Video · 2026-06-08 · 1h 1m · 6 moments

The Economics of AI Usage and What's Next For SaaS | Benedict Evans on a16z

✦ AI generated

timeline · colored by role

Claim

Agentic coding achieved absolute product-market fit this year, with customers pulling it out of developers' hands, which is why the whole industry's focus has narrowed onto it while other use cases remain unproven.

Benedict Evans says agentic coding is the one AI use case with undeniable product-market fit, which is why the industry's attention has narrowed sharply onto it even as the resulting demand has caused a supply/price crunch.

transcript

Benedict Evans: clearly agentic coding started working and so all the focus in tech has kind of narrowed in massively onto that as something that has absolute product market fit in the sense that like the customers are pulling it out of your hands.

explains mechanism · 1supports · 1

Example

Mobile carriers built enormously valuable, high-growth infrastructure but captured almost none of the resulting value because it was competed away as a commodity, and AI infrastructure is likely headed the same way.

Drawing on the mobile data era, Evans notes that despite traffic growing ~2000x, telcos' margins stayed flat for 20 years and all the valuable innovation was captured by companies further up the stack — a likely preview for AI infrastructure providers.

transcript

Benedict Evans: since then mobile data traffic has risen by something like one and a half to 2,000 times and the mobile networks collectively have revenue of about a trillion dollars and they spend about 200 billion dollars a year on capex and the s[t]ocks have been flat for 20 years and all the cool stuff got built by somebody else.

Claim

The chatbot is only a weird, limited first-generation interface that works well for a narrow set of tasks and people, not a sufficient product form on its own — most use cases need dedicated tooling, configuration, and interface design built by people other than the end users.

Evans argues the chatbot is a crude v1 interface, not the final product — most real use cases require dedicated tooling, data setup, and interface design that the chatbot format alone doesn't provide.

transcript

Benedict Evans: the chatbot itself is like a kind of a weird limited v1 UI and there's some things and some people and some kind of task where it works really well but there are most of the others you need a bunch of other stuff you need tooling and it needs to be set up right. It needs to have the right data and it needs to be configured and controlled.

provides context · 1rebuts · 1

Prediction◆

Foundation models are heading toward becoming low-level commodity infrastructure with no pricing power, because there's no network effect or sustainable differentiation between models, so value will accrue to layers built on top of them rather than to the model companies themselves.

Evans lays out his central thesis: since models lack network effects or differentiation, three-to-six competing frontier labs will end up as commodity infrastructure without pricing power, similar to chipmakers or ISPs rather than Windows or iOS.

transcript

Benedict Evans: So therefore, they're low-level infrastructure. And so then, well, do they have pricing power? Well, you're going to have, pick a number, three to six companies making a Frontier model, spending no one knows, no one honest knows, like something between $200 billion and $2 trillion a year on building these models.

gives example · 1supports · 1

Example

Once AI is embedded in an industry like media or law, the important open questions stop being San Francisco technology questions and become industry-specific questions that only people from that industry can really answer — exactly as happened with Netflix, where all the meaningful decisions became Los Angeles/TV questions rather than tech questions.

Evans compares AI's rollout to his earlier 'Netflix isn't a tech company' argument: once a technology is embedded in an industry, the consequential questions become industry-specific (what shows, what legal work, what analysis) rather than technical ones that Silicon Valley can answer.

transcript

Benedict Evans: if you looked at Netflix, this whole thing is enabled by stuff the tech industry is built. But all the questions for Netflix are are TV LA questions like what shows, how many shows, what kind of shows, what should you pay the talent... These are all Los Angeles questions. These are not San Francisco questions.

Prediction◆

Big tech companies cannot sustain their current trajectory of AI capex growth — spending 1.5 trillion next year isn't feasible without massive borrowing, and there's a hard financial ceiling since there simply isn't $10 trillion a year available to spend on AI infrastructure.

Evans argues there's a hard financial ceiling on AI infrastructure spending — companies already devoting over 50% of revenue to capex can't sustain much higher growth, since the money simply doesn't exist at that scale.

transcript

Benedict Evans: Clearly, like those companies could not spend 1.5 trillion next year or if they did, they'd have to borrow it and they certainly couldn't sustain that level of spending um for any length of time. Um, and so there's a certain point at which like that growth has to slow down because like there isn't any more money.

Highlight slides

Models Are Becoming Commodity Infrastructure✦ from: Foundation models are heading toward becoming low-level commodity infrastructure with no pricing power, because there's no network effect or sustainable differentiation between models, so value will accrue to layers built on top of them rather than to the model companies themselves.Massive, Uncertain Spend Without Pricing Power✦ from: Foundation models are heading toward becoming low-level commodity infrastructure with no pricing power, because there's no network effect or sustainable differentiation between models, so value will accrue to layers built on top of them rather than to the model companies themselves.AI Capex Has a Hard Ceiling✦ from: Big tech companies cannot sustain their current trajectory of AI capex growth — spending 1.5 trillion next year isn't feasible without massive borrowing, and there's a hard financial ceiling since there simply isn't $10 trillion a year available to spend on AI infrastructure.Why Growth Must Slow✦ from: Big tech companies cannot sustain their current trajectory of AI capex growth — spending 1.5 trillion next year isn't feasible without massive borrowing, and there's a hard financial ceiling since there simply isn't $10 trillion a year available to spend on AI infrastructure.

Related episodes

Video · The Cognitive Revolution

Fable's Back, AI Engineer Recap, & SambaNova

→

Video · All-In Podcast

OpenAI vs Anthropic IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts

→

Video · The TWIML AI Podcast

Foundation Models Are Coming for Enterprise Databases [Jure Leskovec] - 768

→