Frontier AI models are commoditizing within months, so the biggest fortunes of the intelligence era will not be made by the labs building them. That is the central argument of a new essay by Will Manidis and Joe Roberts, titled The Commanding Heights, which says durable wealth will accrue to whoever controls the chokepoints that AI must pass through: distribution, payment rails and a stablecoin-based dollar.
The essay, published on October 29, borrows its frame from Lenin's 1922 defense of private trade: the state held "all the commanding heights," the vital branches through which everything else flowed. In today's version, a chokepoint is not a fleeting bottleneck but a position that durably captures a share of the economic activity it enables — and whose grip strengthens as that activity multiplies.
Why AI labs may not capture the wealth
The authors compare the moment to the railroad age: Gutenberg went insolvent, and the great railroad fortunes were captured by men who never built a locomotive. Rational capital allocation at the level of a single lab or hyperscaler, they argue, can destroy capital in the aggregate, because spending keeps rising even as model capability diffuses across the economy.
Their essay also points to a policy bind: American growth, household wealth and credit now share one common exposure — AI spending, stock prices and model capability rising together. That, they write, is why Washington is likely to lean on increasingly central backstops, from retirement savings to the Treasury, to keep that line rising.
- Training costs rise roughly 3.5x per year, keeping lab and hyperscaler spending on itself even as returns fall.
- Frontier models become commodities within months, pushing revenue toward early adopters who embed them in existing workflows.
