English translation of the Italian 10x31 episode, first aired on 19 may, 2026.
When something collapses, it creaks first.
For a few days now, some coincidences that have got me thinking. And thinking is an almost subversive activity in these times, when all incentives only reward our knee-jerk reactions.
Let me put a few pieces together and see if you, too, start to see what I see.
The first thing, of course, is that insurance companies are, as one, stopping issuing policies for damages caused by artificial intelligence.
These are people who watch their money closely, and the fact that they don’t want to pay for certain things has a very simple explanation: the risk of having to pay is just too high.
We have a first sign: after almost five years of relentless propaganda, those who are supposed to help mitigate risks see the risks as too great. Interesting. Let’s move on.
Second point: major banks are repackaging the risks of loans to mega data centers into consumer-market financial instruments called… Collateralized Debt Obligations.
If that doesn’t ring a bell, we’re talking about the CDOs whose consequences we saw in the 2008 crash. In that case, CDOs repackaged the risk of subprime mortgages; today, they repackage the risk of loans to mega-datacenter builders. The risk being, of course, that mortgages and loans won’t be repaid.
In 2008, the crash was due to the fact that the vast majority of CDOs were tied to mortgages that had been granted to just about anyone and their dog, people who could never repay the loan. Today, CDOs repackage loans issued to highly reputable artificial intelligence entrepreneurs for the construction of data centers of a scale comparable to the mystical visions Sam Altman paints for his investors: behemoths of 160 square kilometers, essentially a square twelve and a half kilometers on each side, practically all of Paris inside the Boulevard Périphérique, all 21 Arrondissement of it.
There’s just the teeny tiny problem that there isn’t enough electricity to power them, and in fact, one of the few data centers actually built (Musk’s in Minnesota, also a very little one) runs on about fifty gas turbines because the local electricity supply isn’t even remotely sufficient.
In theory, there’s a gold rush in data center construction, because the bigwigs of Artificial Intelligence are talking left and right about their astronomical investments (of other people’s money), which will pay off as soon as Artificial Intelligence, in two or three years, works as they’ve been promising us every six months for the past five years.
In reality, most data centers have barely begun construction, and as Ed Zitron tells us, the money to complete them (and above all, to make them profitable) simply does not exist.
Another piece of news: the Artificial Intelligence bandwagon is doing everything it can to latch onto government contracts, first and foremost those in the defense sector.
Anthropic pulled out all the stops to get into the Pentagon, then using the Israeli-U.S. aggression against Iran as a promotional opportunity. Remember the headlines: “Anthropic helps the Pentagon plan 1,000 missions in the first 24 hours,” including the bombing of a middle school—which, curiously, no one wanted to take credit for.
Of course, two and a half months later, the world’s most powerful military, aided by Artificial Intelligence, doesn’t know how to extricate itself from yet another defeat; the Strait of Hormuz, which was open before the attack, is now paying tribute to Iran, and three U.S. Navy aircraft carriers (the stuff they advertise in Top Gun) are keeping their battle groups as out of sight as possible to avoid losing an aircraft carrier to two poor souls on a speedboat packed with TNT.
OpenAI, on the strength of its non-existent achievements, has instead managed to secure the analysis of data from decades of nuclear tests, a feat for which an OpenAI representative had to show up handcuffed to a briefcase to install ChatGPT on the heavily secured supercomputers at the Los Alamos research center.
How on Earth a language model, which must be trained separately on how many “r”s there are in “raspberry”, can ever be useful for data analysis is beyond my comprehension, but evidently the promise of AI magic can enthrall even the Los Alamos people.
Meanwhile, in response to no-one’s requests, manufacturers continue their race to cram so-called Artificial Intelligence into everything; Google just announced a Googlebook where Gemini interprets every move and decision the user makes, naturally via a voice interface. I'm sure it will be a huge hit in open-plan offices.
One is almost tempted to think that all this effort, in the face of no real market demand, seems to have the sole purpose of raising the price of a return to the status quo ante, once the speculative bubble bursts. But then, one ought to be suspicious and even distrustful.
Outside of the economic news, AI CEOs are giving non-stop interviews to sycophantic journalists, with free rein to claim unburdened by evidence just how important their work is, that Artificial Intelligence is changing everything, that work (obviously ours, not theirs) will change forever, and that since what they’re doing is so important, perhaps the government should step in to help them if they were ever to face financial difficulties.
Now, dear CEO, it’s one of two things: either you’re selling a product that someone wants, in which case it’s unclear why you wuould incur financial problems, and especially why the government should help you overcome them.
Or, you already have financial problems because for four years you’ve been subsidizing a product that no one is willing to pay enough to cover your costs, and in this case, your financial problems are called “the speculative bubble you rode into town,” and since the free market exists, you should deal with them yourself.
Just to reiterate, Anthropic spends between $8 and $13 for every dollar of revenue (the source is the Harvard Business Review, not “Communist Revolution”). To date, the shortfall continues to be covered by investor money or government contracts: as always, the taxpayer ends up footing the bill. But sooner or later subsidies will run out, and I don’t see companies lining up for the privilege of spending five or ten thousand dollars a month per programmer just to let them play with Claude.
We’ve known for at least two years that AI balance sheets don’t hold up, even with a crutch. We’ve known for at least two years that the famous “investments” being announced every week are actually a circular deal where everyone promises the same hundred billion dollars (typically in the form of discounts) and then books them as revenue.
"Wolf of Wall Street" meme. Sam Altman: Sell me this pen; Jensen Huang: I'll lend you the money to buy it.
There’s a meme going around, adapted from The Wolf of Wall Street. In the first panel, Sam Altman, pen in hand, says, “Sell me this pen.” In the second, Jensen Huang, CEO of NVIDIA, says, “I’ll lend you the money to buy it.”
It’s funny, but with the amount of money at stake right now, we should be terrified.
Anyone who isn’t hallucinating and hears the numbers peddled by Sam Altman and his associates immediately realizes they’re dealing with a snake oil salesman. And I’m not the one saying it: his own CFO says that, with the financials it has, OpenAI can’t afford to attempt an IPO this year.
We’re living in a speculative bubble whose scale dwarfs even 2008. US GDP growth in recent years is due exclusively to the speculative overvaluation of AI stocks, but as long as everyone believes it, the speculation holds.
Musk, who is smarter than Altman, has merged xAI and its losses into SpaceX and its secure military contracts, and will most likely go public within the year.
None other than the Wall Street Journal tells us how S&P is trying to rewrite its own rules to include OpenAI, SpaceX, and Anthropic in indices like the S&P 500. The idea is that the rules (which until now required companies to wait a year after going public and, above all, to report profits in order to be included in the index) would be ignored in the case of Initial Public Offerings large enough to enter directly among the top 100 stocks by market capitalization.
The fact that the value of Anthropic, xAI, and OpenAI is purely imaginary apparently doesn’t worry anyone.
And who buys the stocks included in the indices? Investment funds and pension funds. That is, savers.
In Italy, today’s Il Sole 24 Ore, Sunday, May 17, 2026, still has the nerve to run the headline:
Italy's "Il Sole 24 Ore": From credit to high-tech, profits up 20%. 2026 earnings drive Wall Street.
From credit to high-tech, profits up 20%. 2026 earnings drive Wall Street.
I, who, unlike Il Sole, am not financed by the Industrialist Association, look at the same world and come to slightly different conclusions.
Up until two years ago, one could say (and many did say) that critics of Artificial Intelligence were speaking out of bias, or that they were acting in bad faith, the usual Luddites and enemies of progress.
Until one year ago, one could still argue (and many did argue, conveniently ignoring circular investments) that the level of investment signaled an expanding industry.
But today the cracks are coming from within the establishment itself: banks, insurance companies, stock exchanges; not exactly people who jump in or out of an investment just because it’s trendy.
And the establishment is preparing for the collapse, dumping on a bamboozled public securities that will lose 100% of their value the moment the market wakes up from its trance.
Just like in 2008, the real economy will be pulverized by the collapse of the speculative economy’s imaginary values. Governments will step in to save the culprits, and we plebs will foot the bill.
Our only consolation will be remembering the names of the scoundrels who created this disaster, and of the useful idiots, cheerleaders, and politicians who enabled them. And, this time, pass the bill to them.
All of them.
Storie Spettinate
in reply to LINKioni • • •Scimmia di Mare
in reply to Storie Spettinate • • •Scimmia di Mare
in reply to Scimmia di Mare • • •Artematiko
in reply to Scimmia di Mare • • •qua l’unico che sembra essere se stesso è Cristiano - dal gruppettino.
Io e Joda stiamo in crisi d’identità
@storiespettinate @LINKioni @sio
Storie Spettinate
in reply to Scimmia di Mare • • •DajeLinux
in reply to Storie Spettinate • • •