Will profits accrue mainly to chipmakers, energy providers, cloud platforms, frontier labs, application companies, data owners, or AI-enabled incumbents?
So far value has pooled in chips and cloud. Whether the model layer or the application layer captures more going forward — and whether revenue justifies the capex — is Sequoia's '$600B question.'
View on the map → · Open in Browse →
The margin data finally moved: SemiAnalysis documents frontier-lab inference margins rising sharply as agentic demand raises token prices, against Thompson's case that incumbents win through integration and compliance and Smith's argument that OpenAI's spending is a winner-take-all bet that could leave profits with chips, clouds, and applications.
In consuming intelligence, you are creating intelligence. And what you create should belong to you
Microsoft's CEO argues model providers capture customers' proprietary knowledge through usage exhaust — an institutional position from the largest AI reseller, pushing orchestration layers and open models.
The age of low gross margins for frontier model providers is over.
Margin analysis: value shifting from chips and clouds to model labs; Anthropic inference gross margins from 38% to over 70% as agentic demand raises token prices.
why pay for software when you can just ask AI to write your own application
Incumbent software survives AI through integration, identity, and compliance rather than code; SaaS faces in-house displacement.
A gross-margin framework separating scalable-investment losses from structurally unprofitable unit economics.
This isn't a business model. This is Pascal's Wager.
Competition, cash burn, and weak vertical integration could shift AI profits to chipmakers, clouds, and applications.