Nvidia Has Been the King of AI. This ETF Bets the Next Winners Will Be Somewhere Else.

The iShares Expanded Tech-Software Sector ETF (IGV) has gained 15% this quarter after falling more than 14% through June, while the VanEck Semiconductor ETF (SMH) has declined 14% this quarter following an 80% gain through June. The article argues that slowing AI-development expectations and rising security concerns are shifting leadership from AI infrastructure and semiconductors toward software and cybersecurity providers such as CrowdStrike, Cloudflare, and Zscaler. The outlook is bullish for software, though it is primarily an investment thesis rather than company-specific fundamental news.

Analysis

The software rebound is more likely a factor rotation than evidence that AI development has become structurally less threatening. The decisive fundamental distinction is between vendors whose products become a control plane for AI adoption—identity, endpoint, cloud/network security, observability—and horizontal application software whose seat-based pricing and implementation revenue remain exposed to AI-driven deflation. CRWD, ZS and NET fit the former group, but their recent strength raises the bar for net-new ARR and operating-margin upside; broad IGV exposure would dilute that quality with slower-growth SaaS names facing greater pricing risk.

Security spending can be countercyclical only when it is tied to mandatory controls, incident response, or insurer/customer requirements. A slower infrastructure capex cycle could pressure NVDA’s incremental revenue expectations over the next 1-3 quarters, but it does not automatically release budget to software: enterprise CIOs may instead preserve cash until AI governance standards and measurable ROI emerge. The more durable 6-18 month beneficiary is the vendor able to monetize AI-agent identity, data-loss prevention, and inference traffic protection, rather than a generic "software recovery" basket.

Consensus appears to be extrapolating relative-price momentum: semis down versus software up. That trade is vulnerable if hyperscaler capex remains resilient or if software earnings reveal that AI features are being bundled rather than monetized. Falsify the security-over-broad-software thesis if CRWD/ZS billings and remaining performance obligations decelerate materially at the next earnings cycle, or if NET fails to sustain enterprise growth and gross-margin progression; conversely, upward revisions to these metrics would justify further multiple support despite elevated valuations.

Key Decisions for Investors

• Prefer a 3-6 month quality-security basket—long CRWD, ZS and NET—over IGV. Size modestly after the momentum move; add only on post-earnings pullbacks where billings/RPO guidance is maintained or raised. Thesis requires durable enterprise growth, not AI narrative beta.
• Express rotation selectively with long CRWD or ZS versus short IGV in equal dollar beta-adjusted sizing for 1-3 months. This isolates mandatory security spend from lower-quality application software; exit if IGV relative strength breaks down less than 5% or if security bookings guidance is cut.
• Do not initiate a directional NVDA short solely on a purported AI-development slowdown. Use NVDA as the hedge leg only against an overextended security long book if evidence emerges of hyperscaler capex cuts or order-pushout commentary; absent that evidence, the semiconductor earnings-risk asymmetry remains unfavorable for shorts.
• Set an earnings watchlist for CRWD, ZS and NET: net-new ARR/billings, RPO, large-deal duration, AI-product attach rates, and sales-efficiency trends. A shift from paid security modules to free AI feature bundling is the key warning that current valuation premiums are not supported by incremental monetization.