2 Top AI Stocks To Put On Your Shortlist

The article argues that the AI capex cycle is still expanding, citing Meta's raised 2026 capex estimate to $145B and more than $100B expected from SpaceX's Terafab initiative. It highlights ACM Research as a beneficiary of semiconductor fab expansion, with management signaling 20-30% 2026 growth, and Ouster as a lidar/autonomy play with improving L3 chip performance and recurring software revenue. Overall, the piece is bullish on under-the-radar AI infrastructure names rather than broad-market mega-caps.

Analysis

The setup is less about “AI winners” broadly and more about a narrowing funnel of capex winners with pricing power, long lead times, and visibility into orders two to six quarters out. ACMR looks like a leverage play on an equipment bottleneck that should persist as fabs keep extending buildouts; the market usually underestimates how long suppliers can ride a transition from pilot spend to full-scale line utilization. The key second-order effect is that as hyperscalers and frontier-model builders keep pulling for compute, foundry/tooling budgets become less discretionary and more infrastructure-like, which supports multiple expansion for niche suppliers with installed-base service and higher-value process steps.

The bigger competitive dynamic is that this is likely to pressure the incumbents less than expected and instead force them to defend share with faster product cycles and heavier customer support spend. That can compress gross margin at the large toolmakers if they need to bundle more services or price aggressively to hold sockets, while smaller differentiated suppliers can see operating leverage if they are single-source or qualification-constrained. TSM remains a structural beneficiary of the capex wave, but the more important read-through is that supply chain bottlenecks may keep shifting upstream, which extends cycle duration rather than ending it.

OUST is a different trade: it is a pick-and-shovel beneficiary of autonomy adoption, but the market is likely to overestimate the speed of revenue conversion from “AI at the edge.” Industrial and robotics deployments are capex-light compared with semis, so the revenue ramp should be slower and more back-end weighted, with the software attach rate doing most of the margin work. The risk is that lidar remains a feature, not a moat, if OEMs standardize on lower-cost sensors or if autonomy budgets get pushed out in a softer macro.

The contrarian view is that sentiment is already crowded around AI infrastructure, but still under-allocated to the enablers with the longest project duration. That favors buying dips in the more rate-sensitive, underfollowed names rather than chasing mega-cap AI beta. The highest-risk mistake here is confusing a multi-year capex cycle with a straight-line earnings path; volatility will likely come from quarterly ordering lulls, not from thesis failure.

Key Decisions for Investors

• Initiate a starter long ACMR position on any post-earnings pullback, with a 6-12 month horizon; target a 20-30% re-rating if order visibility improves, but cap downside with a 15% stop because this is a cyclical equipment name.
• Pair trade: long ACMR / short a basket of higher-quality but more crowded semi-capex leaders (e.g., AMAT/LRCX) for 3-6 months; thesis is greater multiple expansion in the underowned name if bottleneck scarcity persists.
• For OUST, prefer a call spread rather than outright equity: buy 6-12 month upside exposure and finance with a higher strike sale; this captures the adoption upside while limiting drawdown if autonomy budgets slip.
• Add TSM on weakness as the cleanest way to express extended AI capex duration; hold 6-18 months and use a trailing stop rather than fixed profit target, since the main risk is sentiment compression rather than fundamental deterioration.
• Avoid chasing META on the headline capex print; if anything, use it as a liquidity event to rotate into suppliers with lower consensus ownership and clearer operating leverage.