The Artificial Intelligence Trap in the Markets

How Financial Automation Commits Millions of Savings Without Anyone’s Consent

There are moments in financial history when danger does not hide in complex products, obscure derivatives, or the darker corners of the banking system. Sometimes it stands in plain sight. It enters through rules. It settles inside the indices. It arrives in technical, neutral, administrative language — and that neutrality is precisely what makes it hard to name.

That is what seems to be happening now with the artificial intelligence boom.

Before going further, it is worth clarifying a few terms.

An IPO, or Initial Public Offering, is the moment when a company that was previously private begins selling its shares to the public on the stock market. Put simply, it is when a company “goes public”, and ordinary investors can buy a piece of it.

A stock market index is a basket of companies used to measure how a part of the market is performing. The Nasdaq-100, for example, groups together many of the large technology companies listed on Nasdaq.

An index fund is a fund that does not try to choose companies one by one. It simply copies an index. If the index contains Apple, Microsoft or Nvidia, the fund buys Apple, Microsoft or Nvidia. If tomorrow another large company enters the index, the fund buys that company too.

A company’s free float is the part of its shares that is actually available to buy and sell on the market. If a company is worth an enormous amount of money but releases only a small part of its shares to the public, the real supply stays thin. That can drive the price upwards, because many buyers are competing for very few available shares.

Now we can come to the point.

In May 2026, Nasdaq introduced important changes to the entry rules for the Nasdaq-100. The so-called fast entry rule allows a newly listed company to be included in the index after only fifteen trading days. The previous minimum free-float requirement of 10% was also removed. And for companies with a low free float, the index applies a special rule that may compel index funds to buy more shares than the real supply would naturally justify.

Put simply: a gigantic company can go public while making only a very small part of its shares available to the public, and still enter a major index very quickly. When that happens, passive funds have to buy.

Not because they have analysed the price.

Not because they believe the business is healthy.

Not because a fund manager has consciously decided to take that risk.

Because the index tells them to.

The passive market does not think. It replicates. And when the index changes, the money of millions of people moves — automatically, without deliberation, without anyone being asked.

This is where the great expected IPOs come in: SpaceX, OpenAI and Anthropic. We are not talking about small companies or marginal start-ups. We are talking about companies whose combined valuation could exceed three trillion dollars.

A valuation is the estimated price of an entire company. If people say that a company is worth one trillion dollars, it means the market is treating the whole business as if it were worth that amount.

In the case of SpaceX, some estimates speak of a possible stock market listing at a valuation close to $1.75 trillion, while Morningstar values the company much lower, at around $780 billion.

That difference is not a technical detail. It is an open wound in the numbers — and wounds of that size do not stay open indefinitely without something giving way.

If these companies enter the major indices quickly, millions of people will become indirect shareholders without having made any conscious decision. Anyone with a private pension plan invested in global index funds may find themselves holding these companies automatically.

Not because they chose them.

Not because they understand them.

Not because they explicitly accepted that risk.

But because they are inside the index.

Here the key idea appears: we may not only be facing a possible valuation bubble. We may be facing an earnings bubble — which is a different and harder problem.

A valuation bubble occurs when the price of a company rises much faster than its real profits. It is like a house that produces a normal rent suddenly being sold at an absurd price simply because everyone believes the price will keep going up.

An earnings bubble is more dangerous. Here, the problem is not only that the price is high. The problem is that the profits themselves may be inflated by accounting, revaluations or paper gains — so that the company looks healthier than it actually is.

To see the distinction clearly: one thing is to make money because you sell more products and your customers pay you. Another thing is to declare that you made more money because a stake you own in another company is now “worth more” on paper.

The first is operating money.

The second may be an accounting gain. And an accounting gain can dissolve if the valuation falls.

Before 2008, banks and property companies also showed profits. Not all the numbers were invented in the simplest sense. The problem was that the assets supporting those profits were overvalued. When the real value of those assets was revised, profits that had looked solid turned out to be fragile structures built on estimates.

Today, the accounting circuit around artificial intelligence is beginning to show a troubling architecture.

The large technology companies are investing enormous amounts in AI infrastructure: data centres, chips, energy, servers, networks and cloud capacity. At the same time, start-ups such as OpenAI and Anthropic are major future customers of that same infrastructure.

A start-up is a young company, usually growing quickly, that is still building its business model. It may carry an enormous valuation on paper while not yet generating stable profits.

The technology giants invest in the start-ups. The start-ups rent computing power from the technology giants. Then the technology giants record the rising value of their stakes in those start-ups as profit.

The circuit is elegant. It is also, for exactly that reason, worth examining with some care.

Money goes out as investment. It comes back as cloud revenue. The private valuation rises. That rise gets recorded as an accounting gain. That gain reinforces the growth narrative. The new narrative supports new funding rounds. The new funding rounds sustain new computing contracts.

As long as everyone holds the valuations, the system functions.

But the circuit runs in both directions. If real AI demand does not justify the spending — if the start-ups do not generate enough revenue, if private markets revise their valuations, if the cloud contracts lose credibility — the accounting gains can reverse. What once inflated profits can begin to destroy them. The elegance of the circuit does not protect against its own inversion.

That is why simply speaking of an “AI bubble” misses the structure of the problem.

The technology may be real. Artificial intelligence may transform medicine, defence, education, software, logistics, science and productivity. None of that guarantees that the investor who enters at the peak of the euphoria will profit.

History is unambiguous on this point.

The American railway boom of the nineteenth century built real infrastructure that transformed the United States. But many investors who entered during the euphoria lost capital. The same happened with fibre optics in the 1990s. The internet did not disappear. Fibre did not stop being useful. The demand for data did not die. But many of the first investors paid absurd prices, and a second generation built durable businesses by buying cheap assets after the bankruptcies.

The technology won. The investor who entered at the wrong moment did not. That distinction is not peripheral — it is the whole question.

The macroeconomic context makes the present moment sharper.

The 30-year US Treasury bond is debt issued by the United States government. When that bond pays around 5%, it means the American state must offer substantial interest rates for investors to lend it money over the long term.

Interest rates function as gravity in the markets. When rates are low, almost everything can float: expensive shares, distant projects, cheap debt, future promises. When rates rise, gravity increases. Money costs more. Distant promises contract in present value. Debts press harder against balance sheets.

Financing is no longer cheap. The large technology companies are issuing substantial amounts of corporate debt to fund the AI build-out. At the same time, the United States must refinance a significant portion of its outstanding debt — around ten trillion dollars this year.

To refinance debt means to pay old debt by issuing new debt. A family that borrowed cheaply ten years ago and must now renew those conditions at a higher cost discovers that the burden has grown without the house changing. The same logic applies to sovereign balance sheets, only at a scale where the margins matter more.

High interest rates make government refinancing more expensive. They press on corporate balance sheets. They reduce the present value of future profits. They punish companies that depend on distant growth.

For someone with a private pension, the exposure is direct. If their fund sits inside global indices, they may find themselves buying these companies at the moment the indices absorb them.

For someone who depends on a public pay-as-you-go pension, the exposure travels a different path.

A public pay-as-you-go pension does not work like a private account invested in the stock market. Today’s workers pay contributions, and that money funds today’s pensioners. There is no individual pot where each person holds their own shares.

But the public system does not live outside the financial world. If interest rates rise in the United States, risk premia in Europe can rise in turn.

A risk premium is the extra cost a country carries to borrow money compared with a country considered safer. When investors see more risk, they demand more interest. When a state must finance itself at higher cost, it loses room — and within that room are pensions, healthcare, and the welfare state.

Two channels of transmission, then.

The private channel: index funds, pension plans, ETFs, passive portfolios — all of which replicate automatically what the index includes.

The public channel: interest rates, sovereign debt, refinancing costs, fiscal pressure on the spending that reaches ordinary households.

This does not mean a crisis is inevitable. Markets can sustain excess for years. They can punish the prudent and reward the reckless for longer than seems rational. But the architecture of risk is already visible — and visibility is the first condition for naming it.

What we may be watching is a transfer of risk from private markets to public markets. Large private funds invest early. Valuations rise. Companies prepare for public listing. They enter the indices. And once they enter the indices, the final buyer is no longer the sophisticated capital that entered first.

It is the passive saver.

It is the worker with a private pension.

It is the citizen whose money is being committed to that story without anyone explaining it to them.

That is the actual problem — not the technology, not the companies, not even the indices as such, but the chain that ends with someone who was never invited to understand what they were buying.

The question is not whether artificial intelligence is real. It is.

The question is not whether SpaceX, OpenAI or Anthropic are important companies. They are.

The question is different: who is buying at the end of the chain?

Because when a private company arrives on the market with a gigantic valuation and enters the indices quickly, it is not only raising capital. It is transferring risk — moving exposure from the investors who priced it early to the instruments that absorb it automatically.

And if that transfer happens at the moment of maximum euphoria, the public may end up holding what the earliest investors needed to sell.

A civilisation can generate transformative technology and, at the same time, pay too much for it. A company can be extraordinary and still be overvalued. An index can follow transparent rules and still propagate systemic risk. A passive fund can be diversified and still mechanically buy into a bubble.

The trap is not artificial intelligence.

The trap is the financial automation that routes people into it without their knowledge and without their consent.

For years, ordinary citizens have been told not to worry too much. That they do not need to understand. That it is enough to work, save, pay taxes, contribute to the system, trust the experts. The message has been consistent: live your life, the system knows more than you do.

Every great crisis reveals the same structure: when the moment of loss arrives, the system stops being abstract. The loss comes down to the family table. It comes down to wages, pensions, the price of food, the cost of the mortgage, the future of the children.

The gains were sophisticated. The losses became domestic. That has been the pattern, and there is no reason to expect this episode to differ.

While everything rises, the language belongs to the banks, the funds, the indices, the regulators, the experts. When everything falls, the blow is no longer absorbed by a technical term. It lands on the ordinary family.

The housewife who has never heard the word IPO.

The worker who cannot tell you what sits inside his pension fund.

The pensioner who discovers that the state now says there is no room.

The young person who finds that the promise of progress can arrive with debt, insecurity, and inflated prices already baked in.

This is not only a financial question. It is a moral one — and the distinction matters.

One thing is to invest consciously and accept risk with open eyes. Another thing is for risk to be packaged, automated, and distributed among people who were never offered the chance to understand it. That second operation, dressed in the language of diversification and passive investing and the long run, deserves to be named for what it is.

Artificial intelligence may be a real revolution. It may change the world. It may open a new technological age. But none of that justifies turning the ordinary citizen into the final buyer of a euphoria that others constructed and others will exit before them.

True progress should not require darkness to function.

If something is genuinely good, it can be explained clearly. If an investment is genuinely solid, it can withstand simple questions. If a company is worth trillions, it can show real profits — not only promises, cross-contracts, and accounting gains that dissolve when the private valuation is revised.

And if pension money is going to enter there, people have the right to know. Not a disclaimer in paragraph forty-seven of a prospectus. The right to actually know.

Behind every index, there is a life.

Behind every fund, there are years of work.

Behind every pension, there is accumulated tiredness, early mornings, worn bodies, sustained families, and silent sacrifices made in the reasonable belief that the system would hold.

We are not speaking only about numbers.

We are speaking about the savings of a lifetime.

The final question is not technological. It is not even financial.

It is human: who carries the risk when the music stops?

If the answer is once again the ordinary people — the ones who were told not to worry, the ones who trusted the experts, the ones who never saw the chain they were standing at the end of — then what has been described here is not progress.

It is an elegant transfer of danger.

And that transfer, however much it dresses itself in innovation, indexation, artificial intelligence, and the language of the future, must be named.

Because no one should discover too late that their pension was not simply invested in the future.

It was financing someone else’s exit.