AI's Profit Mirage: Investors Fund the Boom, Not Customers

The numbers don't add up. Economist Torsten Slok, now chief economist at Apollo Global Management, has crunched the data and found that the AI boom is being paid for by shareholders, not by paying customers. In an analysis published on August 7, he broke the AI value chain into four groups and uncovered a profit structure that contradicts every classic economic model. The result is a sector where the entities selling the finished product are the ones bleeding cash.

The upside-down economics of AI

Using data from PitchBook and Bloomberg, Slok examined the profitability of the four links in the AI chain: models & applications, cloud & compute, energy & grid, and semiconductors & equipment. The results are startling. The chip makers and equipment suppliers enjoy a stunning 41% operating margin. Meanwhile, the companies building AI models and applications are operating at a staggering negative 59% margin.

"The profits of the AI boom are currently funded by investors rather than earned from consumers."

This inversion means that every dollar of high-margin hardware revenue is effectively a transfer of the capital that deficit-running application companies have raised from venture capital and public markets. The cash flows upstream, but the value does not flow back down yet.

The $1 trillion question mark

Goldman Sachs now forecasts that AI investments will surpass $1 trillion in 2026. The staggering sum underlines the scale of the bet. Yet the technology has barely proven its economic worth. Its impact on productivity remains difficult to measure, and for many enterprise customers, the return on investment is still theoretical. The entire edifice rests on the assumption that the revenue will eventually materialize.

Names like Nvidia, Microsoft, OpenAI, Amazon, and AMD are all caught up in this gravitational pull. Nvidia's chips are selling, but they are being paid for largely with money that companies like OpenAI and Anthropic have raised, not with profits they've generated from end users.

When the music stops

Slok's concern is not a distant worry. Some companies are already facing negative free cash flow. If end customers do not quickly generate a concrete return, the massive funding could dry up abruptly. The fragile profitability revealed by Slok suggests that the AI industry is not yet a self-sustaining economic engine. It is a machine that runs on faith, and faith can evaporate fast.