AI won’t pay off for the S&P 493 until CEOs rebuild the work

AI has yet to deliver broad financial gains outside major technology companies: revenue per worker and margins are rising at the Magnificent 7 but flat across the other 493 S&P 500 firms, while BCG estimates AI spending has climbed to 3.3% of revenue from 1.7% in late 2025. Azure revenue grew 48% and Google Cloud 103% in the quarter ended Sept. 30, but enterprise buyers have not broadly converted AI use into measurable returns. The author expects widespread, specific AI benefits across the S&P 493 no earlier than the end of 2027; Barclays’ roughly 5,000 announced job cuts were attributed partly to AI, but the bank has not quantified AI’s contribution.

Analysis

The investable distinction is between selling compute and capturing productivity: cloud providers can monetize AI demand before customers prove enterprise-wide returns. That supports near-term revenue at Microsoft (MSFT) and Alphabet (GOOG), but does not by itself establish durable margins or attractive returns on their AI infrastructure investment. Watch whether cloud growth translates into sustained earnings and cash generation, not just capacity demand.

For adopters, the likely lag is organizational rather than purely technical. If approvals, exception handling and controls remain human-paced, labor savings may show up as isolated capacity gains rather than consolidated margin expansion. Cutting staff before redesigning workflows could instead raise service failures, rework and control risk. Barclays' cited reductions do not establish AI-driven productivity; the article says the AI contribution is unspecified. Treat them as a cost action, not proof of monetization.

Over 1–3 months, results and guidance can test cloud demand and infrastructure returns; over 6–18 months, a more useful signal is measurable cost per accepted outcome at adopters. A second-order risk for infrastructure sellers is that model substitution, smaller models and deterministic automation lower compute consumed per task, weakening the link between AI adoption and cloud revenue. The contrarian point: flat aggregate productivity today may understate value in specific workflows, but widespread gains are not a near-term base case. No broad adopter short is justified without company-level evidence; a cloud-versus-adopter trade also risks confusing demand growth with provider profitability.

Key Decisions for Investors

• Keep MSFT and GOOG on a relative-quality watchlist, not an automatic AI-premium add. Add only if upcoming results support durable cloud growth alongside improving earnings or cash returns; reassess if growth decelerates while infrastructure spending rises.
• Do not short Barclays solely on this report. Require evidence that restructuring costs, service quality or revenue are deteriorating; the disclosed layoffs do not isolate AI’s contribution.
• For adopter exposure, screen earnings calls for workflow-level cost per outcome, exception/rework rates and realized margin changes—not AI seats, usage or announced headcount reductions. Treat absent metrics as a watch item rather than a short signal.
• Monitor model efficiency and workload mix over the next 6–18 months. Faster migration to smaller/open models or deterministic automation could reduce compute intensity and challenge the assumption that enterprise adoption scales cloud economics.