Decoding the AI Stack - Compute, Models and who captures the Value

If you are trying to understand where the real investment opportunity in AI lies, start with one question: will the companies spending the most on AI necessarily capture the most value from it?

Not necessarily.

Hyperscalers are committing enormous amounts of capital to chips, power and data-centre infrastructure. At the same time, China is demonstrating that competitive AI capabilities may also emerge through a more capital-efficient route. This creates an important distinction between where capital is being deployed and where pricing power may eventually remain.

To understand that distinction, we first need to look at how the AI stack is structured.

The Stack: Factories Below, Products Above

At its simplest, the AI industry can be understood in two layers:

●       Infrastructure layer: Power, chips, servers and data centres that provide the computing capacity AI systems need.

●       Application layer: Models, copilots and software products that customers and businesses actually use.

Most hyperscaler capex is flowing into the infrastructure layer, on the assumption that companies controlling more compute today will be better positioned to dominate the models and applications built on top of it tomorrow. But owning the factory does not automatically mean capturing the greatest share of value from the products running on it. That is where the investment debate begins.

The Funding Problem Hiding in Plain Sight

The scale of AI investment is widely discussed. What receives less attention is how these investments may eventually be funded. Google, Meta, Amazon and other hyperscalers are increasing their spending as demand for computing capacity continues to grow.

For some, these investments may eventually exceed what operating cash flows can comfortably support, increasing their reliance on external funding. The real question is not how much they are spending, but whether the infrastructure being built can generate returns high enough to justify both the capital committed and the cost of funding it. This is where credit markets begin to offer an important second opinion.

What Credit Spreads Are Signaling

Equity markets may remain optimistic about AI-led growth, but credit markets are beginning to show greater caution. Wider bond spreads and higher CDS (credit default swaps) premiums suggest that lenders are paying closer attention to rising AI spending, potential borrowing needs and the time these investments may take to generate returns.