ON Semiconductor's Bull Trap Over: Brilliant AI-Driven Portfolio Renewal

ON Semiconductor's diversification into data centers and exits from legacy businesses are expected to accelerate growth despite mixed automotive demand and softer industrial conditions. New product launches, customer wins and accretive M&A support a projected data-center opportunity of roughly $115,000 per rack by 2030. The outlook points to accelerating FQ3 2026 growth after relatively solid H1 2026 performance and trough-like FY2022-FY2025 metrics.

Analysis

ON is attempting to replace a cyclical automotive/industrial earnings profile with higher-growth power-semiconductor content in AI infrastructure. The relevant valuation question is not rack-level content potential, but how much of that addressable content ON can win versus Infineon, STMicroelectronics, Vishay, MOSFET specialists, and analog incumbents such as ADI and MCHP. Incremental data-center revenue should carry better factory utilization and gross-margin leverage than the company’s trough-cycle industrial mix, but the earnings impact will be immaterial until design wins convert into qualified production orders—typically a 12-24 month lag.

The near-term setup hinges on whether the guided acceleration reflects broad order recovery or customer inventory normalization. A guidance beat can expand ON’s multiple over the next 1-3 months if management identifies hyperscaler/ODM programs, power-device attach rates, and backlog conversion; vague commentary around pipeline value should be treated as promotional rather than evidence of revenue durability. Automotive remains the key offset: weaker EV volumes, silicon-carbide pricing pressure, or delayed OEM platform launches could absorb much of the data-center upside before it reaches consolidated margins.

The contrarian view is that AI-power enthusiasm may already be capitalized into semiconductor peers, while ON’s differentiated exposure is more dependent on power architecture transitions than on GPU shipment growth alone. Nvidia-led rack deployments benefit suppliers only where higher-voltage conversion, power distribution, and thermal constraints require new components rather than lower-cost substitutions. The thesis is falsified if FQ3 growth is driven by distributor restocking, gross margin fails to improve sequentially, or data-center design-win disclosures do not translate into a rising share of revenue within the next two earnings cycles.

Key Decisions for Investors

• Watch-list ON for a post-earnings long entry rather than chasing guidance: initiate only if management pairs revenue acceleration with sequential gross-margin expansion and quantifies production—not pipeline—data-center wins. Target a 6-12 month rerating versus MCHP/STM; exit if margin guidance declines or automotive/industrial inventory days rise.
• Use a 1-3 month pair trade only on confirmed data-center traction: long ON / short STM. ON has greater potential multiple upside from mix improvement, while STM retains more direct exposure to a slower EV and industrial recovery; size small because both remain exposed to cyclical analog/power demand.
• For a lower-beta expression, consider long SOXX versus ON until customer concentration, data-center revenue contribution, and acquisition economics are disclosed. The missing data prevents underwriting a standalone ON earnings step-up with confidence.
• Set an earnings alert on three metrics: data-center revenue or design-win conversion, company gross margin, and automotive book-to-bill. A sequential revenue beat without improvement in the latter two would favor fading an initial positive reaction rather than adding exposure.