GitHub

There is a website called “Is AI Profitable Yet?” whose sole mission is to answer one of the most important—and most uncomfortable—questions of today's technology industry: does artificial intelligence make money anymore? The visual response is absolutely overwhelming:

The short answer is a priori a big NO, but be careful, because that answer is in a certain sense misleading. The graph effectively shows how the companies that are building frontier models are burning money as if there were no tomorrow, and they all spend much more than they earn.

The four that appear with long red bars (expenses) and very short green bars (income) are precisely the companies that are betting almost everything on the future of AI. Amazon, Alphabet, Microsoft and Meta have not stopped increasing their capex (capital expenditures) in recent years, and that logically means that their accounts are in the red.

In fact, the announcements of these "hyperscalers" in their latest financial results have not only failed to soften that capex, but have driven it even further. The combined capex of these technologies by 2026 is expected to amount to $725 billion, 25% of all global military spending.

But the message of "everyone is losing money" is dangerous, because what all these companies are doing is investing in their future even though in doing so they are running out of cash flow. There are two clear examples that can alert us.

The first is Amazon, which did not stop losing (investing) money for years and then became the giant it is today. The second, Uber, a company to which the same thing happened: it lost (invested) money for a decade, and although it does not have the size or success of Amazon, today it is an absolute world leader in its segment.

This leaves us with a clear message: not being profitable by investing in your future is not the same as not being clear about the economic model. And all these companies are very clear about the economic model of AI: it is to invest today to earn (a lot) tomorrow.

Nvidia is the big winner, but not the only one

The great irony of AI is that for now the big business does not seem to be in AI, but in selling infrastructure to those who try to do business with it.

It is the same thing that happened during the gold rush in mid-19th century California: those who amassed stable fortunes were not the miners who searched for gold, but those who provided them with services and tools.

There are several well-known examples: Levi Strauss saw the need for sturdy clothing, Samuel Brannan bought all the shovels, picks and pans he could in the area, and Henry Wells and William Fargo founded the famous postal and financial services company that allowed money and supplies to be sent safely to gold seekers.

Nvidia is basically doing that: (making and) selling blades. This has caused absolutely extraordinary growth in the stock market, and in the last three years it has become the most valuable company in the world and has not stopped breaking market capitalization records.

Here it must be clarified that the estimates on that website are striking, but they do not mean that these companies are in any way bankrupt. Google / Alphabet continues to make billions of dollars every quarter, and the same goes for its rivals.

All those red bars don't mean that AI is smoke: just that we're footing the bill for the experiment. One that could go wrong, of course, but one that could also go really, really right. The phrase that best summarizes this "AI fever" is the one that Mark Zuckerberg uttered a few months ago:

"We're going to invest aggressively. Even if we lost a couple hundred billion dollars it would be a bummer, but it's better than being left behind in the race for superintelligence."

Neither Zuckerberg nor his rivals seem upset about losing $200 billion right off the bat. They certainly do not seem to be wrinkled despite the fact that at the moment there is a reality in the market: AI already works technically, but what it does not do is work economically for those who invest in frontier models.

Here, however, there are a couple of notable notes. The first, the fact that Anthropic apparently plans to end the quarter making money, something unusual and promising. The second, that this website only shows Nvidia as the winner of this AI race, but that company is by no means the only one that has managed to make gold with this technological fever.

In fact, we are seeing how a large number of technology companies have grown extraordinary in recent months thanks to the demand for hardware and components such as memories. Micron. SK Hynix and Samsung are the big beneficiaries of this situation, but they are not the only ones either. These days we have seen how PC manufacturers barely grow in revenue from those PCs, but they are doing so with servers.

There are more winners. There are photolithography equipment manufacturers such as ASML or Applied Materials, but also electrical, liquid cooling, networking, storage companies, and of course companies specialized in data center construction.

This website answers the question in a very limited way, because the AI segment is not only the one in which OpenAI, Anthropic, Microsoft, xAI or Google operate. What's happening is simply that the big business of AI is currently not where everyone thinks it is.

AI is being very profitable. The problem is that perhaps we are looking in the wrong place.

In Xataka | The problem is not spending a lot of tokens, it's that most of them are being wasted