One of Stanford's original AI gurus says productivity liftoff has begun after doubling in 2025

Recent debate centers on whether AI is now boosting macro productivity: Apollo economist Torsten Slok warns that employment, productivity and inflation data have yet to show effects, while Stanford’s Erik Brynjolfsson points to revised BLS figures—2025 job gains cut to 181,000 from 584,000 (vs 1.46m in 2024) and fourth-quarter GDP tracking up 3.7%—to estimate U.S. productivity rose roughly 2.7% in 2025 (vs a 1.4% decade average). Capital Economics’ Stephen Brown corroborates by noting ICT output rose even as employment fell, suggesting a nascent shift from heavy AI investment to measurable productivity gains, though several more periods of sustained growth are needed to confirm a long-term trend.

Analysis

Winners are clear: GPU/AI chip leaders, hyperscale cloud vendors and a small cohort of “power users” automating end‑to‑end workflows will capture outsized productivity and margin gains; Brynjolfsson’s ~2.7% 2025 productivity vs a 1.4% decade average implies potential operating-margin tailwinds of ~50–200bps for adopters over 12–36 months. Losers are entry‑level, labor‑intensive service providers and legacy outsourcing firms where tasks are easily automated; expect revenue mix shifts and pricing pressure in low‑value services over 6–24 months.

Supply/demand is bifurcating: demand for datacenter capacity, GPUs, advanced nodes and ASML tooling will stay tight for 6–18 months, supporting hardware pricing power and capex intensity; energy and copper demand edges higher regionally, while increased automation should be disinflationary on unit labor costs. Cross‑asset: durable productivity gains are bullish for equity multiples and long duration bonds (if disinflation crystallizes), compressive for FX carry currencies and supportive of IG spreads tightening if corporate cashflows improve.

Tail risks include abrupt regulatory action (EU/US fines or restrictive AI rules), a reversion in BLS measurement (data revisions), sudden chip supply shocks, or a Fed tightening response that negates disinflation; key short windows are next 1–3 BLS prints and upcoming earnings seasons, medium 3–12 months for capex cadence, long 12–36 months for structural adoption. Hidden dependency: benefits concentrated in a handful of firms—productivity may not diffuse without material capex by mid‑caps.

Consensus is underweight the mid‑cap “power users” and overweights only the Magnificent 7; this creates relative‑value opportunities to buy AI infrastructure and selected enterprise adopters while shorting labor‑exposed service firms. Catalysts to watch: BLS monthly revisions, Fed dot moves, ASML shipment data, and corporate capex guidance over the next 3–9 months.

Key Decisions for Investors

• Establish a tactical 2% notional position in NVDA via a 9–12 month 1:1 call spread (buy ATM, sell strike ~+20%) to capture continued datacenter GPU scarcity; target gross return 30–70% or max loss = premium, exit on +40% position move or if NVDA guidance weakens on the next two earnings calls.
• Overweight MSFT and GOOGL with a combined 4% fund weight (2% each) over 3–9 months to play cloud + enterprise AI monetization; trim if either stock outperforms by +15% or if management cuts FY capex guidance by >10% sequentially.
• Implement a 6–12 month pair trade: long a small-cap/SMID AI adopter basket (examples: PLTR 1.0%, DDOG 1.0%) and long semiconductor exposure (AMD 1.5%) financed by a 1.5% short in staffing/entry‑level exposed names (MAN 0.75%, RHI 0.75%) to capture margin divergence over 6–18 months.
• Add conditional macro hedge/positioning: if two consecutive quarterly BLS-derived productivity prints exceed +2% y/y, rotate 3–5% of portfolio into long‑duration Treasuries (e.g., TLT or 10y futures) to capture disinflation; if 10y yield falls below 3.5%, scale into duration in tranches (25% increments).