Three Bubbles, Three Lessons for Investors and Operators
Not all bubbles are created equal. The Tulip Bubble was a 17th-century Dutch mania where rare tulip bulbs traded for prices higher than houses. When confidence collapsed in 1637, values crashed overnight, wiping out fortunes and cementing its place as one of history’s first recorded asset bubbles.
In more recent times, we’ve seen three technology-driven bubbles—each offering a different lesson about the interplay of capital, innovation, and technology adoption.
The Internet Bubble (1990s): Infrastructure Without Demand. The dot-com era poured trillions into laying fiber, building portals, and chasing “eyeballs.” Stamp your slide deck with “Internet Business” and you increase the probability of getting funded. Most dot-com companies failed because they lacked a repeatable model of customer demand. But the capital wasn’t wasted: the infrastructure—broadband, browsers, e-commerce rails—enabled the platforms that defined the next two decades. The winners emerged long after the hype-cycle ended.
The Era of Abundant Capital (2009–2021): Demand Without Sustainable Economics. A period when cheap capital, digital platforms, and global optimism converged to fuel an unprecedented boom in innovation, valuations, and faith in exponential growth. The government stimulus during COVID (2020–2021) supercharged the economy and markets. Startup, VC and PE growth was fueled by ZIRP and low operating costs from open source and the cloud, enabling rapid scaling without sustainable economics. When rates rose, valuations collapsed, revealing a platform economy built on demand but lacking pricing discipline.
Today’s AI Bubble: Real Demand, Real Utility, but Real Constraints. The AI boom began quietly around 2017, accelerated between 2020 and 2022, and erupted into mainstream adoption in 2023. Unlike past speculative cycles, this surge is anchored in genuine utility—copilots, automation, and productivity gains across industries. Demand is strong, but growth is constrained by capital intensity, data center bottlenecks, and geopolitical tensions. While valuations remain inflated, success will hinge on execution, efficiency, and talent, not hype. The boom has also triggered a talent crunch and record pay for technical roles, even as the tech market overflows with undifferentiated startups. With IPOs scarce and M&A highly selective, only those with true defensibility, scale discipline, and measurable traction are likely to endure.
The challenges are both unique and systemic:
Capital intensity: Training and deploying frontier models requires billions in GPUs, datacenters, and talent. This sharply concentrates power into a handful of players.
Geopolitical overlay: AI is now a national security priority. U.S.–China competition and export controls make this cycle unlike any prior one.
Unproven ROI in the Enterprise: High costs, scarce talent, and uncertain returns give many CFOs reason to resist the AI wave and slow adoption.
Hype Cycle Saturation: While the technology is real, inflated valuations and an overcrowded startup ecosystem mean many AI players won’t survive the shakeout.
For investors and operators, the question isn’t if AI will endure—it will. The question is who survives the arms race and how value will be captured across the stack: models, infrastructure, applications, and distribution channels.
Just as the Internet bubble left behind infrastructure, and the post-Obama bubble left behind scaled platforms and open source software, the AI bubble will leave behind durable AI infrastructure, applications and usage patterns. But the journey there will be volatile, capital-intensive, and brutally competitive.