Sample post: placeholder content to preview the site design. It contains no real data, statistics, quotes, or sources and is not investment advice.
Where Value Might Accrue as AI Shifts from Training to Inference
A sample note on how to think about the AI value chain as usage grows: infrastructure, models, and applications, and the questions that separate durable positions from temporary ones.
Sample post. Placeholder content for the site preview. Generic framing only: no real data, statistics, quotes, or sources. Not investment advice.
The question
As AI moves from being built to being used, the economics may shift from one-time training efforts toward ongoing inference. That raises a simple question with complicated answers: who captures the value?
Three layers
- Infrastructure: chips, data centers, power, and networks that run models.
- Models: the systems themselves, whether proprietary or open.
- Applications: products that put AI to work inside a specific workflow.
What to look for
Durable positions tend to combine something scarce (capacity, data, distribution, or trust) with a reason for customers to keep paying. This sample note will be expanded into a full analysis with read-throughs.
Investor read-throughs
What this means for you
Three lenses on the same analysis. Not investment advice.
Lens 01
Public Markets
Separate companies exposed to AI usage from those exposed only to AI build-outs.
- Ask whether revenue grows with ongoing usage or with one-time infrastructure spending.
- Look for disclosure on how AI features affect pricing, retention, and margins.
Lens 02
Private Markets
Underwrite distribution and data advantages, not model access alone.
- Model capabilities can be widely available; proprietary workflows and customer relationships are harder to copy.
- Stress-test gross margins under different compute-cost assumptions.
Lens 03
Entrepreneurs
Build where you own the workflow and the outcome.
- Choose problems where AI improves an outcome a customer already pays for.
- Design so that falling model costs help your margins rather than erode your differentiation.
This analysis is for informational purposes only and is not investment advice or a recommendation to buy or sell any security.
- #inference
- #value-chain
- #applications
Newsletter
Get the next deep dive by email.
Related analysis
The Power–Compute–Connectivity Loop: A Framework for Investing Across AI Infrastructure
AI demand, chip supply, power availability, and network capacity form a single loop. This sample deep dive lays out a framework for locating the binding constraint, and what each investor type might do about it.