Everyone is asking the same question right now: Is Artificial Intelligence a bubble? 🫧
We see the stock prices soaring, the massive capital expenditure from Big Tech, and the frenzy to buy GPUs. It feels eerily similar to the late 90s. But to understand where we are going, we have to understand what actually happened during the Dotcom crash.
Spoiler alert: The bubble wasn't about the websites. It was about the plumbing. 🛁
Here is your survival guide to the infrastructure wars, the energy crisis, and why a 19th-century economic theory might just save the tech industry.
The Lesson of Dark Fiber 🧶
In the 1990s, telecom companies spent billions laying fiber optic cables across the oceans and continents. They built infrastructure for a future that hadn't arrived yet. When the bubble popped in 2000, 90% of that cable was "Dark Fiber"—unused, unlit, and seemingly worthless. Investors lost everything. 📉
But here is the twist: That infrastructure didn't disappear.
That cheap, overbuilt bandwidth is exactly what allowed Google, Facebook, and Netflix to thrive a decade later. We are seeing the exact same pattern today, but instead of cables, it’s Data Centers and GPUs. We are building the roads before we have the cars. 🏎️
The Smart Child Fallacy 👶
We tend to think of current Generative AI as a super-genius. In reality, it’s more like the smartest child in the world who learned everything via flashcards. 🃏
Current LLMs predict the next word based on massive text training. They know that "Earth" is the "third" planet from the sun because they've seen those words together billions of times. But they don't understand the physical concept of "third" or "gravity" the way a human child does through experience.
Why does this matter for business? Because we are hitting a wall with text. To get to the next level—General Intelligence—AI needs to understand the physical world. It needs to know what the back of a gold nugget looks like without ever seeing it. 🪨✨ This requires massive amounts of compute power to simulate reality, not just predict text.
The DeepSeek Shock amp; The Efficiency Trap 📉
Recently, a relatively unknown Chinese startup, DeepSeek, released a model that rivaled top-tier US models for a fraction of the cost and compute. The market panicked. 😱
If AI models become 50x more efficient, do we really need all these multi-billion dollar data centers? Will demand for chips collapse?
Enter Jevons Paradox. 💡
In the 1800s, William Stanley Jevons noticed that as steam engines became more efficient with coal, coal consumption didn't go down—it skyrocketed. Why? Because when a resource becomes cheaper and more efficient, we find new, massive ways to use it.
If AI compute becomes 10x cheaper, we won't just use it to write emails. We will use it to:
- 🧬 Simulate protein folding for drug discovery.
- 🎬 Generate real-time Hollywood movies.
- 🌍 Model climate change solutions.
We aren't going to spend less; we are going to do more.
The Real Bottleneck: Electricity ⚡
The constraint isn't the chips anymore; it's the plug.
We are talking about building data centers the physical size of Manhattan. 🏙️ The energy demand is growing faster than at any point in human history. We are effectively trying to light the entire world with whale oil before we've fully scaled the modern power plant.
The companies that win the next decade won't just be the ones with the best algorithms; they will be the ones who can secure Gigawatts of power without crashing the grid. 🔌
The Verdict? 🌱
Are we in a bubble? Probably. Many AI startups will vanish like Pets.com. 🐕
But the "forest" is growing. The infrastructure being built today—the energy grids, the data centers, the efficient chips—is the soil for the next twenty years of innovation.
Don't bet against the forest just because a few saplings wither. 🌲🚀
Discussion 0 comments