5 Stocks to Buy if AI Investing Comes Back Into Style

The article argues that AI investing is reaccelerating, highlighting strong fundamentals across Nvidia, Broadcom, Taiwan Semiconductor, SoundHound AI, and Nebius. Key growth figures include Nvidia's 73% recent revenue growth with analysts expecting 79% in Q1 and 85% in Q2, TSMC's 45% March revenue growth, and Nebius' projected 522% revenue growth in 2026. Broadcom is targeting more than $100 billion in annual AI-chip revenue by end-2027, reinforcing a constructive view on the AI hardware and infrastructure buildout.

Analysis

The important signal here is not that AI is “back,” but that the trade is broadening from pure model-training exposure into the picks-and-shovels layer that monetizes every incremental dollar of capex. That favors the infrastructure complex first: foundry, networking, custom silicon, and data-center capacity all get levered to the same spending cycle, but with different margin profiles and timing. If the next leg of AI demand is real, the market should stop paying only for headline GPU growth and start rewarding the suppliers that can convert backlog into multi-year visibility.

The second-order winner is likely the supply chain bottleneck, not the loudest brand names. When hyperscalers shift more budget toward custom silicon and distributed inference, foundry capacity and advanced packaging become the gating factor, which can keep pricing power elevated even if end-demand normalizes. That also implies the competitive field gets narrower: smaller AI software names may see sharp multiple expansion only if they can prove durable usage and low churn, because infrastructure enthusiasm usually raises the bar on software monetization.

The risk is that this move is too consensus on the obvious names and too optimistic on the smaller ones. Large-cap AI leaders can still outperform, but much of their upside may depend on investors extrapolating peak-growth narratives into 2027, which is vulnerable to any capex digestion or a pause in hyperscaler ordering over the next 1-2 quarters. For the smaller high-beta names, the key issue is not growth, it is funding and execution: if equity markets tighten or buildout schedules slip, these names can give back a large portion of gains quickly even with good topline growth.

The contrarian read is that the best risk/reward may be in relative-value expressions rather than outright longs. The market is likely overpaying for optionality in the most crowded AI beneficiaries while still underpricing the operational leverage in the semiconductor supply chain and data-center enablers. If AI spending holds, the next outperformance may come from firms with cleaner revenue visibility and less narrative dependence than the pure-play AI software names.

Key Decisions for Investors

• Long TSM vs short a basket of high-multiple AI software names over the next 3-6 months; thesis is that supply-chain pricing power and fabrication utilization are more durable than usage-based narratives if AI capex stays elevated.
• Buy NVDA on pullbacks only, with a 1-2 quarter horizon; favorable risk/reward remains, but use it as a high-quality core rather than chasing after momentum because the stock is increasingly a sentiment trade around guidance revisions.
• Add AVGO as a medium-term long and consider call spreads into the next earnings cycle; custom silicon monetization should capture more of the hyperscaler spend shift, with upside if order visibility extends beyond the current fiscal year.
• Treat NBIS as a tactical long, not a strategic core position; use small sizing or call options for 3-6 months because the upside is large if utilization ramps, but execution and capital intensity create meaningful drawdown risk.
• Avoid chasing SOUN outright; if you want exposure, prefer a call spread or pair long SOUN against a stronger infrastructure name, since the stock is most vulnerable if the market rotates back toward cash-flow quality.