The news
Apollo Global Management Chief Economist Torsten Slok reported that artificial intelligence investments have produced zero payoff for corporate America outside the technology sector. The assessment covers the S&P 493, the S&P 500 minus its largest tech constituents. Slok’s finding rests on observed capital expenditure patterns and the absence of corresponding gains in output or margins at non-tech firms.
Context
Firms in banking, manufacturing, retail, and health care have directed substantial budgets toward AI models, cloud capacity, and integration projects over the past several years. Those outlays have increased demand for chips, data centers, and software licenses supplied by a narrow set of technology companies. Before the current wave of model deployments, productivity gains at most firms came from incremental software adoption and process redesign rather than large-scale AI infrastructure.
Detail
Slok’s comment appeared in a Bloomberg Technology article published October 25, 2026. It draws on his review of corporate financial data and spending trends across the broader index. The report contains no company-specific breakdowns or quantified productivity metrics. The statement records present outcomes rather than predictions about future returns.
Why it matters
Non-tech companies continue to fund AI initiatives even as the economist records no visible lift in revenue per employee, operating margins, or unit costs. This pattern indicates that current spending functions primarily as a transfer to infrastructure and model providers rather than a broad productivity tool. Engineering teams at traditional firms therefore face a direct test: any new deployment must show measurable efficiency or revenue impact within the current budget cycle, or further commitments should be deferred.
Founders and product leaders at non-tech companies now have clearer evidence that blanket adoption carries immediate costs without offsetting gains. Roadmaps that tie every AI project to a specific operating metric—such as reduced support tickets, faster order processing, or lower defect rates—become the minimum standard for continued funding. Teams that skip this step risk accumulating technical debt and vendor lock-in while core business metrics remain unchanged.
For investors, the observation reinforces the existing concentration of market returns. Technology suppliers capture the capital deployed across the index, while downstream users absorb higher operating expenses without corresponding improvements in earnings. This imbalance can persist as long as competitive pressure and vendor marketing keep budgets elevated. Portfolio managers tracking the S&P 493 may therefore treat AI-related line items as a cost center until concrete productivity data appears in earnings releases.
Internal technical staff should also adjust expectations. Claims from vendors that models will deliver rapid returns lack support in the current data for most industries. Pilot programs that begin with narrow, instrumented use cases and require documented before-and-after measurements offer the only reliable way to separate actual gains from projected ones. Budget holders who insist on these controls will have an easier time deciding which projects to scale and which to abandon.
The economist’s assessment does not rule out future productivity improvements. It simply states that none have registered so far for the majority of companies in the index. That distinction matters for planning cycles that run through 2027 and 2028.
---
Sources:
No comments yet