2 Hidden Industrial Stocks That Will Benefit From the AI Infrastructure Build-Out Not Named SpaceX or Intel

The article argues that ON Semiconductor and Rockwell Automation are underappreciated AI beneficiaries, with ON's AI data center revenue about $250 million in 2025, or 4.2% of sales, expected to double to $500 million in 2026 and rise to 7.7% of revenue. Rockwell is positioned to gain from AI inference in industrial automation through Nvidia-integrated digital twin software, but the piece is largely a long-term bullish commentary rather than a new company-specific catalyst.

Analysis

The market is still pricing AI as a hardware supercycle, but the cleaner second-order trade is the monetization layer that comes after build-out: power delivery and industrial inference. That favors ON and ROK because both are levered to the “everywhere AI” phase, where workloads migrate from centralized training clusters to distributed execution, creating demand for power electronics, controls, and software that sit closer to the edge. The important implication is that revenue quality should improve faster than headline growth suggests, because inference tends to be stickier and more recurring than capex-driven infrastructure orders.

ON has the more asymmetric near-term setup because its AI contribution is still small enough that incremental wins can change the mix meaningfully, yet large enough to matter to estimates over the next 4-6 quarters. The second-order effect is that every new AI data center design decision can expand ON’s content per rack, while adjacent markets like energy storage and microgrids can piggyback on the same power-management stack. That creates a path for margin expansion even if non-AI end markets only recover modestly, making consensus upside more likely to come from mix than from unit volume.

ROK is a longer-duration compounder, but also less “headline obvious,” which is why the stock can lag until the market fully prices software-plus-inference value creation. The real option here is not just automation penetration; it is the ability to make Rockwell’s installed base more valuable by embedding AI into workflows, which should lift wallet share and service attach rates over time. If digital twin adoption accelerates, Rockwell’s economics could shift from cyclical industrial hardware multiples toward higher-quality recurring software-like monetization.

The main risk is timing: AI infrastructure spending can decelerate before inference revenue fully offsets it, creating a 2-4 quarter air pocket for suppliers perceived as AI-exposed. For ON, the key watch item is whether AI revenue growth outpaces any normalization in auto/industrial, otherwise the re-rating stalls; for ROK, the risk is that customers trial AI tools without broad deployment, delaying monetization. Consensus may be underestimating how much of the value accrues to enabling infrastructure rather than model owners, but it may also be overestimating the speed of that shift.

Key Decisions for Investors

• Go long ON on any 5-8% pullback; target a 6-12 month hold where AI mix expansion can drive estimate revisions and multiple support, with downside limited by diversified end markets.
• Initiate a pair trade: long ON / short a crowded AI infrastructure beneficiary basket; thesis is rotation from build-out to inference over the next 2-3 quarters, capturing relative multiple compression in the more consensus names.
• Build a starter long in ROK only on weakness, framed as a 12-24 month software re-rating story; upside comes from digital twin and AI attach, but patience is required before revenue inflection shows up.
• Use call spreads in ON rather than outright equity if looking for a catalyst trade into next earnings; the setup is estimate-upside driven, but valuation sensitivity remains tied to auto/industrial macro.
• Avoid chasing near-term outperformance in the obvious AI infrastructure winners; relative alpha should shift toward less obvious picks like ON/ROK as inference becomes the dominant incremental spend.