Artificial Intelligence (AI) Stocks Have Crushed the Broader Market in 2026. Here Are 2 Top Stocks to Buy Before They Soar Higher This Earnings Season and Beyond

AI infrastructure spending is projected to stay strong, with Alphabet/Amazon/Meta/Microsoft capex in 2026 forecast to jump 77% to a record $725B (vs. $500B prior expectation), supporting Nvidia and Celestica. Nvidia’s EPS is cited at $4.77 for fiscal 2026, with consensus implying ~50% EPS CAGR over the next three years and a potential 129% upside to ~$484 by fiscal 2029 (30x earnings, $16.15 EPS). Celestica revenue rose 53% YoY to $4.05B in Q1 and Q2 guidance of $4.3B implies 49% YoY growth; adjusted earnings rose 80% YoY with Q2 midpoint guidance pointing to +61%, and the stock is expected to gain ~58% to ~$568 (30x, $18.95 EPS).

Analysis

The market is increasingly paying for AI infrastructure exposure twice: once for the obvious silicon bottleneck and again for the “picks-and-shovels” layer that turns capex into deployable systems. That dynamic favors CLS more asymmetrically than NVDA near term, because CLS can still see earnings revision upside without needing the same level of multiple expansion; the stock’s recent drawdown is more interesting than its absolute valuation. NVDA remains the higher-quality long-duration compounder, but at this point the better trade is likely around catalyst timing rather than thesis discovery.

The second-order winner set extends beyond the names in the article: networking, optical interconnect, power management, and datacenter electrical gear should keep seeing order strength if hyperscaler budgets stay elevated. The risk is that headline capex numbers can overstate actual revenue timing; if customer spend is front-loaded into land/build phases, suppliers like CLS could see a short-term surge followed by a digestion period. For NVDA, the key question is not demand but mix and shipment pacing—if earnings only confirm already-embedded expectations, the stock can still be flat despite strong fundamentals.

Contrarian view: the consensus is treating “more AI capex” as automatically bullish for every layer of the stack, but the next leg of the cycle may be more selective. GPU leaders may get less incremental multiple expansion than the infrastructure enablers because the former are already priced as strategic monopolies, while the latter still trade with some skepticism. What would falsify the setup: a visible cut to 2026 hyperscaler capex, softer order commentary from datacenter suppliers, or any sign that backlog is converting slower than expected into revenue and gross margin expansion.

Key Decisions for Investors

• Go long CLS into its next earnings print later this month; this is the cleaner near-term catalyst trade because the recent pullback gives a better entry and the market is underpricing bottom-line operating leverage. Risk/reward: upside if revenue/gross margin guidance is raised; falsify if the company only reiterates and backlog conversion looks sluggish.
• Use NVDA as a quality long, but prefer call spreads or a staggered entry ahead of customer capex updates rather than chasing spot after a strong print. The trade works if hyperscalers raise 2026 capex again; it fails if guidance is merely confirmed and the stock de-rates on “good but expected” results.
• Pair trade: long CLS / short a basket proxy of AI mega-cap capex beneficiaries where valuation is already rich relative to next-quarter visibility. The goal is to isolate earnings revision upside from multiple compression risk; cover the short if hyperscaler commentary materially improves order conversion visibility.
• Set a watch item on NVDA forward revenue beats versus raised guidance: buy-the-dip only if the company pairs strong demand commentary with evidence of shipment acceleration. If the market starts fading good prints, that is a signal the AI capex story is transitioning from acceleration to normalization.