The Pulse
Apollo Economist Finds No AI Payoff Outside Tech
Apollo chief economist Torsten Slok says revenue per worker and profit margins remain flat for companies outside the technology sector, despite heavy AI investment. His October 24 analysis points to concentrated token spending and says the
AI.info Team · Written with AI · fact-checked ·
Companies are spending heavily on artificial intelligence, but Apollo chief economist Torsten Slok says the financial returns remain concentrated in technology. In a Daily Spark analysis published October 24, Slok points to flat revenue per worker and profit margins across the S&P 493, his label for the companies outside the “Mag 7,” while both measures are rising for the seven largest technology firms. He also says there is still no sign of AI lifting productivity growth.
The contrast is between growing investment and an absent broad-based payoff. Apollo’s argument is not that AI has produced no value anywhere; it is that the gains have not yet appeared in the corporate indicators Slok considers most consequential outside tech. His assessment puts the burden of proof on the companies buying AI services, rather than on the technology firms selling them.
Revenue and margins diverge between the Mag 7 and S&P 493
“Revenue per worker is rising for the Mag 7 but is flat for the S&P 493,” Slok writes. Apollo reports the same divide in profit margins: rising for the Mag 7, flat for the rest of the index. The analysis says its revenue-per-worker figures are adjusted for inflation to the fourth quarter of 2019.
“And there are still no signs of AI boosting productivity growth.”
Torsten Slok, partner and chief economist, Apollo
For its productivity chart, Apollo uses a utilization-adjusted total factor productivity series from the Federal Reserve Bank of San Francisco, adjusted for cyclical changes in how intensively labor and capital are used. The source list names the San Francisco Fed, the U.S. Bureau of Labor Statistics and the research of John Fernald. That measure tracks the wider economy; it does not, on its own, establish whether AI caused a change or lack of change.
Token costs look like an expense, not a growth engine
Slok says mainstream companies are pulling back and beginning to treat token spending as a cost that weighs on margins, rather than as a source of growth. He contrasts that caution with venture-backed firms, which he says are spending more on tokens. Apollo’s charts cite the Ramp AI Index and data from economist Ara Kharazian at Ramp for its discussion of spending concentration and token use.
That divide matters to Slok’s economic argument: if AI-agent use and token purchases remain confined to a small slice of the economy, their aggregate effect will be limited. The analysis says many established companies do not need the latest model and are unwilling to pay for one. Apollo provides no spending total or company-by-company breakdown in the note, so its claim is about the direction and concentration of activity, not a quantified estimate of the whole market.
Cheaper models still have to improve business results
Slok’s test for a wider payoff is whether companies can use increasingly inexpensive AI to raise revenue growth or profit margins. He argues that business decisions depend on judgment and context, and that factors such as trust, accountability, workplace culture and fairness shape reputational risk. In his view, cheaper access to model output does not automatically supply those capabilities.
Apollo’s conclusion is therefore narrower than a verdict on AI’s long-term potential: it sees no current evidence of improving revenue growth or margins outside technology. The analysis does not prove that AI has had no effect on individual firms, nor does it isolate AI as the cause of the flat aggregate measures. For Slok, the broader economic case will depend on whether the S&P 493—not just the technology companies building and selling AI—can show higher revenue or margins in the data.