You feel it, don’t you? That tightness in your chest when you see another billion-dollar funding round for an AI startup. It’s the intoxicating cocktail of FOMO and dread. You want to ride the wave of the next great technological revolution, but a quiet voice in the back of your head keeps whispering the same two words: dotcom bubble.
You aren’t crazy. The anxiety is entirely justified. We are currently watching valuations detach from reality at a speed that makes 1999 look conservative. But to call this an “AI bubble” is to miss the most dangerous part of the story.
We are not in an AI bubble; we are in an AI-lab bubble.
Let’s be clear about one thing right now: Large Language Models (LLMs) are not a fad. The technology is transformative. It works, it scales, and it is going to fundamentally rewrite how we interact with software. The tech is here to stay.
The problem isn’t the code. The problem is the business models of the companies writing it.
Look at the high-profile AI labs. They are raising capital at valuations that require them to not just dominate the software market, but to essentially consume the entire global economy. Yet, what is their actual moat? What do they own that prevents someone else from doing the exact same thing for a fraction of the cost?
The answer is terrifying: nothing.
The technology will eat the world. The companies building it will likely be eaten by it.
The dotcom era taught us a harsh lesson. The internet was real. Webvan and Pets.com were not. The infrastructure survived; the over-leveraged front-runners went to zero. We are repeating history, but with a terrifying twist. In the AI era, the core product itself has zero marginal cost to replicate.
Enter the cheap, open-source models flooding out of China and decentralized developer communities. When a competitor can clone your billion-dollar model’s performance in a few weeks for a few thousand dollars, you don’t have a monopoly. You have a science experiment.
When your billion-dollar moat can be cloned for a fraction of a cent, you aren’t a tech monopoly. You’re a charity.
Sam Altman and his peers are asking you to bet on a future where they achieve AGI and solve all of humanity’s problems. It’s a beautiful story. It’s also the perfect cover for the fact that their current revenue streams can’t justify their astronomical burn rates.
So, who actually wins the AI war? The same people who always win during a gold rush: the ones selling the shovels. The real value in this era isn’t going to accrue to the flashy AI labs burning cash to give away free chatbots. It will accrue to the boring infrastructure providers—the chipmakers, the cloud hosting platforms, and the commodity model providers who drive the price of intelligence down to zero.
The revolution is real. The valuations are a lie. Don’t let the hype blind you to the math, and whatever you do, don’t be the last one holding the bag when the music stops.
FAQ
Q: Isn't OpenAI's brand and first-mover advantage enough to protect them?
A: No. Brand loyalty means nothing when a free, open-source model offers 95% of the capability at 0% of the cost. First-mover advantage is a myth in software when the underlying tech can be reverse-engineered or replicated almost instantly.
Q: So should I stop investing in AI startups entirely?
A: Stop investing in AI startups that are just thin wrappers around existing LLMs. Invest in companies using AI to solve deep, unsexy, industry-specific problems where the data itself is the moat, not the model.
Q: Who actually makes money in this market then?
A: The commodity providers. The companies driving the cost of intelligence to zero will commoditize the AI labs out of business. The real money is in hardware, cloud infrastructure, and the boring enterprise software that actually integrates this tech into daily workflows.