ON — ON Semiconductor Corporation

Sentiment is genuinely mixed, and the price action says so: as of 2026-10-04, ON was up 9.96% in one week and 17.35% in one month but still down 6.94% over three months, which is the signature of a sharp bounce rather than a sustained re-rating. The September 16, 2026 Analyst Day ($11B revenue by 2030, high-voltage 800V AI data-center power) sold off because the payoff is weighted to 2028+, and the early-October amendment of the Synaptics deal from $7B all-stock to $5.7B all-cash removed dilution but committed debt. Analyst positioning is split nearly down the middle - 24 Buy vs 22 Hold and 1 Sell - with a $99.50 consensus target but a more recent $91.27 average, and the current price sits below both.

The three-year arc is a textbook cycle, not a structural decline. Revenue peaked at $2.181B in Q3 2023 and fell every quarter to $1.446B in Q1 2025 - a 34% drawdown - then ground back to $1.603B in Q2 2026, the strongest quarter in seven and +9.2% YoY, with the prior quarter at +4.7%. Gross margin compressed from 47.3% (Q3 2023) to a 34.9% trough in Q4 2025 (the 20.3% print in Q1 2025 was charge-distorted), and operating margin fell from 31.5% to 16.1%. Critically, two quarters carry outsized one-offs: Q1 2025 posted operating income of -$573.7M and EPS of -$1.15, and Q1 2026 posted operating income of -$53.4M and EPS of -$0.08 - in the latter case gross margin was a healthy 38.5%, meaning the loss came from below the gross-profit line (impairment/restructuring), not from the core franchise. Excluding those, operating margin has recovered from 13.2% (Q3 2025) to 16.1% (Q2 2026). FCF is running at $1.32B TTM, flattered by capex collapsing to $21.9M and $34.3M in the last two quarters versus $225.8M a year earlier.

As of 2026-10-04, ON's 53.7x TTM P/E and 27.9x TTM EV/EBITDA are artifacts of depressed trailing earnings (TTM EPS of $1.615 includes two charge-hit quarters), so they should be discarded. The meaningful number is the 26.5x the $84.89 price pays on the $3.198 next-FY consensus EPS - which itself assumes roughly an 80% earnings jump as margins normalize - against a PEG of 1.20. Relative to its own history, 5.33x sales versus a 5-year average of 3.88x is a genuine 37% premium, and the 4.59x book is well above the trough-era norm. Verdict: rich on trailing metrics, roughly fair-to-slightly-rich on forward earnings, and requiring the recovery to actually land.

The market is already paying for a recovery that has only two quarters of revenue evidence behind it and zero quarters of normalized margin evidence: 26.5x next-FY EPS that itself embeds an ~80% earnings step-up. What is NOT in consensus numbers is the AI data-center power roadmap - content per rack scaling via 800V high-voltage sidecars - which management sized at an $11B revenue company by 2030. Read it as: the cyclical normalization is priced, the electrification/AI content story is a free option, and the September Analyst Day selloff plus the still-negative 3-month return (-6.94%) tell you the market refuses to pay for 2028 payoffs today. If Q3 (2026-11-02) shows revenue growth decelerating back below +5% YoY, the forward multiple compresses quickly; if it prints +10%+ with margin expansion, the stock has room toward the $99.50 consensus.

As of 2026-10-04, RSI(14) stood at 65.49 - firm but not yet overbought - with the price 15.05% above its 20-day SMA (~$73.8), 10.76% above its 50-day (~$76.7), and 5.51% above its 200-day (~$80.5). The shares have just reclaimed the 200-day, which is the constructive part; the 15% extension above the 20-day after a 6.0% single-session and 9.96% weekly surge is the caution flag. This is a recovering trend that is short-term stretched, not an oversold entry. Timing implication: chasing $84-85 is poor risk/reward; the $77-80 zone (50-day/200-day confluence) is where the setup improves, and a loss of the 200-day would invalidate the recovery thesis.

• Federal Reserve rate adjustments have improved liquidity but heightened volatility for high-multiple growth and semiconductor equities, and ON's ~2.0 beta means it amplifies both directions.
• A prolonged inventory digestion cycle in automotive and industrial end-markets - ON's two largest - is the single biggest swing factor in whether the current revenue inflection holds.
• Generative AI data-center infrastructure capex is a structural tailwind, with high-voltage 800V power architectures increasing ON's addressable content per rack.
• EV electrification adoption pacing directly drives the premium silicon-carbide franchise; any slowdown in EV unit growth hits mix and gross margin disproportionately.
• US-China trade policy, tariffs, and semiconductor export controls introduce supply-chain and end-demand risk given meaningful China automotive and industrial exposure.

• Closing and integration of the amended $5.7B all-cash Synaptics acquisition (announced early October 2026) - removes share dilution but consumes essentially all of ON's $3.51B cash and adds debt.
• Q3 2026 earnings on 2026-11-02: the first real test of whether the Q2 revenue inflection (+9.2% YoY) and margin recovery extend.
• Execution against the September 16, 2026 Analyst Day 2030 roadmap ($11B revenue), specifically design wins and customer adoption for high-voltage AI data-center power sidecars.
• Sustained capex discipline - capex fell to $21.9M and $34.3M in the last two quarters versus $225.8M a year prior; if that is structural rather than timing, FCF stays elevated.
• Any broad automotive/industrial restocking signal that converts the two-quarter revenue upturn into a durable cycle.

• The recovery is only two quarters old and sits on top of a 34% revenue drawdown; if auto/industrial digestion resumes, the 26.5x forward multiple has no earnings cushion.
• Capital allocation risk: $5.7B in cash (about 17% of market cap) for Synaptics at roughly 5x sales, funded with debt and consuming the entire cash balance, materially changes the balance sheet that currently supports the strong financial-health score.
• The 2030 targets pay off mostly in 2028 and beyond, as the post-Analyst-Day selloff demonstrated; the market has shown it will not fund that timeline at a premium.
• Trailing quality is genuinely poor right now - TTM ROIC of 0.7%, ROE of 1.6%, operating margin of 11.7% - and there is no way to fully separate 'cyclically depressed' from 'structurally out-competed' until margins recover.
• Valuation: 5.33x sales versus a 5-year average of 3.88x, and 27.9x TTM EV/EBITDA, leaves limited room for disappointment.
• Beta near 2.0 means any reversal in rate expectations or AI/tech sentiment produces an outsized drawdown.

This is a reasonable but not compelling risk/reward at $84.89 on 2026-10-04. The good news is real: revenue has inflected (+4.7% then +9.2% YoY), gross margin is back near 38-39%, operating margin has recovered to 16.1%, FCF is $1.32B TTM, and net debt was just ~$0.14B as of the July 3, 2026 balance sheet - plus a genuine, unpriced AI data-center power option. The problem is that the stock already pays 26.5x a next-FY EPS estimate that assumes an ~80% earnings jump, just rallied 17.4% in a month, and is about to take on $5.7B of debt for Synaptics while the 2030 roadmap pays off after 2028. I would want to own this on a pullback toward $77-80 rather than chase it here.

Blended three ways using the FMP data, then adjusted for optionality. (1) 24x forward EPS - a mid-cycle multiple for analog/power semis in a recovery year, below the 5-year average P/E of 52.6 but above the 15-19x the stock traded on peak 2023-24 earnings - applied to the $3.198 next-FY consensus EPS = $77. (2) 15x FY2027E EBITDA of ~$2.0B (roughly $7.0B revenue at a 22% operating margin plus ~$0.55B D&A) less ~$0.2B net debt = ~$77 per share. (3) A 4.25% FCF yield on ~$1.35B of normalized FCF (TTM FCF of $1.32B, with capex marked back up toward 6-8% of revenue from the current unsustainably low 2-3%) = ~$82. The ~$79 mean, plus a ~10% premium for the AI data-center / 800V power content opportunity that consensus numbers do not yet carry, nets to $87.