TL;DR — Key Takeaways
- AI’s financial payoff remains uneven. Apollo’s analysis shows rising revenue per employee and profit margins among the Magnificent Seven, while the remaining S&P 500 companies show relatively flat trends.
- Productivity gains don’t always translate into financial returns. Deloitte found that 66% of surveyed organizations reported productivity or efficiency improvements, but only 20% reported increased revenue.
- Enterprise AI costs are drawing greater scrutiny. Token spending and other operating expenses can constrain adoption as businesses seek measurable returns.
Despite massive corporate investments in AI, most U.S. companies have yet to see measurable gains in revenue, profit margins or productivity, according to a new analysis from Apollo Global Management, a major U.S. investment firm.
Apollo chief economist Torsten Slok examined the performance of the S&P 500, comparing the seven tech companies with the remaining 493 firms. His findings indicate a sharp divide between the financial performance of leading tech giants and the broader corporate market.
The Magnificent Seven, which includes NVIDIA, Microsoft, Alphabet, Amazon, Meta, Apple and Tesla, have enjoyed substantial growth in revenue per employee. This figure climbed from approximately $550,000 in 2022 to nearly $800,000 this year, according to Apollo.
In contrast, revenue per employee among the other 493 companies has remained near $550,000 since 2019.
Profit margins reveal a similar gap. While the seven tech leaders have expanded their margins significantly since 2023, the remaining S&P 500 companies have seen little improvement, with margins hovering around 10%.
Perhaps most concerning for enterprise AI buyers, Apollo found no clear evidence that AI has boosted productivity across the broader U.S. economy. “The bottom line is that the AI story will be decided in the S&P 493, not the Magnificent Seven, and so far there is no evidence that revenue growth or margins are improving outside of the tech sector,” Slok wrote.
However, some industry experts were skeptical of the report’s conclusion, given that AI is still relatively new in the enterprise.
“The implementation of AI tools at a meaningful scale has only been ramping up this year, which makes it too early to measure productivity gains,” Gil Luria, Head of Technology Research at D.A. Davidson, told Techstrong.ai.
Supported by Related Research
The findings raise questions about the return on enterprise AI spending, particularly as companies move beyond experimental deployments and seek measurable business results.
Research from other consulting firms supports Apollo’s concerns. A PwC survey of 4,454 CEOs found that 56% reported neither higher revenue nor lower costs from AI investments. While 30% reported revenue gains and 26% reported cost reductions, just 12% achieved both. Deloitte reports that 66% of surveyed organizations achieved productivity or efficiency gains from AI, 40% reduced costs, and 20% increased revenue
McKinsey offers a somewhat more optimistic assessment, suggesting that enterprises are approaching a stage where AI pilots and early deployments could begin producing returns. Yet its research also found that the share of businesses reporting improved earnings from AI remained unchanged.
A key challenge is the cost of running AI applications, particularly those that consume large amounts of tokens, the units used to process information through AI models. Slok noted that established companies are becoming more cautious about token expenses, while some venture-backed startups keep spending aggressively.
The demand for AI processing is also sensitive to pricing, Apollo found. When costs rise, businesses may reduce usage or adopt less expensive open-source models.
This creates a potential challenge for the leading tech vendors investing billions of dollars in data centers and next-gen computing infrastructure. If mainstream businesses resist higher AI operating costs, the market for premium AI services could prove more limited than infrastructure spending suggests.
But lower prices alone may not solve the enterprise AI profitability problem. Slok argues that business intelligence requires judgment, including an understanding of organizational culture, accountability, trust and ethical considerations. While AI models can process information and generate content, these more complex decision-making requirements remain difficult to automate.
To be sure, the concentration of AI-related growth may have broader economic impact. JPMorgan previously estimated that just 42 AI-related stocks accounted for 60% to 80% of S&P 500 earnings growth, stock returns and capital spending since January 2024. Economist Paul Krugman has also warned that massive investments in AI data centers could divert capital from housing, manufacturing and other sectors while pushing interest rates higher.
