The valuations of the AI ​​giants are sustained because we want to believe in them

In the wonderful The Big Shortthe characters discover that the mortgage market subprime It’s a kind of house of cards. The data doesn’t add up and the valuations don’t make sense, but the system still works because everyone pretends it works. Until it stops working.

That dynamic summarizes very well what is happening in the technology sector.

OpenAI follows palming money for each consultation What do we do to ChatGPT but it’s already worth half a billion. AI startups increase their valuations tenfold even though they have no real recurring revenue. The funds They continue opening fat rounds for what they continue to be wrappers of AI whose only technical difference with the competition is the marketing paragraph. A domain ending in ‘.ai’ serves to make the investor take his back off the backrest and whisper to the person next to him.

Nobody asks about EBITDA. Nobody expects profitability in five years. Not in ten. You have to keep turning the crank.

This, to another extent, We already experienced it with the dotcom bubble. What has changed is that we have twisted the loop of self-deception. There was more naivety in the ’90s: Too many people actually believed that Pets.com would revolutionize the dog food trade. Now almost everyone senses that this is more fragile than it seems, but no one can afford to be the first to say it. Because whoever says it, loses.

The CEO who admits his “AI integration” is just a wrapper of OpenAI with little uniqueness is left without the next round. The fund that does not invest in AI remains like a dinosaur. The CTO who says “this is cool but it’s not improving our productivity” risks being replaced by someone more enthusiastic: in this industry, a frown sells little. So many nod, many applaud, many pretend to see the complete revolution when perhaps they are only seeing the beginning.

Meanwhile, we are often seeing how the distance between narrative and reality continues to widen.

There are companies laying off employees justifying it as a “strategic reorganization towards AI”, when in reality they have burned capital on technology that does not work for them or at least does not pay off. Products are launched by releasing pigeons, fail six months later, and no one mentions the corpse because they are already busy announcing the next one. The metric of success is sometimes no longer “this solves a real problem” but rather “this got us another round of funding,” when not “this earned me a promotion.”

The curious thing is that this economy of belief can be sustained for many years. As long as there is liquidity and rates allow for financing losses indefinitely, as long as no one has clear incentives to break the consensus, the theater continues.

But there are two problems:

1. ANDThis dynamic destroys the sector’s ability to distinguish the real from the performative.. When much of the discourse is narrative and few ask about fundamentals, companies that really build something valuable can become difficult to distinguish from those that only know how to raise capital. Good engineers and good products sometimes get hidden amidst a lot of extremely well-funded mediocrity.
2. This economy constantly needs new believers. Like other speculative cycles, it works as long as there are more people entering than leaving. And if the music stops—when rates change, when investors demand tangible returns, or when customers stop paying on promises—there may not be enough chairs.

Here is the fundamental difference with the dotcom bubble: AI does have real and demonstrable value. ChatGPT solves specific problems, Claude Code development skyrockets and models improve quarter by quarter. Nobody believes at this point that they are vaporware like thirty years ago. There are companies using AI to improve margins, accelerate processes and automate tasks that previously required entire teams.

The issue is that The gap between the value that technology generates today and the capital it absorbs is considerable. And as long as that gap exists, the sector works more by consensus than by fundamentals. It’s not that everything is smoke, it’s that there is too much capital chasing too few profitable applications in the short term.

No one knows when the adjustment will come, if it ever comes. AI may end up justifying all bets and this will be seen in retrospect as the moment when the giants of tomorrow were built. Some of this capital may even end up funding breakthroughs that truly change entire industries. Many cities today have subways because someone more than a century ago decided to build tunnels and lay roads, assuming brutal costs without an immediate return. At the time it may have seemed like financial madness, but thanks to that today we don’t go by bus.

The difference is that this was public money betting on the long term. This is private capital waiting to multiply in less than a decade. And that difference matters, because it changes the incentives: whoever builds public infrastructure can wait two generations to see the return. Whoever raises a Series B round needs metrics in the next quarter.

So the optimistic scenario exists, but it coexists with another less rosy one: that a large part of the sector is playing the same game (believing because it is necessary to believe, investing because everyone invests) without really knowing where the bottom is.

For now, we just keep painting the ships red and acting like that makes them fly faster. Maybe I will. Maybe not. We’ll know when someone dares to check if the painting was what mattered.

In Xataka | AI needs 650 billion a year to sustain itself. The problem is who will put them on the table

Featured image | Xataka