Jun 8, 2026 · 1h 0m · a16z
The Economics of AI Usage and What's Next For SaaS | Benedict Evans on a16z
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this interview on The a16z Show, tech analyst Benedict Evans and host Erik Torenberg explore the evolving AI ecosystem, analyzing foundation model commoditization, the disruption of SaaS and software engineering, and the economic frameworks governing enterprise AI adoption.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 7.7% of the talking time here. How this is scored →
The host as informed peer 3.1 Guest teaching 4.8 Guest disagreement 2.1 The host pushing back 1.0
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 5:20 Guest Dismisses Career Structure Predictions
Benedict directly rejects speculation regarding AI's impact on engineering org charts, claiming anyone asserting they know engineering team structures in three years would be insane.
Hardest push from the host ▶ 48:29 Host Challenges CapEx Risk LogicSteph/Erik cites Google's CEO regarding under-investment risks and explicitly presses the guest on whether current CapEx levels are reaching an unsustainable tipping point.
Biggest teaching moment ▶ 11:20 Telecom Infrastructure Trap AnalogyBenedict educates the host on historical infrastructure economics, showing how mobile network operators spent hundreds of billions on CapEx while all economic value accrued to application layer players.
The host holds their own ▶ 14:55 Host Delineates SaaS vs Hardware MarginsThe host demonstrates deep market knowledge by contrasting traditional SaaS application layer margins with cloud infrastructure and hardware value capture, specifically referencing Nvidia.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Reflecting on 'AI Eats the World' and Recent Industry Shifts | 2 | 3 | 1 | 0 | Host opens by asking the guest to reflect on what has changed in AI over the past year. Guest outlines key shifts, focusing on agentic coding product-market fit and infrastructure capacity constraints. | |
| Agentic Coding and Software Engineering Dynamics | 3 | 5 | 3 | 1 | Host asks if the coding explosion was foreseeable and how it reshapes engineering orgs. Guest explains why software developers naturally built software tools first and firmly dismisses early predictions about future engineering careers as premature. | |
| Model Lab Strategies and Enterprise Point Solutions | 2 | 4 | 2 | 0 | Host inquires about OpenAI's strategy shifts. Guest contrasts OpenAI's broad product surface area with Anthropic's focused strategy on coding and highlights enterprise point solutions. | |
| Historical Tech Parallels and Infrastructure Economics | 3 | 6 | 2 | 1 | Host asks how early AI adoption compares to mobile platform shifts. Guest delivers an extended historical analysis comparing current token pricing crunches to 2009 mobile data bottlenecks and telco infrastructure value accrual. | |
| Value Accrual Across the AI Tech Stack | 5 | 4 | 3 | 2 | Host offers a structured framework contrasting SaaS app margins with hardware/cloud value accrual. Guest counters that historical parallels show possibilities rather than predictive outcomes during early technology cycles. | |
| Why Foundation Models May Become Commodities | 3 | 6 | 2 | 1 | Host prompts guest to explain his thesis that foundation models are not standalone products. Guest details four core structural arguments, including lack of network effects, UI limits, and commodity pricing wars. | |
| On-Device Intelligence and Domain-Specific Questions | 2 | 5 | 1 | 0 | Host asks what emerging questions Benedict is most focused on next. Guest highlights on-device execution, domain-specific restructuring in law/finance, and fundamental physical uncertainties. | |
| Automation Frameworks and Economic Elasticity | 2 | 5 | 2 | 0 | Host asks which non-coding use cases could yield daily active engagement. Guest presents economic automation frameworks involving price elasticity, Jevons paradox, and unlocking previously cost-prohibitive tasks. | |
| AI Impact on Advertising, E-Commerce, and Retail | 3 | 5 | 2 | 1 | Guest discusses how AI transforms retail and advertising by understanding product semantics rather than simple metadata. Host asks about rebuilding legacy platforms, prompting guest to emphasize novel applications over cloning old software. | |
| Rethinking SaaS, Workflows, and Corporate Adoption | 4 | 5 | 2 | 2 | Host asks if AI will de-consolidate the SaaS landscape. Guest outlines the three enterprise software tiers (big iron, vertical apps, informal spreadsheets) and analyzes how probabilistic tools fit into current workflows. | |
| AI-Native Interfaces and Task Deconstruction | 4 | 4 | 2 | 1 | Host asks whether AI-native software will discard traditional front-end UIs in favor of direct agent queries. Guest argues that software value centers on exception handling and distinguishing tasks from overall jobs. | |
| AI CapEx Limits, ROI Measurement, and Productivity | 5 | 5 | 2 | 2 | Host cites Big Tech executive statements on under-investing risks and asks if a CapEx limit or token ROI reckoning is near. Guest analyzes financial limits, comparing AI infrastructure spending to global telecom and energy sectors. | |
| Foundation Model Strategy and the Long-Term Arc of AI | 3 | 5 | 3 | 2 | Host asks how foundation model labs should adapt given massive fundraising alongside commoditization risks. Guest clarifies his thesis as an analytical challenge to prove non-commoditization and reflects on long-term technology adoption arcs. |
Statements from this episode (27)
Evans: Tech industry cannot spend $10 trillion annually on AI infrastructure
“We can't spend 10 trillion dollars a year on our AI infrastructure, because there isn't 10 trillion dollars a year there to spend on it.”
Evans: Chatbots and foundation models are not standalone products
“I don't think foundation models. A product. I don't think a chatbot is a product. I think the value will be further up.”
Evans: Current generative AI technology cannot drive daily consumer usage
“We don't see a way that consumers will use this daily rather than weekly with the technology we have right now.”
Evans: AI coding's impact on engineering teams will take years to settle
“It's going to take a couple of years for this all to settle down, you know, if nothing else because of the pricing, you know, you've got this enormous crunch between the demand and supply and hence the pricing.”
Evans: Future structure of software engineering careers remains unpredictable
“I don't think anybody can possibly say they kind of know what the market structure is going to look like or what the career of a software engineer is going to be in three years time. I think it would be, you'd be insane to think that you could know that yet.”
Evans: Anthropic succeeded in AI coding by focusing strategy with less capital
“And then Anthropic, with having less capital raised, said, no, we're going to focus on coding, and they got coding working. Whether that was, like, a deliberate strategy or, kind of, they stumbled into it is, you know, for other people to say, but, like, clear…”
Evans: Enterprise AI adoption succeeds through specific back-office automation
“And there's a lot of places where corporations are using it to automate some like specific back office process where you're not asking the user to work out what they do with the new tool. Instead, you're saying, okay, here's a problem that we can solve.”
Evans: Telecom stocks stayed flat despite 2,000x mobile data growth
“Mobile data traffic has risen by something like one and a half to 2000 times, and the mobile networks collectively have revenue of about a trillion dollars, and they spend about two hundred billion dollars a year on CapEx, and the stocks have been flat for 20 …”
Evans: Tech industry will spend $1T to $2T on AI CapEx
“Over the next couple of years, we've got like a trillion or two trillion dollars of capex coming down the pipe, and the models get a hundred x, 200 x, if it's more efficient every year.”
Evans: Infrastructure providers historically fail to capture value compared to OS
“Chip companies didn't capture the value. ISPs didn't capture the value. Mobile network operators didn't capture the value. Windows and iOS did, but they were doing something else.”
Evans: Analysts should focus on unpredictable areas rather than solved tech
“One of the characteristics of tech is that the moment that you understand something and you know how it works and what's going to happen, Is the moment you should move on to something else, you should always be looking for the quiet places where we don't know …”
Evans: Foundation AI models are commodities and low-level infrastructure
“The models are kind of diff commodities, and the chatbot isn't the right UI or the right product, and the companies aren't going to be able to build all of that stuff themselves, so therefore they're low-level infrastructure.”
Evans: Only 3 to 6 companies will build frontier AI models
“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 two hundred billion dollars and two trillion dollars a year on building these models.”
Evans: Current AI compute and token scarcity is transitory
“This situation right now is transitory. You know, we're in this extreme scarcity, and then we have a pricing system, and we have a free market, and we have a surge of capex, and like a trillion dollars of capex.”
Evans: AI differs from past tech shifts because physical limits are unknown
“The way that all of this is sort of fundamentally different from previous platform shifts, is that with, you know, three G or the iPhone or the web or whatever it was, You didn't know what was going to happen next, but you knew the physical limits. Like, you k…”
Evans: Coding is the only AI use case with clear product-market fit
“The place that's got product market fit right now is coding. Nothing else has equivalent product market fit right now.”
Evans: Internet distribution shift devastated newspapers but left movie studios unchanged
“Like, if we'd been back in the late nineties and we'd said, you know, internet will destroy the value of physical distribution, it turned out that meant completely different things for newspapers and movie studios. Like, newspapers got completely screwed by th…”
Evans: Global advertising is a $1T market while retail is $25T
“Advertising is a trillion dollars and retail is 25 trillion dollars.”
Evans: AI integration is driving ad revenue gains at Google and Meta
“Which is, of course, why you see the ad numbers and the, you know, the conversion rates shooting up in the, every quarter from Google and Facebook. Because they're rolling all of this into their ad systems, and their recommendation engines, and their predictio…”
Evans: Breakthrough tech adoption creates new paradigms rather than simple automation
“Whenever you get a new technology, you start by doing the old thing, but more. More spreadsheets. More PowerPoints. More email. Better email. But the important stuff is not doing the old thing, but more. It's doing something new that you couldn't have done wit…”
Evans: Typical large US enterprise runs over 1,300 software applications
“Vertical software and typical big US company has, like, three to 400 test apps and then, like, another thousand apps that they've built, bought, or built themselves internally running on-prem.”
Evans: AI will drive 10x to 100x increase in total software creation
“What does this do to software, and the answer is, more software, like, way more software. I mean, all software companies exist to solve problems created by other software companies, and that was the joke in security, like, All security software exists to solve…”
Evans: LLMs will master standardized tasks but struggle with novel reasoning
“LLMs are going to be very good at anything where you can describe how people do it. And where what you want is the way anybody would do that, and not so good at where you can't really explain why you did it like that, and where you're doing it differently to t…”
Evans: Microsoft, Meta, and Google spending over 50% of revenue on CapEx
“Microsoft, Meta, and Google are all on, in line to spend over 50% of revenue on CapEx this year.”
Evans: Enterprise AI productivity gains will be competed away
“These things become competitive necessities, and everybody has to buy it and use it. But the cost saving or the productivity gain that you get from it just kind of gets competed away, so you don't get to charge more for it.”