The Eternal Return of Tech Mania
Every major technology wave follows the same arc: breakthrough, bubble, bust, then boring productivity. We're watching it happen with AI right now, and the script is so familiar it hurts.
The thing about bubbles is they're simultaneously right and wrong. The dot-com investors who proclaimed the internet would change everything were correct. They were just early, overleveraged, and backing the wrong horses. Today's AI enthusiasts making similar proclamations will probably meet the same fate. Not because AI won't transform everything, but because they're confusing proximity to the future with understanding it.
The Pattern Recognition Problem
I've been thinking about this pattern recognition problem. In 1999, Harvard MBA students were dropping out to join startups. In 2017, everyone was launching an ICO. In 2021, celebrities were shilling NFTs. Now in 2024, every startup is "AI-powered" and VCs are writing checks to anything that mentions GPT-4. The specifics change but the psychology remains constant: the fear of missing the next platform shift overwhelms rational analysis.
What's particularly interesting about the current AI boom is how it differs from previous bubbles. Unlike Pets.com or most NFT projects, AI demonstrably works. You can use ChatGPT right now to write code, analyze documents, or automate workflows. This isn't vaporware. NVIDIA's revenue growth isn't based on promises but on actual demand for compute. The fundamentals are stronger.
When Real Utility Enables Fake Value
Yet this makes the bubble dynamics more dangerous, not less. When technology has real utility, it becomes harder to separate genuine value creation from speculation. Every bubble contains a kernel of truth wrapped in layers of noise. With AI, that kernel is larger than usual, which paradoxically enables more noise to accumulate around it.
Consider the venture capital dynamics at play. A typical AI startup today takes an existing foundation model, adds a thin wrapper, targets a specific vertical, and raises millions at an inflated valuation. This isn't necessarily bad. Many successful companies are essentially productivity arbitrage: they take a general-purpose technology and make it useful for a specific context. The problem emerges when hundreds of startups do exactly the same thing, all chasing the same markets, all claiming revolutionary breakthroughs.
The Expectation Curve vs. The Reality Curve
The taxonomy of bubbles reveals another pattern. Technology adoption follows an S-curve, but human expectations follow a different trajectory entirely. We overestimate change in the short term and underestimate it in the long term. Right now, we're in the overestimation phase for AI. People expect AGI next year and full automation by 2030. When these expectations aren't met, disillusionment will follow.
But here's where it gets interesting. The post-bubble phase is often when real value gets created. After the dot-com crash, the companies that survived or emerged weren't the ones with the best Super Bowl ads. They were the ones that figured out sustainable business models. Amazon survived by ruthlessly focusing on customer experience and logistics. Google emerged by solving search better than anyone else. The noise died down, and signal emerged.
The Anatomy of Post-Bubble Winners
For AI, the post-bubble winners will likely share certain characteristics. They'll own their own models or have genuine technical differentiation. They'll solve specific, valuable problems rather than chasing broad platform plays. Most importantly, they'll have what I call "compound moats": defensibility that grows stronger over time through data accumulation, network effects, or workflow lock-in.
The smartest approach for builders right now is counterintuitive: ignore the hype. Not the technology, but the hype. Focus on building something people want, to use the YC motto. The best AI companies being built today are probably the ones you haven't heard of, quietly solving boring problems in unfashionable industries.
The Investor's Dilemma
For investors, the calculus is different. Some exposure to the AI wave makes sense, but the highest returns will come from identifying the signal among the noise. Look for founders who would be building their company even without the AI angle. Look for products where AI is a means, not an end.
History suggests this bubble will pop within a few years. When it does, AI won't disappear any more than the internet disappeared after 2000. Instead, it will become infrastructure: boring, essential, and ubiquitous. The companies that survive will be those that used the bubble's capital to build something lasting.
Timing Is Everything (And Nothing)
The meta-lesson is about timing. Being early is often indistinguishable from being wrong. Being late means missing the opportunity entirely. The sweet spot is being right about the direction but patient about the timeline. In AI, we're still early in absolute terms but late in bubble terms. Plan accordingly.
The future is already here, as William Gibson said. It's just not evenly distributed. With AI, it's not evenly understood either. And in that gap between distribution and understanding, between hype and reality, lies both opportunity and danger.
As witnessed the .com burst, i am kinda strong in my beliefs that this AI bubble will also burst like back then, as the author said.
"The bubble will burst but AI is here to stay"
Great take away. Loved it! 🤘🏻