How sustainable is the AI rally? Jefferies weighs in

AI stocks have driven more than 80% of the S&P 500’s gains in 2026, but Jefferies says the rally is earnings-supported rather than multiple-driven, with the AI basket’s forward EPS estimates up more than 30% since mid-2025 and PEG at 0.6x. First-quarter beats were strong — 86% for earnings and 82% for sales — yet the market is rewarding beats mainly in AI while broader sectors see muted reactions. Management optimism on earnings calls reached 95%, but the U.S.-Iran conflict is weighing on sentiment and helped push oil higher.

Analysis

The key takeaway is not simply that AI is carrying index returns, but that the leadership is becoming more self-reinforcing through revisions, not multiple inflation. That matters because a rally driven by upward estimate momentum can persist longer than one driven by valuation expansion, especially when the rest of the market is barely growing earnings at all. The hidden risk is concentration: if one or two AI sub-industries keep compounding, passive flows and benchmark pressure will keep forcing capital toward the same crowded winners, even if breadth remains poor.

The second-order effect is that valuation dispersion inside AI is now more important than the sector call itself. The market is rewarding businesses with near-term monetization and penalizing those with longer-dated capex payback, which creates a cleaner long/short opportunity than an outright beta trade. That also implies the next leg of underperformance may hit the “story” names first if earnings upgrades slow, while memory/compute beneficiaries can keep outperforming as long as supply remains disciplined and order visibility stays tight.

Geopolitical noise is a real volatility catalyst, but near term it is more likely to affect factor rotation than fundamentals. Higher oil can pressure consumer-sensitive and industrial names through margin and demand channels, yet the market is currently telling us that macro shocks are being filtered through earnings quality: beats only matter when they come with revisions. The contrarian view is that consensus is overpricing the durability of top-down AI enthusiasm and underpricing how quickly this becomes a stock-picker’s market within the theme.

Key Decisions for Investors

• Stay long the AI basket, but express it via the cheapest subsectors rather than the obvious leaders: favor memory/compute suppliers over hyperscalers and chip designers for the next 1-3 months. Risk/reward is better because revisions are still positive while PEGs remain compressed.
• Fade relative valuation in the expensive AI laggards with a pair trade: long a memory/compute proxy against short a chip-design or semiconductor-capex proxy. Target a 6-10% spread over 1-2 quarters if estimate revisions continue to favor monetization over capex intensity.
• For SMCI, avoid chasing strength outright; use a call spread or wait for a 5-8% pullback before initiating. The name can continue to benefit from AI infrastructure spend, but it is more vulnerable to any disappointment in order cadence or gross margin normalization.
• For APP, treat as a higher-beta monetization play and buy on weakness only if ad spend and AI-driven optimization remain intact. Best expression is a medium-dated call spread, since upside can persist but sentiment can reverse quickly if breadth narrows.
• Keep a tactical hedge against oil-driven factor rotation by shorting a consumer/margin-sensitive basket for 2-6 weeks. The goal is not an energy call; it is protection against higher input costs and weaker consumer sentiment if geopolitical headlines intensify.