The AI Bubble Myth: Why Everyone Is Getting It Wrong w/ Jordi Visser

Quick Overview

Jordi Visser argues that the current AI boom is fundamentally different from past speculative bubbles like the Dot-Com era because the massive capital expenditures are tied to tangible compute and power needs, meaning the underlying investment is not purely speculative, although he acknowledges that many current AI companies are overvalued and face a high risk of failure.

Key Points: AI capital expenditure is on track for over $400 billion in 2024 and $500 billion in 2025, significantly higher than the Dot-Com bubble's peak. The core driver of AI CapEx is compute and power, encompassing data centers, cooling systems, semiconductors, and transformers, unlike the pure speculation of the Dot-Com era. Visser predicts that over 50% of the companies spending heavily on AI today will fail, citing high debt levels and negative free cash flow yields among many. He compares the current situation to the Dot-Com bubble, noting that while many companies failed, the underlying technology (the internet) proved transformative, suggesting AI will be similarly disruptive. The concentration of wealth is extreme: the top 1% own one-third of assets, while the bottom 50% own very little, contributing to the K-shaped economy argument. Visser suggests that while the term 'bubble' is often used as clickbait, the reality is a massive, concentrated investment cycle where many participants will fail, but the core technology will win.

Context: The discussion takes place between John Gillen and guest Jordi Visser, who runs Visser Labs and 22V Research. They analyze the massive capital expenditures currently flowing into Artificial Intelligence (AI) infrastructure and debate whether this spending spree constitutes an economic bubble comparable to the Dot-Com era, focusing on the tangible components driving the investment.

Detailed Analysis

Jordi Visser asserts that the current AI investment cycle is not a traditional bubble because the spending ($400B+ in 2024, projected $500B+ in 2025) is directed toward real infrastructure needs: compute power, data centers, cooling systems, semiconductors, and transformers. This is a fundamental difference from the Dot-Com bubble, where investments were often speculative without immediate underlying physical infrastructure needs. Visser, however, predicts a massive shakeout, estimating that over 50% of the companies currently spending heavily on AI will fail within five years because they lack revenue, are overvalued, and carry significant debt. He agrees with Mark Zuckerberg's sentiment that companies are willing to spend billions on AI for a chance to win big, but notes that this creates a race to obsolescence where many will lose. Visser points out that the concentration of wealth—the top 1% owning one-third of assets while the bottom 50% own almost nothing—is fueling a K-shaped economy, and the current AI spending dynamic is an output of this structure, not necessarily a sign of a broad market bubble, though individual AI companies are certainly vulnerable.