Mizuho raises ON Semiconductor stock price target on AI server demand

Mizuho raised its price target on ON Semiconductor to $150 from $130 while keeping an Outperform rating, citing AI server ramps, 800-volt architecture adoption, and tighter supply in AI server MOSFETs. The firm also sees improving industrial demand and normalized inventories, offsetting some weakness in low-voltage power products. ON Semiconductor has surged 193% over the past year, though it still trades at 88.56x earnings and is viewed as overvalued by InvestingPro.

Analysis

The market is increasingly rewarding the power-semiconductor layer that sits closest to AI infrastructure bottlenecks, but the second-order implication is that valuation dispersion inside the analog complex should widen further. ON is being re-rated not just on growth, but on scarcity value in high-voltage and thermal-management content where lead times remain stretched; that creates a stronger relative case for suppliers with mix exposure to AI server power stages than for broader analog peers tied to slower industrial cycles. The key read-through is that AI capex is no longer only a compute story — it is becoming a power-density story, which should keep pricing power elevated for the few vendors with qualified architectures.

Near term, the biggest risk is that expectations have outrun the actual revenue inflection. With the stock already discounting a multi-year AI content ramp, any delay in 800V adoption, design-win conversion, or server OEM qualification could trigger a sharp multiple reset over the next 1-2 quarters even if fundamentals remain healthy. The setup also looks vulnerable to a broader factor unwind: if rates or inflation data pressure long-duration growth again, the most crowded winners in semis tend to de-rate first, and ON’s elevated multiple makes it especially sensitive to that air pocket.

The contrarian point is that the bullish case may be less about absolute demand and more about supply discipline and channel normalization, which is inherently more cyclical than the market is pricing. If industrial inventory destocking has truly ended, the next leg higher may be modest unless AI power demand broadens beyond a few hyperscaler programs; that argues for selectively owning the best-positioned names rather than the whole group. The market may also be underestimating substitution risk: as AI power architectures evolve, vendor concentration can shift quickly, so today’s winner can lose share if qualification cycles migrate to different topologies or if customers dual-source aggressively.

Bottom line: the trade is strongest if you express it as a relative winner versus slower-growth analog peers, not as an outright momentum bet in a crowded high-multiple stock. The highest-conviction payoff likely comes from pair construction or using options to define downside into the next catalyst window, rather than chasing spot after a strong multi-month run.

Key Decisions for Investors

• Long ON / short TXN as a 3-6 month relative-value pair: ON should retain the AI power optionality premium while TXN is more exposed to slower industrial normalization; target 10-15% spread capture, but keep sizing modest given ON’s valuation risk.
• Buy ON Jan-2026 call spreads rather than stock to express the AI server ramp thesis: upside participates if 800V content inflects, while premium paid caps damage if qualification or ordering slips over the next 1-2 quarters.
• Add ADI on weakness versus ON for a cleaner analog compounder exposure: ADI’s multiple is easier to defend if the market rotates from scarcity stories back to durable free-cash-flow quality.
• Use a tactical hedge against a factor reversal by shorting a basket of high-multiple semis against ON longs if inflation/rates re-accelerate; ON’s elevated P/E makes it one of the first names to de-rate in a growth selloff.
• If already long ON, trim into strength above prior consensus target levels and re-enter on post-catalyst pullbacks; the risk/reward is better after a 10-15% reset than chasing continued multiple expansion from here.