Corporate AI spending has not produced a visible financial gain for most large US companies, according to an analysis by Apollo Global Management chief economist Torsten Slok. He examined the S&P 500 after excluding Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla, a group he calls the “S&P 493.”
Profit margins and revenue per employee in that broader group remain roughly flat, Slok said, offering no aggregate evidence that AI has increased productivity outside the largest technology companies. He argues that cheaper machine intelligence does not automatically produce better business judgment, which also depends on accountability, trust, culture and unwritten rules.
The finding challenges the assumption that widespread experimentation should already appear in corporate results. It is not proof that every non-technology deployment has failed. An index-level comparison can hide successful projects inside individual companies, and benefits may take time to alter staffing, revenue or margins. Other costs and economic changes can also offset gains.
The practical question for buyers is therefore narrower than whether AI is useful in general. Companies need to connect each deployment to measurable cycle time, error rates, revenue or avoided cost rather than counting licenses and pilot programs as progress. Slok’s data suggests that, at least so far, those local improvements have not combined into a clear financial shift across corporate America.