Apollo Economist: AI Has Delivered No Financial Returns for Most Companies

After years of massive spending on infrastructure, models, and tools, a recent economic report raises a question that can no longer be ignored: where are the actual financial results of artificial intelligence on corporate balance sheets? The answer offered by Torsten Slok, chief economist at Apollo Global Management, is blunt to the point of discomfort: outside the technology sector, AI has yet to translate into any meaningful financial return for major US companies.

The "493" Concept and What It Reveals About Earnings

Slok built his analysis on a striking approach, excluding from the S&P 500 index seven tech giants: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla. What remains is what he calls the "S&P 493" — the companies that represent the broad real economy, spanning manufacturing, retail, financial services, healthcare, and logistics.

Looking at this group, the analysis found no clear trace of artificial intelligence in the metrics that should improve first:

• Profit margins: Essentially flat, with no improvement that can be linked to technology adoption.
• Revenue per employee: Also unchanged, despite being one of the most important measures of actual productivity.
• Revenue growth: No evidence of acceleration driven by AI tools.

In Slok's own words, the technology — despite enormous investment and widespread hype — has delivered "zero financial return" for US companies outside the tech sector.

Is It Simply a Matter of Time Lag?

The most common explanation for this gap is that non-tech companies move slowly, and that the gains will emerge later as applications mature. But Slok appears skeptical of this argument, holding that converting "cheap intelligence" into revenue growth and margin expansion is a far harder task than assumed.

His core argument concerns the nature of business itself. He describes artificial intelligence as "an excellent calculator," but adds that intelligence in a business environment is mostly judgment and discretion, not computation. The decisions that determine a company's fate are bound up with accountability, trust, institutional culture, unwritten rules, ethical trade-offs, and notions of fairness — elements that shape reputational risk and ultimately affect revenue and profit. Current models, in his view, do not understand this layer of work.

Consistency With Other Studies

This conclusion does not stand in isolation. Several recent studies have reached a similar finding: so far, generative AI has not produced large, measurable productivity gains at the macroeconomic level. Successful cases do exist, but they remain confined to narrow tasks or limited teams, and are not scaling widely enough to show up in consolidated financial statements.

This divergence between talk of radical transformation and the accounting results recalls what economists call the "productivity paradox" — the interval between the spread of a general-purpose technology and the appearance of its actual impact, a period that can stretch for many years during which money is spent on learning and re-engineering processes before the gains are harvested.

What Does This Mean for Companies in Practice?

This analysis should not be read as a verdict of failure for artificial intelligence, but rather as a call for greater realism in planning. In our assessment, there are clear practical lessons:

• Start with the problem, not the tool: Projects that begin with "we want to use AI" usually end as pilots that never reach production.
• Tie every initiative to a financial metric: Shorter cycle times, lower cost to serve a customer, or a higher conversion rate — in numbers that can be tracked.
• Redesign the process itself: Bolting an intelligent model onto a legacy workflow produces only marginal improvement.
• Invest in data and governance: Data quality and usage controls are what separate a prototype from a trusted system.
• Keep humans in the decision loop: In areas carrying legal or reputational risk, human judgment remains a requirement, not a luxury.

Conclusion

The most important message in the Apollo report is not that artificial intelligence lacks value, but that value does not materialize automatically through spending alone. The difference between the companies that reap the gains and those that remain stuck in pilot mode will not lie in budget size, but in the ability to rebuild processes and measure impact rigorously. Until that happens at scale, the numbers will stay far quieter than the marketing rhetoric surrounding the technology.

✦ بقلم فريق دروب أيديا