Episode summary
We Asked the Man Who Mapped the AI Economy If the Boom Is Real — And Who Keeps the Money
Excess Returns
· · 1 hr 15 min · Hosts: Kai Wu · Guests: Azeem Azhar
Summary
Azeem Azhar argues that the AI boom is backed by a real demand base, not just hype, and that his bottom-up model points to about $110 billion of generative AI spending over the prior 12 months ending June 2026, or $175 billion annualized. He says growth remained as fast in summer 2026 as in December 2025, which surprised him because he had expected demand to slow.
He breaks the AI economy into chips, hosting, foundation models, and applications, and argues that pricing power depends less on stack position than on competition and bottlenecks. In his view, hosting and memory are critical constraints, while value is starting to accrue more to foundation model labs like OpenAI and Anthropic as well as to applications. But he also thinks open source, vertical integration, and infrastructure competition could squeeze margins over time, leaving consumers with much of the surplus.
On enterprise adoption, Azhar says most firms are still at the “light bulb” or “motor” stage of AI, not the full workflow redesign needed for productivity gains. He expects investors to look for evidence in earnings-call language, hiring shifts, ramped AI spend, and whether CEOs are talking about revenue rather than just cost savings. The durable winners, he says, will be the companies that use AI to redesign workflows, strengthen judgment, and convert proprietary data and human know-how into a real advantage.
Key takeaways
- Azhar estimates $110 billion of generative AI spending in the last 12 months to June 2026, annualizing to $175 billion, and says growth has not meaningfully slowed.
- He thinks hosting and memory are the clearest infrastructure bottlenecks, with memory becoming the tightest constraint in five- to ten-year demand models.
- Azhar says open source creates pricing and business-model pressure on proprietary model labs, but does not yet close the market to them because AI spend is still early.
- He argues most enterprises are only at the “light bulb” or “motor” stage, so real productivity requires workflow redesign, not just more copilots.
- For investors, he recommends tracking quantified AI claims on earnings calls, hiring shifts, and whether spending is moving from cost savings toward revenue and new markets.
Notable quotes
“our model our estimates came in at uh for to the year to the June 2026 $110 billion uh for the previous 12 months exly China being spent on generative AI”
“the amount of value being captured by the foundation labs specifically anthropic and open AI has started to uh increase proportionately”
“you can't go from co-pilots to an AI native firm by just adding more co-pilot licenses, right”
“most companies are at stage one or stage two. um and the advanced ones at that at that stage too”
How the episode opens
Over the past few episodes, a central topic of discussion has been the multi-t trillion dollar AI investment boom. Massive capital spending has helped propel markets higher while leaving stock indices increasingly concentrated in companies tied to the AI buildout. Some of the biggest questions facing investors today concern the sustainability of AI demand, the pace of technological progress, and where importantly economic value will ultimately acrew across the AI stack.
To help answer these questions, I'm glad to introduce my friend Azimar, a technologist, investor, author, and entrepreneur who has been thinking deeply about the impact of frontier technologies for many years. Azim is the author of The Exponential Age: How Accelerating Technology is Transforming Business, Politics, and Society, and the founder of the excellent exponential view newsletter. Azim, thanks for coming and welcome to the intangible economy.
I'm so excited to have this conversation with you, Kai. I've learned so much from your essays uh over the last couple of years, so looking forward to it. — Me, too. All right, so let's start with your report on the state of the AI economy, which you just released a couple weeks ago. Congratulations, by the way. It was excellent.
Um, so the the idea is this that the supply side of the AI boom is is relatively visible. We can observe chip sales, capital spending, and and data center construction, but really the demand side is is much harder to see. you attempted to reconstruct that demand from the bottoms up. Um, what did you find?
Well, we found that uh there was a real economy uh uh under underneath all of this underneath all of those uh orders that are improving Jensen's collection of leather jackets. Uh is is a company's spending real money.
From the episode’s captions: roughly the first two minutes, lightly cleaned. Keep watching on YouTube →
Topics
- ai economy
- ai stack layers
- ai infrastructure bottlenecks
- open source vs proprietary models
- enterprise ai productivity
Watch the episode
Summary, takeaways and quote selection by Octagon. Quotes are short excerpts; the full conversation is in the episode.
More from Excess Returns
-
We Asked Fidelity’s Director of Global Macro Who Wins If AI Spending Slows -
An AI Bubble Could Hit Stocks. Inflation Could Hit Bonds. How Do You Diversify and Keep Both? -
Tech Is Priced for a Boom. 86% of the Economy Grew 0.7%. What If AI Slows? -
Tech Is the Cheapest Since 2022. Walmart Is 46 Times Earnings. Where Is the Bubble? -
This $1.5 Trillion Manager Expects More Fed Hikes. So Why Is He Still Bullish? -
AI Stocks Cracked. Margins Could Be Peaking. So Why Is a Liquidity Expert Watching the Debt Wall?
Excess Returns, this episode and its artwork belong to Excess Returns. Octagon is not affiliated with the show; this page is an independent summary.
