Jul 15, 2026 · 23m · another-podcast
Token pricing
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Benedict Evans and Tony Karen Brown analyze the future economics of artificial intelligence token pricing, assessing whether foundation model creators can establish durable competitive moats or will succumb to the commoditization seen in previous technological infrastructure cycles.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 94.8% of the talking time here. How this is scored →
The hosts as informed peer 8.2 Guest teaching 0.2 Guest disagreement 0.7 The hosts pushing back 1.2
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 14:10 Challenging model development costs
Tony briefly interrupts Benedict's point about Big Tech building new models by pointing out that doing so was expensive.
Hardest push from the hosts ▶ 14:12 Dismissing high cost barrier framingBenedict swiftly counters Tony's remark, clarifying that the capital expenditure was not trillions and is readily affordable for major tech giants.
Biggest teaching moment ▶ 11:17 Psychological framing of pattern huntingTony shifts the discussion from technical analogies to the psychological need for humans to hunt for recognizable patterns to feel in control.
The host holds their own ▶ 7:15 Detailed economic distinction between telecom and fiberBenedict showcases his structural analysis by explaining how fiber costs are 80% fixed trench digging whereas mobile networks incur real marginal infrastructure costs.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Market Disequilibrium and S-Curve Uncertainty | 8 | 0 | 0 | 0 | Benedict delivers an extended analytical framing on current token pricing disequilibrium, comparing the current early S-curve uncertainty to the dot-com era in 1997. Tony simply summarizes the approach. | |
| Frontier Model Dynamics and Infrastructure Economics | 8 | 0 | 0 | 0 | Benedict deconstructs the frontier model landscape, price-performance curves, and hyperscaler analogies to cloud infrastructure without pushback from Tony. | |
| Historical Parallels: Fiber, Telecom, and Semiconductors | 9 | 0 | 0 | 0 | Benedict demonstrates deep domain knowledge comparing fixed-cost fiber overbuild to marginal-cost mobile telecom and Rock's Law in semiconductor fabrication. Tony asks a brief clarifying question. | |
| Hunting for Analogies and Analyzing Market Power | 8 | 1 | 2 | 3 | Tony prompts why people hunt for historical analogies, then interjects that foundation models were expensive for Musk and Zuckerberg. Benedict immediately pushes back, noting that for Big Tech it was well within budget and not prohibitive. | |
| The Mechanics of Strategic Moats and Network Effects | 8 | 0 | 1 | 2 | Benedict breaks down the definition of power and network effects across Windows, iOS, and social platforms, refining Tony's comment to emphasize that execution must precede network effect moats. | |
| Enterprise Realities and Commodity Infrastructure | 8 | 0 | 1 | 2 | Benedict addresses the limits of bottom-up enterprise adoption and commoditized infrastructure. When Tony offers a soothing takeaway on slow adoption, Benedict gently categorizes it as tangential before concluding with his thesis. |
Statements from this episode (18)
Evans: Agentic Coding Has Driven Token Demand Up Multiple Orders of Magnitude
“And as hopefully everyone listening to this will understand, like in the last six months or so, the sudden product market fit of agentic coding means that demand for tokens has gone up by many orders of magnitude, and the model labs have been sort of scramblin…”
Evans: AI Model Inference Alone Generates 40% to 50% Gross Margins
“And meanwhile, we sort of know that you have positive gross margins on inference alone of sort of 40, 50%, but you've got the cost of building the next model, and you don't know where the cost The cost will move, or the cost of the next model will move, and yo…”
Evans: Early Tech S-Curves Make Scale Obvious but Mechanics Unknowable
“There is a stage in the S-curve in the development of a new, big new technology Where, like, at the early part of the curve, like, it's suddenly very obvious that this is going to be huge, and everyone's very excited, but you don't know how any of it's going t…”
Evans: Anthropic Leads AI Frontier, OpenAI and Gemini Trail by 5%
“Because right now we've got, you know, Anthropics ahead at the moment. We have OpenAI and Gemini five percent behind, like not far behind. Grok just, SpaceX Grok Nazi model just came out with like a model that's in like the top At the top of some of the benchm…”
Evans: Applications Captured Most Cloud Value, Not Hyperscalers
“Because that's kind of what happened with cloud. For example, there's only three hyperscalers, but, and yet like most of the applications are built by other people and most of the value is captured by people who run on AWS.”
Evans: AI infrastructure is built behind demand, unlike dot-com fiber
“The narrow problem with that, of course, is that that was built out ahead of demand, whereas this is being built out behind demand.”
Evans: Mobile networks are a better analogy for AI than fiber
“So the more interesting comparison or the more relevant comparison is mobile because mobile does have marginal cost. Like if your mobile traffic doubles, you kind of have to go out and build, not exactly double, but kind of double the infrastructure.”
Evans: Telco stocks were flat for 20 years while apps captured value
“Giant industry, trillion dollars of revenue, two hundred billion dollars a year of CapEx stocks have gone nowhere in 20 years, and all the value is built by other people.”
Evans: DeepSeek Proves Anyone With Two Billion Can Build Foundation Models
“Like anyone who's got a couple of billion dollars can make a foundation model which is what DeepSeq showed us as well.”
Evans: Sam Altman's Windows and utility analogies for AI are flawed
“Sam Altman talked about windows, which is a really bad analogy because windows had network effects. He also talked about water companies and electricity companies, which are like completely the opposite. They're regulated utilities with crappy margins selling …”
Evans: AI foundation models become commodities if frontier progress slows
“If the frontier slows down, then this stuff is definitely a commodity.”
Evans: AI Model Makers Lack Value Capture Due to Identical Technologies
“Because right now there isn't a path to value capture because right now you've got a lot of people doing basically the same thing with basically the same technology and the same data and the same chips.”
Evans: Platform power is the ability to force people to do what they avoid
“Power isn't sophistication or complexity or doing clever things or being impressive. Power is the ability to make people do something that they don't want to do.”
Evans: Google Search Network Effects Will Defeat Microsoft's Spending
“Google search has network effects, so Microsoft can spend, doesn't matter how much money Microsoft spends, so Google search is still better.”
Evans: OpenAI's new ChatGPT work product is a confusing disaster
“I was playing with a new chat GPT work product, which is clearly not that it's a complete disaster. It's chaotic, confusing mess”
Evans: Bottom-up enterprise adoption only ever succeeded for Slack and Notion
“This is like the great fallacy of enterprise software that you can just sell to the users, and it's like it only ever worked for Slack and one other company. Maybe Notion, but even Notion has capped out.”
Evans: LLMs lack broad product-market fit beyond software developers
“The LLM itself Is not a great product for most people. Usage is a mile wide and an inch deep, and you have this kind of polarization between people where this really, really works, and they really, really have product market fit, which is basically software de…”
Evans: Model makers will lose pricing power within five years
“And the paradox is like right now they can name their price, but that isn't where we're going to be in five years. You can argue about how quickly the infrastructure gets built out and how fast the GPUs arrive, blah, blah, blah, blah. Fine. But that's a supply…”