3 AI-Related Stocks Worth a Closer Look After Pullbacks

AI-linked names (Vertiv VRT, Applied Optoelectronics AAOI, Innodata INOD) have pulled back sharply from 2026 highs despite intact growth narratives, creating “opportunity” setups. Vertiv reported sales up 24% YoY to $3.27B and adjusted EPS up 60%, while raising 2026 organic sales growth guidance to 30%-32%; Applied Optoelectronics revenue jumped 86% YoY to $191.9M with data center revenue up 140%+ and 800G volume more than doubling; Innodata posted record quarterly revenue of $92.1M (+58% YoY) with adjusted EBITDA up 92% and maintained at least 40% 2026 revenue growth.

Analysis

This is mostly a positioning reset, not a thesis break. In AI infrastructure, the market is increasingly punishing crowded longs first and asking questions later, but VRT is still the cleanest way to own the physical bottleneck: power, thermal management, and liquid cooling are the gating items before incremental GPU deployments can translate into usable capacity. That should support a higher quality multiple than chip-adjacent or purely narrative AI names, especially if hyperscaler capex remains sequentially firm over the next 1-2 quarters.

AAOI is the highest-beta expression and the least forgiving. Even with visible demand, optics stocks tend to trade on gross margin durability, inventory turns, and working-capital intensity; if 800G/1.6T pricing softens or customers pull forward less than expected, the market can re-rate the stock faster than the revenue line decelerates. Competitors and adjacent suppliers like COHR, LITE, and Ciena can absorb some of the growth if procurement shifts or qualification timing slips, so this is better treated as a tactical trade than a core compounder.

INOD is different: the key question is whether AI services remain scarce enough to sustain premium growth, or whether large customers internalize more of the workflow and force pricing down. The consensus may be underestimating how quickly labor-scaled services can reprice once model builders optimize in-house, so the upside is real but the risk is multiple compression around each print. What would falsify the bullish setup across the group is a cut in hyperscaler capex, evidence of margin erosion, or any guide-down that implies the 2026 growth rates are peak-velocity rather than a durable run-rate.

Key Decisions for Investors

• Go long VRT on continued weakness, with a 3-6 month horizon; best risk/reward in the basket because the bottleneck is physical capacity, not sentiment. Reduce/exit if next guide implies organic growth slips materially below the current 30%+ run-rate or if hyperscaler capex commentary turns cautious.
• Use a long VRT / short SMCI pair to isolate the infrastructure beneficiary from the more valuation-sensitive AI hardware trade. This should work over 1-3 months if the market keeps rewarding earnings visibility and punishing narrative multiple risk.
• Trade AAOI only tactically: buy on post-earnings pullbacks or use call spreads rather than cash equity, given the volatility and execution risk. Stop out if 800G volume growth slows materially or gross margin compresses for two consecutive quarters.
• Maintain only a smaller starter long in INOD, or wait for the next print before sizing up. The setup improves if customer diversification keeps broadening; it breaks if revenue growth falls below 30% or operating leverage stalls.
• Set an alert basket on AI infra capex commentary: if hyperscaler spending or backlog indicators roll over, cut exposure across VRT/AAOI/INOD quickly because the next leg lower would likely be multiple compression before fundamentals visibly weaken.