You’re looking at the AI revolution all wrong, top economist says: 40% unemployment and a 3-day work week are the same thing

Key framing: a 40% unemployment scenario is roughly equivalent to everyone working 60% of prior hours (e.g., a 3-day week); historically U.S. annual working hours fell ~40% from ~3,000 to ~1,800 since 1870. Corporate behavior is currently retaining AI efficiency gains (executives increasing output/headcount rather than cutting hours), while evidence shows AI is creating job churn (≈5M jobs created vs 4.8M destroyed monthly) and raising worker fatigue. Policy action (an 'AI dividend', more holidays, or regulation) is the proposed lever to distribute gains; Tabarrok also flags upside from AI-driven medicine — a cited full cancer cure could imply a ~$50 trillion global economic boost, with even 10% mortality reductions transformative. Near-term market impact is limited and more thematic/structural than price-moving.

Analysis

Corporate behavior will determine whether AI becomes a consumer leisure windfall or an earnings multipler for incumbents. If managements choose to monetize per-employee throughput rather than return hours to workers, expect sustained capex into GPUs, data fabric, and proprietary LLM stacks that amplify gross margins and revenue-per-employee for cloud and systems vendors over the next 6–24 months.

Second-order winners are non-obvious: experiential consumption and high-margin live entertainment should outgrow traditional consumer staples if discretionary time expands, while commercial office real estate and middle-market staffing firms face secular demand erosion as firms compress headcount or decentralize tasks. Supply chains for semiconductors, power/cooling, and data-center construction will see demand cascades that benefit upstream equipment and specialty contractors for multiple years.

Key catalysts that will re-rate these trades are legislative or labor policy (wage floors, AI taxes, mandated reduced hours), major healthcare breakthroughs enabled by AI, and corporate earnings commentary quantifying revenue-per-employee gains. Tail risks: rapid regulatory clampdowns, generative-AI reliability failures that stall enterprise deployments, or a macro recession that forces firms to reverse hiring/automation plans — any of which could unwind multiple sectors within quarters.

From a portfolio construction standpoint, favor concentrated exposure to AI infrastructure and consumer-experience winners while hedging with short-duration or equity-derivative protection on real-estate and staffing exposure. Time horizons should be 6–36 months; keep position sizing disciplined because technology adoption is uneven and political interventions are increasingly probable.

Key Decisions for Investors

• Buy a concentrated AI-infrastructure pair: Long NVDA (or 18–24 month call spread) + Long MSFT stock. Timeframe 6–24 months. R/R: asymmetric upside from continued GPU/cloud spending; downside: valuation compression if enterprise AI adoption stalls. Use a 15–20% trailing stop on delta-equivalent exposure.
• Initiate a tactical short in office-exposure: Short VNO (Vornado) or buy protection on an office-heavy REIT basket (12–36 months). R/R: secular downside as corporates reconfigure real estate; risk: urban office reopening or property repurposing cushions losses. Size as a hedge against tech longs (10–25% of net exposure).
• Buy leisure/experience exposure: Long LYV or DIS (12–36 months). R/R: capture upside if discretionary hours rise and spend shifts to experiences; risk: cyclical downturns that hit ticketing and travel. Consider selling shorter-dated calls to finance entry and improve margin.
• Take asymmetric biotech optionality on AI-enabled medicine: Long-dated calls on ILMN (24–48 months) or a small basket of genomics/AI-drug-discovery names. R/R: single clinical/AI breakthrough can be multi-bagger; risk: binary clinical/regulatory outcomes. Cap exposure to <3% of portfolio to preserve convexity without overconcentration.