AI Bubble, explained with a map of understanding
- Samuel Fernández Lorenzo
- Dec 10
- 4 min read
The Numbers That Don't Add Up
Imagine someone spending $100,000 to set up a business that only generates $10,000 a year. Normally this would be a red flag that would make more than a few people rethink the business model. But this is, on a much more epic scale, what's happening right now in the AI sector. In 2025, global investment in AI infrastructure will reach $558 billion dollars. It's a dizzying figure. To put it in perspective, it's more than the GDP of entire countries. And do you know how much has been generated in actual revenue? $35 billion, that is, 6% of the investment.
More data. If we look back to the dotcom bubble era (when everyone thought that adding ".com" to your company name would automatically make you a millionaire, until it turned out it wouldn't), companies experienced sky-high valuations that multiplied their actual revenues dozens of times over. To give you an idea, the average at that time was 32x for software companies. And here comes another alarm bell when we look at the current valuations of some flagship companies in the sector.
Anthropic: 37 times its annual revenue.
OpenAI: 30 times its annual revenue.
Nvidia: 25 times its annual revenue.
Palantir, a big data and AI company, trades at multiples of 386 times earnings (P/E ratio)!
The ominous comparisons don't end here. There's an indicator, devised by Warren Buffett, that basically measures the total value of the U.S. stock market divided by GDP; it currently stands at 217%. This means that, according to this metric, the stock market is valued at double what it should be. The last time we saw similar numbers was... yes, you guessed it: just before the dotcom bubble burst.
But the bad omens don't end here. It turns out that when you start investigating the flow of investments within this sector, you begin to find a network of circular financing that, so to speak, is sustaining its own figures. The media provides specific examples like this:
Oracle spends billions buying chips from Nvidia.
Oracle announces a strategic alliance with OpenAI.
OpenAI receives multimillion-dollar investments from Nvidia.
Nvidia sells more graphics cards to companies like Oracle…
Is this a sub-economy that's disconnecting from the rest of the real economy? If so, it resembles a game of musical chairs, where everyone fears the financing music will stop and they'll be left out of the game.
The Logic of Bubbles
In order to more critically evaluate the explanation about the bubble, I'm going to use a map of understanding. It's not intended to be so much a personal map, but rather a map that reflects the understanding of the phenomenon as expressed in the media.
Maps of understanding are a visual instrument that I've defined in my work Everything I Can Imagine: The Algorithm of Understanding, as a way to graphically illustrate a theory about synthetic understanding, which I also introduce in the same work.
An understanding map is composed of nodes—the objects that one tries to understand. In this case, I'm going to identify one node: the financial market. This is what I want to better understand through other objects/nodes, which here will be companies in the sector: Nvidia, Palantir, OpenAI, and Anthropic.
The nodes have properties or states that we want to understand, and we'll represent them with subnodes within the map. In particular, the state of the financial market that interests us is the bubble state. AI companies also have subnodes that express their states, such as their current valuation multiples.
Now then, to understand the bubble, we must connect it through logical relationships within the map. This is where we have to critically evaluate the conditions that characterize a bubble. In the paper What are Asset Price Bubbles? A Survey on Definitions of Financial Bubbles, by Michael Heinrich Baumann and Anja Janischewski, an interesting overview of definitions and characterizations is made, which certainly coincide greatly with the warning signs we mentioned earlier. Specifically, the logic pointing toward a bubble can be distilled into the following associations
Hyperexponential growth of financial assets may imply a bubble.
Nvidia has grown to become the most valuable company in human history (reaching $5 trillion) due to the AI explosion. Its market capitalization is larger than the entire economy of Brazil or Mexico. In fact, only the economies of the United States and China are larger than the company Nvidia.
If there's a disconnect between perceived market value and actual earnings, it implies we're in a bubble.
Here we only need to look at the valuation multiples of companies cited earlier. An analogy I've heard summarizes the current AI situation well: it's like a piece of land valued at billions only because people believe that someday gold will grow on it.
If the market shows historical metrics of previous overvaluation, it could indicate we're in a bubble.
As we've seen, both indicators like Buffett's, as well as valuation multiples, are at dotcom bubble levels.
If companies are investing in their own customers to make revenues look better than they really are, it implies an unsustainable business model in the long term.
The network of circular investments like the one we pointed out earlier between Oracle, Nvidia, and OpenAI could be categorized within this scheme where valuations and revenues are artificially inflated.
We now have all the ingredients in play. Finally, the map uses three colors
Red: that which we don't yet fully understand.
Green: that which we accept as basic axioms.
Blue: that which is strictly understood, that is, relationally.
The key is in the connections: an element is considered "thoroughly understood" (blue) only when we can logically connect it with other elements we already understand (in green or blue) through an understood logical relationship (in green or blue).
For our map, we'll assume as previously understood everything except the bubble state, which is what we're trying to understand relationally. Thus, the final understanding map will be something like the following.
Final Reflection
This understanding map is not intended to be a prediction or a recommendation. It's simply a cartography of how the financial market is interpreting the AI phenomenon at this moment.
What's fascinating (and terrifying) is that we're living through this moment in real time. We're witnessing what many consider the greatest technological transformation of our generation, or the greatest speculative bubble in modern history.
Or perhaps, paradoxically, both things at once.
Comments