AI incest spawned a multiheaded monster

Whatever is thrown at the US stock market, it just shrugs and gallops on, soaring gloriously into the sunrise.

Inflation, deadly wars, global rearmament, yawning supply bottlenecks, inexorably rising energy prices, a shrinking US labour market, enthusiasm for nonsense ‘meme’ stocks, America’s self-destruction as a world leader, protectionism, tariff wars, bubble fears – all not much of a bother. In normal times, every single worry would suffice to tank markets. How long can this continue?

Fearmongers have pointed out how investor exuberance is reminding them of the dot-com bubble of the 2000s, which burst so spectacularly, with the NASDAQ Composite losing 78% and the S&P500 falling 49% in two years.

Doomsayers are particularly worried about the ‘concentration’ of the US stock market. A handful of stocks carry the total valuation of 4,000 companies, seemingly unstoppable uphill.

The ‘Magnificent Seven’, all of them trillion-dollar companies riding on the lure of the AI story, comprise 30.47% of the S&P500 industrial index.

These stocks have indeed fear-inducing valuations, exceeding historical norms. Amazon, with a market capitalisation of $2.84tn, trades at a price/earnings multiple (P/E) of 31.1. It has gained 49.85% year to date, 66.66% in the last five years, and 13.87% in the last six months. Alphabet ($4.8tn market cap, P/E 30), the Google parent, has gained 140% in the last year. Apple ($4.4tn, P/E 36.2) has grown 42% year to date. Meta, the Facebook company ($1.5tn) has stopped growing in value lately and is cheap in comparison (P/E 22). Nvidia ($5.46tn), the chipmaker in the epicentre of the AI drama, has grown 1,418% in five years, with Q1 26 profits up 76%. Tesla, valued at $1.3tn, behaves as always more like a meme stock, rising despite shrinking revenues. It has a P/E of eye-watering 353 � at current earnings, it would take four centuries to just recoup the cost of its shares.

American tech companies have been riding for years on the rule-setting dominance and the legal bullying of the US. We all depend on them in our daily lives.

Judge Nicolas Guillou, at the International Criminal Court, was sanctioned by Donald Trump for daring to issue an international arrest warrant for Benjamin Netanyahu and his former defence minister, Yoav Gallant for crimes against humanity in Gaza.

Guillou cannot Google anymore, he cannot use his credit cards, he cannot use his Microsoft e-mails, cannot order on Amazon. Most banks, thinking of their US exposure, will not heed his payment instructions. He was thrown back to a time without smartphones and the internet, and has to live a life usually reserved for cash-wielding criminals.

These tech behemoths are investing in AI technology as if there will be no tomorrow, shrinking their free cash flow and raising debt when the till runs empty. The ‘hyperscalers’, Amazon, Alphabet, Microsoft, Meta and Oracle, and the AI companies at the root of the excitement, OpenAI and Anthropic, are estimated to invest upwards of $800bn this year and $1.1tn in 2027. Some of the Seven, like Meta and Apple, can afford what they invest or invest with measure.

Others, like Amazon and Oracle, or the still private OpenAI, have bet their house on the future. They are investing in chips and building data centres or renting them long-term from others. It is just the tip of the iceberg. Smaller companies, like chip designers AMD, Intel, Broadcom, Qualcomm or Cerebras, or data centre builders like Nebius follow close behind.

Interestingly, most of the stuff needed to build the dream castle of the times is imported, from cables to chips. They are all unfazed by rising energy bottlenecks, scarcer raw materials and the repercussions of the Gulf war.

Investors are starting to discern the headwinds of growing competition and the weakened credit standing of individual firms. It does not go unnoticed that hyper-scalers like Amazon, Meta or Google have started to design their own AI chips, to the detriment of Nvidia (gross profit 62% so far); or that those chips are made in China or Taiwan; or that AI development is spreading far beyond ChatGPT.

But what nobody seems to be overly bothered about is the incestuous relationship of all the companies involved. Yes, institutional investors and analysts count worry beads but nobody, including the alerters themselves, seems to pay heed. Chipmaker Nvidia injects ‘capital’ into customers like Anthropic or OpenAI, which then dutifully use the money to order Nvidia’s chips. It invests in chip designer Marwell Technology to smooth competition. Meta ‘invests’ in chipmaker AMD, which uses this capital to build data infrastructure for the Facebook parent. Oracle has an order book with OpenAI, amounting to hundreds of billions of dollars, for new data centre facilities that may never be built but parades its future profits as if they would have been earned.

Robin Wigglesworth, financial commentator at the Financial Times, has stumbled upon a curious observation reported by the Fortune newsletter. It demonstrates the incestuousness of the AI biotope. Over half of Amazon’s and Alphabet’s reported net earnings in Q1 26 came from gains of their AI investments in OpenAI and Anthropic. This money flows back to rent compute power from the two hyper-scalers.

As both Anthropic and Open­AI are still private companies, their account­ing worth is defined by repeated funding rounds, each at higher valuations. As Amazon and Alphabet are the pivotal, often sole, fund providers in each round, they essentially bid up the value of the companies themselves. And, hey presto, as the value of their investments goes up, so does their own profitability.

It is one thing when stock market darlings ebb and rise according to their advances and setbacks, but an entirely different thing when they all depend on each other in their earnings and their valuations and the whole stock market on them.

Credit markets, from private credit to the bond markets, will lose in tandem with the stock market, infected by the same behemoth companies. The downturn will go full circle when institutional and retail investors become poorer and consumption suffers.

Like the proverbial Hydra, AI has many faces, or individual players, but all of them are linked together for their survival, like Siamese twins on one body with a single heart. If one stone falls, an avalanche of repricing will pull everything with it. A single event can trigger a cascade of value destruction. Last month, we were already close.

Elon Musk, revengeful like Trump, had sued Altman’s OpenAI to the tune of $150bn, allegedly aggrieved by Altman’s move to turn OpenAI from a charity into a for-profit company.

It would have cut off OpenAI, a still painfully loss-making enterprise, from further funding, investment and debt raising. The giant, planned IPOs of Anthropic, OpenAI and SpaceX, now morphing according to Musk’s wishes from a satellite company into yet another AI play, would have been called off. Had the court not dismissed Musk’s case, it would have collapsed the whole AI entanglement, pulling the rug from under the feet of the stock market, debt markets and the US economy as a whole.

There’s hardly a corner in the US and global enterprise that does not profit one way or the other from the world’s biggest investment boom since the 19th-century railway craze. The lawsuit would have achieved what Trump’s war in the Strait of Hormuz has so far failed to trigger: a crash of epic proportions.