Key Takeaways
- • Apollo Chief Economist Torsten Slok identifies a fundamental structural inversion in AI economics: companies with the highest profit margins operate upstream in the value chain (chipmakers at 41% margins), while companies making AI models and applications show -59% operating margins—the inverse of traditional business models.
- • AI profits are currently subsidized by investor capital rather than generated by end-customer demand, according to Slok's analysis of companies including OpenAI, Anthropic, Microsoft, Amazon, Nvidia, AMD, and Micron.
- • The AI value chain breaks into four layers with dramatically different economics: silicon and equipment (41% margins), energy and grid infrastructure, cloud and compute services, and models/applications (-59% margins)—creating a top-heavy dependency structure.
- • The disparity signals potential bubble dynamics: upstream margins are "real" but ultimately funded by capital raised by the unprofitable downstream layer rather than organic end-user demand, creating unsustainable economic dynamics.
Summary
Apollo's Chief Economist Torsten Slok has flagged a critical structural anomaly in AI industry economics that challenges conventional equity market assumptions about technology adoption and profitability. In a Friday blog post, Slok argues that the AI boom operates under inverted profit dynamics compared to traditional business models, with implications for valuations and sustainability.
The core insight concerns profit margin distribution across the AI value chain. Slok segmented AI companies into four categories and analyzed profitability data from PitchBook and Bloomberg covering major players including OpenAI, Anthropic, Microsoft, Amazon, Constellation Energy, Nvidia, AMD, and Micron. His findings reveal an inverse relationship between product position and profitability: silicon and equipment manufacturers (primarily chipmakers) command 41% operating margins—the highest in the chain—while companies producing AI models and applications operate at -59% margins, substantially underwater on current operations.
This inverts the standard business model architecture. Conventionally, companies selling finished products directly to end consumers capture the highest margins, while upstream suppliers operate on thinner returns. The AI market has flipped this relationship. Slok attributes this disparity to the source of capital flows: "AI boom's profits are currently being funded by investors rather than earned from customers," he stated. Rather than organic demand from end users driving profitability upward through the value chain, investor capital and speculative funding support downstream players, which in turn purchase upstream infrastructure and compute resources.
The structural risk Slok identifies is dependency and unsustainability. The "real" upstream margins in silicon and energy are ultimately funded by capital raised by the money-losing application layer, not by cash generated from organic customer demand. This creates a circular capital dependency: investors fund unprofitable AI companies, which then deploy that capital purchasing chips, compute, and infrastructure, generating the upstream margins. If investor enthusiasm wanes or capital availability tightens, the model compresses—threatening both the downstream companies and the upstream suppliers dependent on their purchasing power.
## HISTORICAL CONTEXT
This analysis arrives amid broader market scrutiny of AI valuations and return-on-investment metrics across the sector, though Slok's focus specifically concerns the structural mismatch between how the market is currently pricing AI companies and the actual cash generation dynamics supporting the industry.
The core insight concerns profit margin distribution across the AI value chain. Slok segmented AI companies into four categories and analyzed profitability data from PitchBook and Bloomberg covering major players including OpenAI, Anthropic, Microsoft, Amazon, Constellation Energy, Nvidia, AMD, and Micron. His findings reveal an inverse relationship between product position and profitability: silicon and equipment manufacturers (primarily chipmakers) command 41% operating margins—the highest in the chain—while companies producing AI models and applications operate at -59% margins, substantially underwater on current operations.
This inverts the standard business model architecture. Conventionally, companies selling finished products directly to end consumers capture the highest margins, while upstream suppliers operate on thinner returns. The AI market has flipped this relationship. Slok attributes this disparity to the source of capital flows: "AI boom's profits are currently being funded by investors rather than earned from customers," he stated. Rather than organic demand from end users driving profitability upward through the value chain, investor capital and speculative funding support downstream players, which in turn purchase upstream infrastructure and compute resources.
The structural risk Slok identifies is dependency and unsustainability. The "real" upstream margins in silicon and energy are ultimately funded by capital raised by the money-losing application layer, not by cash generated from organic customer demand. This creates a circular capital dependency: investors fund unprofitable AI companies, which then deploy that capital purchasing chips, compute, and infrastructure, generating the upstream margins. If investor enthusiasm wanes or capital availability tightens, the model compresses—threatening both the downstream companies and the upstream suppliers dependent on their purchasing power.
## HISTORICAL CONTEXT
This analysis arrives amid broader market scrutiny of AI valuations and return-on-investment metrics across the sector, though Slok's focus specifically concerns the structural mismatch between how the market is currently pricing AI companies and the actual cash generation dynamics supporting the industry.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.