The 4 Best AI Stocks to Buy in May

The article is broadly bullish on AI infrastructure names, highlighting Nvidia's expected 72% revenue growth, Broadcom's AI semiconductor revenue of $8.4 billion in Q1 fiscal 2026 (+106% YoY), and TSMC's raised 2026 revenue growth guidance to 30%. It also points to Nebius as a high-growth AI cloud play, with analysts forecasting 523% revenue growth in 2026 and 206% in 2027. Overall, the piece argues that AI compute demand remains strong and supports further upside across the group.

Analysis

The market is still underpricing how concentrated the current AI spend cycle has become: when hyperscalers are constrained on rack power, the capital budget shifts from experimentation to hard procurement, which tends to extend the cycle for the best-positioned silicon and foundry names. That favors NVDA and AVGO near term, but the bigger second-order winner is the manufacturing bottleneck behind them—TSM’s leverage increases because every incremental design win from custom silicon and GPUs still has to clear the same advanced-node capacity and packaging constraints.

The key distinction is that this is no longer just a GPU trade; it is a power, packaging, and yield trade. As custom ASIC adoption rises, AVGO gains share of wallet even if unit growth normalizes, while NVDA’s risk is not demand collapse but mix pressure if hyperscalers continue substituting away from general-purpose accelerators into lower-cost workload-specific silicon. That substitution is likely gradual, measured in quarters to years, which supports both names now but argues against assuming linear multiple expansion.

NBIS is the highest-beta expression of the theme, but it is also the most fragile: its growth path depends on sustained external capital formation in AI infrastructure, not just customer demand. If funding windows tighten or hyperscalers bring more workload in-house, neoclouds can re-rate sharply lower even while the broader AI complex stays bid. The consensus is probably too comfortable with forward revenue hockey sticks in NBIS-like names and too focused on headline growth versus durability of gross margins and customer concentration.

Contrarianly, the risk is not that AI demand disappears; it is that the trade becomes crowded and the market starts penalizing “good growth” relative to “own the toll road.” That would favor TSM and AVGO over NBIS on a 6-12 month horizon. Near term, though, sentiment and guidance momentum should keep the group supported unless a major hyperscaler signals a capex pause or supply-demand balance in advanced packaging loosens faster than expected.

Key Decisions for Investors

• Maintain core long NVDA/AVGO/TSM exposure for the next 3-6 months, but tilt incremental capital toward TSM on pullbacks: it has the cleanest risk/reward because it monetizes both GPU and ASIC demand with less customer-specific execution risk.
• Add AVGO on any 5-8% correction over the next 1-2 months; custom silicon share gains should support multiple durability even if AI unit growth slows, with better downside protection than NBIS.
• Use NBIS as a tactical long only, not a core holding: buy on weakness for a 1-3 month trade, but hedge with a smaller short in a neocloud or high-burn AI infrastructure peer if available; target asymmetric upside, but assume elevated drawdown risk.
• Consider a pair trade long TSM / short a basket of lower-quality AI infrastructure names over 3-12 months: this expresses the view that the semiconductor toll collectors outperform capital-intensive beneficiaries if AI spend broadens but becomes more selective.
• Do not chase NVDA after strength; instead, wait for any sign of hyperscaler capex digestion and then re-enter on a 7-10% retracement. The stock remains fundamentally strong, but the incremental upside from here is more dependent on continued multiple support than on surprise revenue acceleration.