Introduction

Everyone says we are in an AI bubble, but Zvi asks what that really means and shows why the truth is more mixed than headlines.

What's the problem it solves?

People keep saying AI is a bubble without agreeing on what a bubble is. This piece explains the word, lays out the signs, and shows why the answer is not simple.

Quick Summary

Zvi starts by asking what we mean by bubble. If bubble means any big drop in prices, that can happen and does not prove the tech is fake. If bubble means prices that make no sense vs likely future cash, he says that is not what we see in AI today. He notes many smart people are yelling bubble because deals feel circular, costs are huge, and profits are not clear yet.

He then looks at both sides. On the risk side, some AI firms will get crushed by bigger labs. Hype can run ahead of results. Geopolitics, tariffs, or supply shocks could hit. A scare can trigger a fast drop even if nothing real changed. On the strength side, AI revenue is growing fast, core chips and data centers are still scarce, and overall market valuations are high but not wild. The big tech spend is large, but may be worth it if AI keeps adding value. Even if prices fall, that would not mean AI failed. It might just mean hopes were too high for a while.

The key idea is that bubbles are about value vs expectations. If AI grows slower than hopeful plans, prices can sink. If it grows faster, prices can rise more. Today looks less like dot com toys and more like a heavy buildout that takes time and money. Zvi ends by saying a 20 percent drop over months is very possible, yet he would likely buy more if the long term story stays intact.

Key Takeaways

• Bubble talk rose fast, but the word bubble is fuzzy and often misused
• A market dip alone does not prove a fake boom
• Some AI bets will lose, but picks and shovels may still win
• AI demand and revenue growth look strong vs history, though not guaranteed
• Big capex can scare people, yet can be smart if unit value is high
• Valuations are elevated but not near 2000 levels
• A sharp drop could come from mood, policy, or supply issues, not just tech failure
• Long term value depends on profits, not only cool tech

What to do

• Define bubble before you debate it
• Separate the tech trend from stock price swings
• Stress test your AI plans for slower growth, higher costs, and supply shocks
• Favor clear value paths and unit economics over hype
• Be wary of apps that big labs can steamroll
• Look for durable picks and shovels across chips, power, tools, and infra
• Keep cash and rules for buying if a drawdown hits
• Track real adoption and revenue, not just headlines

‍