What Harry Sonneborn Would Tell Sam Altman
OpenAI has a story. What it needs is a mechanism.
Welcome to the sixth edition of the 2026 season of Drift Signal. As a reminder, I serve as Head of Research at Vsquared Ventures. This is my personal newsletter, which examines macro and markets through the lens of late-cycle investment theory, and the views expressed are my own.
Here is some recently published content that could be of interest:
In Currency of Power, Marieke Flament and I examine how China is open-sourcing its intelligence while America is open-sourcing its currency, and why the two are converging into a new version of an old trap: Wide Open, Locked In.
Also in Currency of Power, Marieke and I interview Kurt Hemecker, CEO of Gold Token SA at MKS PAMP, on tokenising gold, the lessons of Libra, and why privacy and anonymity are not the same thing: Not Every Gold Token Is the Same.
My Vsquared colleagues Thomas Oehl and Benedikt von Schoeler report from our recent deep tech ecosystem event in Munich, where the conversation shifted from whether frontier science works to whether it can be industrialised at scale: New Horizons: The Next Chapter.
Now, in today’s edition, I examine OpenAI’s valuation problem through the lens of business history. Starting from Aswath Damodaran’s diagnosis that Sam Altman pitches like a trader rather than a CEO, I draw a parallel with Ray Kroc’s McDonald’s in the early 1960s: a business with genuine scale and a proven system, but a financial structure that could not support its own ambitions.
The solution, as McDonald’s CFO Harry J. Sonneborn understood, was never about selling more hamburgers. It was about securing a second, structural asset beneath the entire operation, one that every participant in the ecosystem depended on and could not bypass.
I then apply that logic to the AI economy, show why owning infrastructure is not the right answer for OpenAI specifically, and propose that building the global compute market is the asset that could finally give Sam Altman a story worth telling to serious investors.
1/ Sam Altman cannot explain OpenAI’s valuation
10 days ago, I listened to Aswath Damodaran’s recent appearance on the Meb Faber podcast. I highly recommend it because it captures the growing unease surrounding the AI boom. Discussing OpenAI, the “dean of valuation” argues that the company’s valuation cannot be justified by any conventional financial model. In Damodaran’s view, when Sam Altman defends this premium, he pitches like a trader spinning a narrative rather than a CEO explaining a sustainable economic mechanism.
To Damodaran’s point, here are some numbers. OpenAI generated around $13 billion in revenue in 2025, yet its projected cash burn is rising toward $47 billion by 2028. Despite this capital drain, OpenAI’s valuation has recently crossed the $700 billion mark, implying a price-to-revenue multiple that has absolutely no precedent in the history of software.
AI enthusiasts are quick to dismiss such challenges as luddism. But Damodaran’s discomfort does not stem from a denial of the AI opportunity as a whole. Just as many of us, he recognises the importance of the technological shift. Rather, his concern is rooted in the absence of a credible path from the company’s current, negative cash flows to its stated valuation. The narrative told by OpenAI management relies on the promise of future abundance in general, but the story is missing a mechanism that explains how value will actually be captured and returned to OpenAI shareholders.
This leaves us with an interesting financial puzzle. The question is not whether OpenAI is a valuable company, but rather how it could ever become as valuable as Altman claims.
2/ McDonald’s Ray Kroc once had the same problem
In the early 1960s, Ray Kroc, the man who built McDonald’s into a global brand, found himself in a bind similar to Altman’s. He had built an operation with enormous influence, a proven system (the legendary “Speedee System”, pioneered by the McDonald brothers), and a rapidly growing network of franchisees.
Yet beneath the surface of this apparent success, his financial structure was quietly failing. The McDonald’s brand and its detailed operating manual—rules so precise that, as Paul Graham once wrote, they function like a piece of software—were real and valuable assets. However, Kroc soon discovered that extracting thin commissions from the sale of cheap hamburgers could not generate the cash flow required to fund a sprawling empire.
We can understand Kroc’s predicament through the framework of a business’s two distinct financial loops, which I introduced in an edition published in December 2020:
What Kroc had built was what I call a “trading loop”: the franchise commissions on burger sales. High in frequency, thin in margin, and wholly dependent on the daily performance of individual franchisees, this stream of cash kept the lights on but could not fund the expansion of McDonald’s at the scale that Kroc envisioned.
What McDonald’s lacked was a “capitalist loop”: a capital-heavy asset inserted into the production process for the long term, one that would generate increasing returns as the network scaled and give McDonald’s a structural grip over every franchisee in the system. Without it, most of the value Kroc created accrued to others rather than to McDonald’s itself.
As I listened to Damodaran discussing OpenAI’s valuation, I could not help but see a parallel with McDonald’s. Under Altman’s management, the company has successfully built a trading loop—subscriptions, API charges, and a growing base of paying users—on top of a large language model, a software experience and a whole ecosystem that together have captured the global imagination. But under the current subscription model, software and algorithms alone cannot fund the staggering capital requirements needed to sustain them. The question we must ask now is what the much-needed capitalist loop actually looks like in the AI economy.
3/ Capitalism requires more than scale
To understand OpenAI’s current problem, it helps to look to the French historian Fernand Braudel and his three-layer framework of the economy.
At the base sits material life, the informal transactions of daily existence. Above it is the market economy, where merchants engage in relentless, margin-crushing competition. And at the top lies capitalism—an ensemble of processes that allows an enterprise to escape the rigours of the market economy by inserting capital into the production process to pursue increasing returns to scale.
As the economist W. Brian Arthur defined them in his landmark 1996 article published by the Harvard Business Review, increasing returns represent “the tendency for that which is ahead to get further ahead”. They are the mechanisms of positive feedback that reinforce success, and mastering them is what separates a true capitalist enterprise from a mere business trapped in the competition-driven market economy.
Yet increasing returns alone are not sufficient. For such returns to compound and translate into sizable profits, the business needs those two financial loops that fit together. The trading loop generates a constant stream of cash flow from recurring transactions: burgers sold, subscriptions charged, APIs called. The capitalist loop is slower and heavier: capital inserted into a structural asset for a long period, generating increasing returns that compound over time.
One clarification matters here. The trading loop is not the same as revenue. It is the high-frequency, market-economy side of the business: real, but thin on margin and constrained by competition. At McDonald’s, the franchise commissions were the trading loop. By contrast, the brand and the franchise system were assets with the potential to generate increasing returns as the network scaled. But Kroc could not realise that potential without a second, structural asset to anchor McDonald’s capitalist loop.
OpenAI faces the exact same problem. Its subscriptions and API charges are the trading loop: the hamburger commission of the AI economy. The model and the developer ecosystem built around it are assets capable of generating increasing returns. But as Damodaran argues, possessing such assets is not enough. Without a capitalist loop to convert them into durable, compounding returns, they remain a story rather than a valuation. That is precisely what is missing from Sam Altman’s pitch.
4/ Harry Sonneborn understood what Ray Kroc did not
Harry Sonneborn, who would go on to become McDonald’s first chief financial officer (and later its CEO), is remembered as the person who first saw the flaw in Ray Kroc’s original model.