AI Investments Yield Little Financial Return for Non-Tech Firms, Says Apollo Economist

If you spend any time following the markets, you’ve heard the promise. Artificial intelligence will revolutionize every sector, boosting productivity and profits across the board. Yet as I walk through the Financial District, past the headquarters of banks and insurers, a question lingers. Where are the tangible returns?

Apollo Global Management’s chief economist Torsten Sløk recently published a note that cuts to the core of this disconnect. His analysis is blunt. “Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to Corporate America outside of tech,” he wrote. This isn’t just a contrarian take. It’s a data point that demands attention, especially coming from a firm like Apollo whose analysts typically highlight economic strength.

The numbers backing this view are sobering. A separate survey from Cloudflare, cited in the same market discussion, provides the granular detail. It found that:

• 55% of companies have increased their AI investment.
• Only 23% said the value exceeds the cost.
• 58% reported that value merely equals cost.
• 19% found that the cost exceeds the value.
• A majority are in a costly holding pattern.
• You invest to get ahead, not to run in place or fall behind.

The issue isn’t a lack of trying. It’s a harsh reality of economics. Deploying top-tier AI is expensive. The conversation on the trading floor often turns to “token consumption,” the unit of cost for using these large language models. Cloudflare’s survey noted that 89% of companies have limited or redesigned their AI use cases specifically because of rising token costs. When the bill comes due, ambition meets a budget. This has led to a pragmatic, two-tiered approach. For mission-critical, highly sophisticated tasks, companies might stomach the cost for a premium model from OpenAI or Anthropic. For everything else, they are increasingly turning to less powerful, open-weight models that are far cheaper to run.

This creates a fundamental tension. The hype cycle assumes a swift, sweeping transformation. The balance sheet reveals a slower, more selective adoption. The financial payoff for a manufacturing firm or a retailer isn’t just about having AI. It’s about having AI that solves a specific problem at a cost that makes business sense. Right now, for many, that equation doesn’t balance. As one portfolio manager told me last week, “We’re buying the shovels, but the gold rush hasn’t started for most of the miners.”

There’s another hurdle. Beyond cost, there’s performance. The same Cloudflare data pointed out that 98% of companies have pulled back on AI deployment due to issues like failures and hallucinations. If you can’t trust the output, you can’t build a reliable business process around it. This isn’t a minor technical glitch. It’s a fundamental barrier to operational integration. When an airline plans crew schedules or a logistics company optimizes delivery routes, they need certainty. AI that occasionally invents facts or fails outright is a liability, not a tool.

So what does this mean for you? It means looking past the headlines. The staggering investments by hyperscalers like Microsoft, Amazon and Google are real. The venture capital flooding into AI startups is real. But the downstream financial benefit for the average S&P 500 company outside of Silicon Valley remains largely unrealized. This divergence is the critical story. The capital is moving, but the returns are not yet following.

This isn’t a prediction of failure. It’s a call for calibrated expectations. Technological adoption curves are never straight lines. The personal computer, the internet, the smartphone each followed a path of initial excitement, costly experimentation and then, finally, transformative integration. AI appears lodged in the costly experimentation phase for mainstream corporate America. The payoff will come not when every company uses AI, but when every company uses AI in a way that clearly improves its bottom line. Until then, the massive investments Sløk references will show up on earnings reports as an expense, not an engine. The real revolution begins when that flips.