The Maze: Apollo, the investment manager, is questioning whether the artificial-intelligence boom is improving the finances of most large US public companies. Its chief economist, Torsten Slok, published an October 24 analysis contrasting seven technology giants with the other companies in the S&P 500 stock index. The comparison raises a useful question. It does not settle every buyer's return.
The Details:
The gap is between company groups. Slok's comparison shows rising revenue per employee for the seven giants and a relatively flat series for the rest. Revenue per employee means company sales divided by headcount. Apollo adjusts the dollar values for inflation using late-2019 prices, so this is not simply a comparison of rising nominal sales. The seven are Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. The remaining group, labelled “S&P 493,” still includes technology companies outside those seven. It is neither a pure non-tech sample nor the whole US economy.
Sales, margins and productivity answer different questions. Apollo's second chart shows rising profit margins for the seven and little sustained improvement for the other group, whose margin remains roughly around 10%. Margin is profit as a share of sales; it is not sales growth. The chart does not specify the precise accounting profit measure or group weighting. A separate chart compares national productivity growth, including output per hour and a broader efficiency measure. The San Francisco Fed's methodology accounts for labor, capital and how intensively they are used. That national measure is not an audit of AI projects at the 493 companies.
Slok's interpretation goes further than the charts alone. He argues that mainstream businesses increasingly regard AI usage bills as costs and questions whether cheaper models will produce stronger sales and margins. Those are his judgments. The displayed comparisons do not identify which firms adopted which tools, compare adopters with otherwise similar non-adopters, or isolate AI from other changes. Stronger results among the seven therefore do not establish that AI caused the gap. Equally, flat results elsewhere cannot establish that every deployment failed.
MarketMaze Take: Buying AI and selling AI are different businesses. Apollo's strongest contribution is to separate the suppliers' boom from the broader financial payoff buyers hope to receive. Its weakest leap is turning a broad comparison into an absolute verdict. A flat group average is a reason to ask better questions, not a substitute for answering them.
A retailer needs its own denominator. For a merchant outside the US, the useful lesson is the distinction between a supplier's sales and a customer's profit. This US index comparison does not estimate that merchant's return. A cheaper tool only improves the business case if the resulting work delivers enough value after implementation, usage and checking costs. That is a project-level question; the seven-company comparison cannot answer it.
Flat margins leave competing explanations open. Useful AI savings could be offset by other costs, arrive later, or be absent. Apollo's charts do not distinguish those possibilities. Nor do they show how much any benefit reaches customers rather than shareholders. The evidence supports scrutiny of claimed returns. It leaves the size, timing and distribution of those returns unresolved.
Sources: Apollo, TechCrunch, San Francisco Fed
