NVIDIA's CEO predicted the meteoric rise of AI before almost everybody else

Jensen Huang’s quote that "AI is going to eat software" is highlighted as an early warning about the software sector’s vulnerability as agentic AI advances. The article argues NVIDIA remains central to AI infrastructure buildout, while SaaS and broader software companies face increasing disruption but have not been eliminated. This is primarily a thematic commentary piece with limited immediate market-moving information.

Analysis

The market is still mislabeling this as a software obituary when it is more likely a margin-reset event. The first-order beneficiaries are not just NVDA and the obvious GPU stack; the second-order winners are the firms controlling data-center power, networking, and model distribution layers, because AI shifts value from seat-based software pricing to usage-based inference economics. That transition tends to compress gross margins in application software before it expands total addressable spend, so the near-term loser basket is the high-valuation SaaS cohort with weak proprietary data and low switching costs.

For NVDA, the key issue is not whether demand exists, but whether the current cadence of capex can sustain its growth rate once hyperscalers re-optimize around cost per token and energy constraints. If AI agents reduce the need for human-facing workflow software, enterprises may actually increase spend on infrastructure while cutting legacy SaaS budgets, creating a temporary divergence where NVDA outperforms the software index for several quarters even if overall IT spend growth moderates. The hidden risk is supply chain normalization: once packaging, memory, and networking bottlenecks ease, pricing power can move down-stack faster than the market currently assumes.

Contrarianly, the consensus may be overestimating the speed of SaaS extinction. Most enterprises will not rip and replace mission-critical software on an AI headline cycle; instead, they will layer agents on top of existing systems, which delays the revenue hit but also delays the upside for pure-play AI application vendors. That means the more attractive trade is not an indiscriminate short software basket, but a relative-value expression against the most exposed names where AI can directly commoditize core workflows over 12-24 months.

Key Decisions for Investors

• Stay overweight NVDA versus the broad software complex for the next 3-6 months; the setup favors infrastructure monetization ahead of application-layer disruption, with a cleaner earnings revision path than SaaS.
• Initiate a pair trade: long NVDA / short an equal-dollar basket of high-multiple, low-differentiation SaaS names most exposed to AI workflow replacement over the next 12-24 months; target 2:1 downside capture versus upside if AI adoption slows.
• Buy put spreads on the most expensive software indices or ETF proxies into any post-earnings strength; the risk/reward improves if management teams start quantifying AI-driven seat compression or pricing pressure.
• Add selectively to networking and power-infrastructure beneficiaries as a secondary trade behind NVDA; these names can outperform if hyperscaler capex remains elevated even as software budgets get reallocated.
• Avoid shorting the entire software sector outright; use relative-value shorts only, because enterprise adoption will likely stretch the revenue decay curve and can force a sharp squeeze in crowded bearish positioning.