Jensen Huang says some CEOs have a ‘God complex’ when it comes to AI apocalypse warnings, which can create shortages of critical workers

Jensen Huang argued that AI will create jobs rather than eliminate them, estimating the technology has already generated more than 500,000 jobs in recent years. He said demand for software engineers is rising and that AI should expand, not shrink, the market for coding and other professional services. The article also cites Apollo economist Torsten Slok’s Jevons-paradox view that cheaper AI-enabled work will increase total demand across legal, consulting, and financial services.

Analysis

The market is treating AI as a zero-sum labor shock, but the more relevant mechanism is demand elasticity: cheaper code lowers the cost of experimentation, so the volume of software, workflow automation, and custom internal tools should rise faster than budgets shrink. That is structurally bullish for the infrastructure stack and for companies selling picks-and-shovels into expanding developer throughput, while the primary loser is higher-multiple software that has not yet proven it can monetize the same productivity gain on the sell-side of enterprise pricing.

Second-order, this is a sentiment reset for NVDA more than a direct fundamental update. If management teams become less fearful of internal AI adoption, capex hesitation should compress, but the bigger effect is longer-duration: more workloads get pulled from human labor into compute, which supports GPU demand across inference, agentic coding, and enterprise application rebuilds over 12-24 months. The flip side is that the more AI is framed as job-creating rather than job-destroying, the less justification there is for broad software de-rating; that creates a setup for short interest to get squeezed in the weakest “AI will replace us” names.

The contrarian miss is that productivity gains do not automatically accrue to the software vendors. In many enterprise budgets, AI can expand unit volumes while still reducing per-seat pricing power, so winners will be the platforms controlling distribution, model access, and compute, not necessarily application layers with weak switching costs. That means the current selloff in software may be partially overdone, but only for names with clear usage-based monetization; for the rest, this is a margin mix problem disguised as an adoption story.

Catalyst-wise, the next 1-3 months matter for guidance season: if CEOs echo this framing and stop talking about headcount cuts, software demand narratives can stabilize quickly. Over 6-12 months, watch whether AI agents actually increase engineering hiring and cloud consumption; if they do, the market will have to reprice AI from a labor-displacement trade into a capex-and-usage expansion trade.

Key Decisions for Investors

• Stay long NVDA into the next 1-3 earnings cycles, but finance it with calls rather than outright equity; the upside is continued AI capex re-acceleration, while the main risk is a temporary narrative break if enterprise buyers slow pilot-to-production conversion.
• Buy the dip in high-quality AI infrastructure beneficiaries versus short-duration software: long NVDA / short a basket of low-switching-cost application software names over 3-6 months, targeting multiple compression in the short leg if AI fear re-enters the tape.
• Add selectively to APOS only if the thesis is that AI adoption increases content/computing demand rather than disintermediating it; otherwise keep it as a small optionality position, because sentiment-driven AI beneficiaries can lag if monetization remains unclear.
• For event-driven traders, use 2-4 week call spreads on NVDA ahead of management commentary or major platform AI launches; the setup is favorable because a modestly positive narrative shift can force systematic de-risking to reverse.
• Avoid blanket shorts on software for now; instead short the weakest balance-sheet, lowest-retention names if valuation remains disconnected from AI monetization, since the broader ‘software is dead’ trade looks crowded and vulnerable to a 1-2 quarter mean reversion.