case

• Price is the cleanest evidence: $87.37 on 7/17 is below both the 6/26 gap close (~$92.89) and the level that would have marked the post-deal shelf holding. Three weeks of distribution with no reclaim attempt is a distribution pattern, and no buyer has defended the gap.
• Sell-side is walking targets down, not up. TD Cowen went to Hold from Buy on 6/26 ($115→$110), then cut again to $95 on 2026-07-13; Mizuho cut $150→$125 on 2026-07-01. The dispersion between $95 and $150 is the market's way of saying nobody can model the combined company yet.
• The all-stock structure puts dilution on existing holders now and delivers synergies only after a mid-2027 close, gated by a Synaptics shareholder vote, antitrust and foreign-direct-investment clearances, tax opinions and S-4 effectiveness ($235M termination fee payable by Synaptics in specified scenarios). That is a 12-month-plus arb overhang with an exchange ratio of 1.350 fixed, so every ON rally mechanically feeds arb supply.
• Roughly 60% of Synaptics revenue is consumer and wireless the exposure the AI-power buyer was explicitly paying not to own.
• The core remains soft: Q1 2026 GAAP operating margin was negative, with about half of revenue tied to auto and half to Asia-ex-Japan, so China EV softness still drags the blended result. The AI data-center piece is the smallest and fastest-growing slice, not yet large enough to carry the P&L.
• As a second-order AI derivative, the ceiling is set by NVDA/AVGO capex commentary. Broadcom's soft AI guide in early June read straight through to ON on 2026-06-05 (-11.05%), which is how a second-order name behaves when first-order guidance wobbles.

Setup & Price Structure

The June structure is gone. The 52-week range is $44.56–$134.92, and the 134.92 high printed 2026-06-03; six weeks later the quote sits 35% below it. The 2026-06-26 gap has neither filled nor held as support, and the drift from ~$93 to $87 through the first half of July happened on no fresh negative company news, which points to positioning unwind and arb-related supply rather than a headline event. There is no ACCELERATING setup here to buy the theme tag survives, the price structure does not. A constructive re-entry would require a higher low and a weekly reclaim of the gap zone in the low-to-mid $90s, ideally with the 800V peer group (VRT, NVTS, Infineon, NXP) leading rather than lagging. Buying weakness before that reclaim is the definition of averaging into a broken structure. The valuation argument 23% below consensus target is not a setup; consensus targets have been cut twice in three weeks and are still catching down to price.

Catalyst Calendar (next 30 days)

• 2026-08-03 (confirmed) Q2 2026 results (quarter ended 2026-07-03), released after the close, conference call 5:00 p.m. ET. The binary: whether AI data-center revenue held the guided double-YoY pace, whether auto/industrial stopped deteriorating, and whether gross margin landed inside the 38–40% guide. First management commentary on the Synaptics deal in a public Q&A format.
• Ongoing through Q3 2026 S-4 registration statement filing and effectiveness for the Synaptics merger; the proxy will disclose deal background and any competing-bid history.
• No dated FDA/regulatory or index event in the window. The 2026-08-03 print is the only hard date inside 30 days.

Elapsed catalysts

• ~Q4 2026 (est., 3–6 months from 2026-07-07) expected close of the Tarlac, Philippines fab sale to Greatek Electronics. _(passed 12d ago)_

What Would Change Our Mind

The bullish reversal case needs three things in sequence, and none has appeared. First, a weekly close back above the 6/26 gap zone in the low-to-mid $90s on expanding volume, establishing a higher low off the July base. Second, an 2026-08-03 print showing AI data-center revenue still tracking to double YoY and auto/industrial revenue flat-to-up sequentially growth without the legacy drag getting worse. Third, target revisions turning upward: the tell would be TD Cowen or Mizuho reversing their July cuts, since they are the marginal sellers of the story. Absent that, the constructive read stays theoretical. Conversely, a Q2 miss or a 2026 AI data-center guide-down would remove the only leg supporting a 25.8x forward multiple, and the downside reference becomes the $44.56 low end of the 52-week range being materially closer than the $134.92 high.

Correlation Notes

ON trades as a high-beta derivative of AI-capex sentiment with an auto/industrial anchor, which means it participates less on the way up than first-order names and more on the way down. Cluster tells: Vertiv (VRT) and Navitas (NVTS) are the cleaner 800V power reads and break first in both directions; Infineon and NXP carry the same auto/industrial exposure and confirm whether weakness is ON-specific or sector-wide. Since 6/26 the divergence has been ON-specific the SOX and the power-pack have not tracked the drawdown, which isolates the Synaptics deal as the cause. SYNA's discount to the implied deal value is a live gauge of how the market prices closing risk, and a widening spread signals rising doubt about the mid-2027 close.