Back
Automatically translated version. May contain inaccuracies compared to the original.
The global economy is hooked. Not on oil and not on the dollar. On artificial intelligence.
This is no longer a metaphor but the official position. 13 October the IMF in its World Economic Outlook directly acknowledged that global growth rests on investments in AI. Oil above one hundred dollars, expensive money, war, and the forecast for 2026 year is still 3,1%. Data centers, chips, and the power to run them make up for everything else. They even raised forecasts for the Eurozone and the UK specifically because of AI.
This machine has one operating condition: it must constantly receive new money. Morgan Stanley estimates that by 2028 the AI infrastructure will need about 1,5 trillion dollars of external financing. Not from retained earnings, but from loans and other people's capital. This is no longer a technology story but a credit story.
And here it’s worth recalling the previous post. Money is getting pricier. Treasury yields are at highs not seen since 2002, interest rates are high. AI needs ever more capital, and borrowing it gets more expensive month by month.
Now the most delicate point. The market is expecting an almost perfect scenario. From the third quarter analysts expect S&P companies’ profit growth for 500 to be about 31%, and roughly two-thirds of that growth will come from technology and AI. Prices already assume everything will go according to plan.
And here is the first crack. And it’s very telling.
Firmus Grid, an Australian operator of AI data centers backed by Nvidia. The perfect story for 2026: chips, energy, compute capacity, the right investors. The company was preparing one of the largest IPOs in Australia’s history: up to 5,5 billion dollars, 11 Australian dollars per share, a valuation of about 30 billion.
Now the numbers. Less than three months ago, in August, Firmus raised 2 billion at a valuation of 10,5 billion. In a year the company raised over 3 billion of its own capital. It went to the exchange already at a valuation three times higher. Not because the business grew threefold. Because that’s how it was done: AI gets more expensive simply because it’s AI.
A few days before the listing bankers were assuring that demand was healthy. The main backers were large American funds; they were supposed to buy a significant portion of the deal. The funds looked at the valuation, the terms, treasury yields above 5% and said “no.” Not “let’s make it a bit cheaper.” Just “no.”
9 October Firmus officially withdrew the IPO. The explanation was standard: market volatility. But the deal collapsed in forty-eight hours, and everyone in the market understands why. Investors refused to buy a valuation worthy of a top-ten industry player.
It’s important who said “no.” Not retail investors panicking. Not skeptics who have been shouting “bubble” for years. Professional managers who were buying everything yesterday. The same money that the whole machine depends on.
And this is not an isolated case. Other data centers are lining up for the IPO. British Nscale, also backed by Nvidia, reportedly filed for the NYSE with a valuation around 30 billion. Now each such listing will be examined under a microscope, and every banker will remember Firmus.
Previously, in AI they bought anything with the right name. Now they’ve started to do the math.
The chain is simple. Global growth rests on AI. AI rests on new money. New money is getting more expensive, and investors for the first time say “too expensive.” It seems this is the peak. And right now it’s worth thinking where your assets will be at the moment they stop funding the machine.
Dependency isn’t noticeable when the dose is present. You notice it when the dose runs out.
💬 ABleaks_bot
💕 Join
Telegram publication
