ON · ON Semiconductor Corporation · Stock research

Current Thesis

The mid-$80s shelf went the way of the ones before it. ON closed $74.21 on 2026-08-21 against $82.66 on 2026-08-14 — roughly 10% lower over five sessions, 44.6% below the $133.93 52-week high, with RSI(14) at 38.2, weak without being washed out. The 2026-08-03 beat-and-raise is now three weeks old and has bought no price. The leg an investor would be buying — 800VDC rack content re-rating a legacy automotive/industrial franchise — remains intact as an operating fact and is broken as a price structure. Those are two different statements and the tape is only grading the second one.

Bullish and bearish views on ON Semiconductor Corporation

The model's bull view on ON Semiconductor Corporation (ON), in brief: Q2 2026 (2026-08-03): revenue $1,604M, +9% YoY; GAAP diluted EPS $0.56, non-GAAP $0.74; GAAP gross margin 38.4%, non-GAAP 39.3% — inside the 38–40% non-GAAP band guided on 2026-05-04. The bear view: Maximum good news produced a lower low, then another one. Both cases follow in full.

Bull Case

• Q2 2026 (2026-08-03): revenue $1,604M, +9% YoY; GAAP diluted EPS $0.56, non-GAAP $0.74; GAAP gross margin 38.4%, non-GAAP 39.3% — inside the 38–40% non-GAAP band guided on 2026-05-04.
• The "other" line rose 34% sequentially to $400M in Q2, and management lifted the 2026 AI data-center outlook from "double" to "more than double" versus 2025 on the 2026-08-03 call.
• Q3 guidance cleared consensus on both lines: revenue $1.650B–$1.750B against $1.666B consensus; non-GAAP EPS $0.81–$0.93 against $0.83.
• Cash conversion inflected in Q2: net income $226.8M versus $170.3M a year earlier, free cash flow $425.4M, operating cash flow +150% YoY.
• Regulatory risk came off the deal early: the FTC granted HSR early termination on 2026-08-12, about four weeks after the 2026-07-17 filings. The S-4 followed on 2026-08-21, which is the next procedural step toward a vote date.
• Synaptics' CFO resigned on 2026-08-21 with the company declining to run a successor search ahead of closing — behaviour consistent with a management team that expects the merger to complete, though it is an inference from a staffing decision, not a disclosure about deal odds.
• Published targets still sit above the tape: Cantor Fitzgerald reiterated Neutral with a $110 target on 2026-08-17; B. Riley $127 (Buy) and Needham $116 (Buy) from 2026-08-04 have not been withdrawn.

Bear Case

• Maximum good news produced a lower low, then another one. The +7% premarket move on 2026-08-04 was surrendered inside eight sessions; the $82.66 close on 2026-08-14 undercut the pre-print $87.37 of 2026-07-17 and the 2026-06-26 deal-gap area near $92.89; $74.21 on 2026-08-21 undercuts all three.
• The 1.350 ratio is fixed, so that share count is the supply arbitrage works against until close.
• The deal's own arithmetic documents the de-rate: the S-4 shows each Synaptics share implying about $156.25 at ON's $115.74 close on 2026-06-24, and about $109.05 by 2026-08-04.
• The Q3 guide's edge over consensus is roughly two percent at the revenue midpoint and about four cents at the non-GAAP EPS midpoint.
• Allocation priority to AI comes out of automotive, per management's own framing on 2026-08-03, in a segment that is roughly half of revenue. No offsetting automotive design-win disclosure has accompanied it.
• Mid-2027 remains the guided close. The FTC clearance shortened the risk list without shortening the calendar.

Setup & Price Structure

The narrative is dead — the price narrative, not the product line. The label is dated by three failures in sequence: the 2026-06-26 deal gap lost, the 2026-07-17 pre-print level lost, and the 2026-08-03 beat-and-raise fully retraced by 2026-08-14 and then extended lower to $74.21 on 2026-08-21. A narrative that cannot hold a bid on its best fundamental print of the year is not consolidating.

Positioning and crowding observables, stated as observables: price is 44.6% under the 52-week high and 36.1% lower over three months, so any moving average an investor would reference is above the tape and falling — the opposite of the extended-above-a-rising-average configuration. RSI(14) at 38.2 on 2026-08-21 is not a capitulation reading; the November-quarter print is roughly ten weeks out (~2026-11-02, estimated), so there is no imminent earnings catalyst compressing the move. The gap between published targets and the tape is the crowding datapoint that matters: Cantor's $110 reiteration on 2026-08-17 and the 2026-08-04 cluster at $108–$127 all sit far above a $74.21 close, and the August experience was targets converging down toward price rather than price rising toward targets. Separately, the S-4 registration on 2026-08-21 is share issuance into weakness, which quantifies the borrow arbitrage can build against.

What a base would look like: a weekly range that holds the mid-$70s, followed by a weekly close reclaiming the $82.66 area with the S-4 effective and a Synaptics meeting date printed. None of those conditions exists as of 2026-08-21.

Catalyst Calendar (next 30 days)

• 2026-08-26 — NVIDIA fiscal Q2 results, after the close. Management guided the quarter to approximately $91.0B in revenue; published previews put sell-side consensus near $91.85B. ON's power content sits downstream of rack capex, and the precedent is direct: Broadcom's soft AI guide contributed to an 11.05% single-day decline in ON on 2026-06-05.
• ~2026-11-02 (est.), beyond the 30-day window — Q3 2026 results: the first test of the $1.650B–$1.750B guide, of whether the AI data-center line repeats anything like the +34% sequential pace, and of what automotive did after AI got allocation priority.
• Mid-2027 (company guidance) — targeted merger close, which removes the fixed-ratio overhang and starts the clock on the $200M synergy target.

Elapsed catalysts

• TBD, following the 2026-08-21 S-4 filing — amendment and effectiveness of the registration statement, then the Synaptics special-meeting date. The meeting date was still a placeholder in the 2026-08-21 document. Until it exists, the arb has no resolution date to trade against. (passed 5d ago)

What Would Change Our Mind

The structure now on the line is the mid-$70s zone where the decline paused on 2026-08-20/21. A weekly close below $72 confirms the August leg extending with no base formed, and puts the 2026-08-03 print definitively in the category of news the market refused to pay for. A second condition sits three sessions out: if the 2026-08-26 NVIDIA print passes and ON has not reclaimed the $82.66 close of 2026-08-14 in its wake, the second-order AI bid has been tested and did not appear.

The constructive reversal is equally specific and requires two things together, not one. First, a weekly close back above $87.37 — the 2026-07-17 level that preceded the print — which would mean the post-print supply had been absorbed. Second, the S-4 going effective with a dated Synaptics special meeting, converting an open-ended arb overhang into a countdown. Either alone is weaker evidence than the pair.

On the fundamental side, the datapoint that would reopen the re-rate case is Q3 total revenue landing above the $1.750B guide ceiling with the AI data-center line still compounding sequentially — mix gains finally moving the blended number rather than being absorbed by flat automotive.

Correlation Notes

• ON trades as a high-beta expression of AI rack capex through a legacy analog/discrete P&L. The 2026-06-05 session — an 11.05% decline into the worst SOX day since March 2020, on Broadcom's soft AI guide alongside a 172k payroll print — is the cleanest read on that beta.
• SYNA is now an ON derivative by construction: the fixed 1.350 ratio means Synaptics equity tracks ON's tape, and the spread between SYNA and 1.350× ON is the observable for whether arb positions are being built or unwound.
• Roughly half the revenue base is automotive, so ON carries a second correlation to the European auto-semi complex and to EV/hybrid build rates, which is why a strong AI data-center number can be neutralised at the total-revenue line.
• Approximately 60% of Synaptics revenue is consumer and wireless, so the pro-forma company is a less pure AI-power expression than the standalone story implies.