(BFM Stock Exchange) – The director general of the young American shoot recently entrusted American journalists whom he estimated that a bubble was created around artificial intelligence. This question has been debated on the market for more than two years. But, overall, investors do not see a bubble at present.

Sam Altman may be at the heart of the revolution caused by the rise of artificial intelligence (AI), this does not prevent him from issuing an important warning.

The director general of Openai, the American start-up behind the major GPT and Chatgpt language models, said that investors demonstrated “overexcitation” as to artificial intelligence.

“When bubbles occur, clever people are excited by a core of truth,” he told journalists during an evening, according to the words reported by The Verge. “If you look at most of the bubbles in history, like the technological bubble, there was something real. The technology was really important. The internet was really important. People were excited,” added Sam Altman. If the leader believes that this bubble will burst, he sees (obviously) his company survive.

Two details deserve to be provided. First of all, Sam Altman seemed, according to the content of the remarks relayed by The Verge, to evoke investment in AI in its entirety, that is to say by also including the unlisted (“Private Equity”).

Then, it should be noted that Optai is not herself present on the stock market. Sam Altman would not necessarily have the same speech if his company was listed at Wall Street. According to several American media, Openai is preparing to make a sale of shares that would value the company at around $ 500 billion, the highest valuation ever obtained for an unlisted company.

>> Access our exclusive graphic analyzes, and enter into the confidence of the trading portfolio

A bubble only exists when it breaks out

Be that as it may, the director general of the young American shoot put a room in the machine, the question of a bubble in the AI posed for two and a half years now. UBS, for example, was “bubble fears” in the spring of 2023.

More recently, last November, the European Central Bank (ECB), in its “review of the stability of the financial system”, noted the high concentration of American markets on a handful of values. The European institution added that this concentration “raised concerns about the possibility of a bubble of the price of assets linked to AI”.

Last month, Torsten Slok, chief economist of Apollo Global Management, a famous investment fund, wrote in a note that the “Bulle of AI is today larger than that of the Internet bubble of the 90s”.

The economist advanced as an argument that the ten largest capitalizations of the S&P 500 displayed multiple scholarship holders much higher than in the 90s, with a course ratio on benefits expected within 12 months located between 25 and 30, against less than 25 in 2000.

However, this observation fails to specify that the S&P 500 in its entirety has also seen its multiples appreciated clearly since the 1990s-2000.

Slacing the question of a bubble is always delicate. Technically, a bubble occurs when the prices of assets, such as actions, bonds or raw materials, rise to the point that these prices clearly exceed their “intrinsic” value (which would be defined, for example, by fundamentals, such as supply and demand).

But more pragmatic, the debate on a bubble is mainly resolved when the said bubble exploded. “A Krach does not anticipate, by construction, otherwise it does not occur,” explained Julien Nebenzahl, director of savings solutions at Etoro, on BFM Business last year. We really recognize a bubble “only when it begins to deflate”, he added.

The managers respond in the negative

In the case of artificial intelligence, market specialists tend to call for measure.

“It is really difficult to consider that it is a bubble. The speed of ‘pricing’ (the price formation on the market, editor’s note) was very fast but ultimately it is mainly large groups which benefit from this ‘IA’ effect and the market is perhaps generous in its valuation anticipations”, noted in 2023, Alexandre Baradez, market analyst for IG France. “These companies already have a ‘very profitable core business’, AI only adds value. We can head towards consolidation”, but not towards “the bursting of a potential bubble”, he judged.

The results of Nvidia, the big winner of artificial intelligence at Wall Street, tend to rule out the idea that the development of AI on the markets is decorated with fundamentals. Over its last financial year published, its income jumped 114%, its profit of 145%. His generation of cash from his operational activities had been multiplied by more than two, at 64.1 billion dollars.

“If we only look at NVIDIA’s figures and the prospects of its customers, it is difficult to see how AI would be a bubble. It is rather an opportunity to seize”, judged last summer Christopher Dembik, of Pictet Wealth Management.

Bank of America questions fund managers every month. In his latest survey, which surveyed 169 participants from July 31 to August 7, 52% of the managers questioned estimated that AI actions did not evolve in a “bubble”, when 41% thought the opposite. During the previous monthly survey, these rates were 54% and 37% respectively.

In a note published at the end of July, Bank of America underlined that the recent increase in technological values went “against the punctual and volume prices, typical of stock bubbles”. The establishment believes that, at present, no volatility signal suggests that a bubble can explode. However, the bank warned that bubbles can take years to train and then deflate.

Gigantic “capex”

One point can however make an eyebrow: the Mega-Déspens of companies in AI. Microsoft, Meta, Alphabet and Amazon have multiplied the announcements of “CAPEX” (investment expenditure) to invest massively in the infrastructures necessary for the development of AI, like data centers.

Microsoft mentioned an amount of $ 30 billion in the current quarter (and therefore 120 billion in annualized pace). UBS expects more broadly that IA expenses leap 67% to 375 billion dollars in 2025 before 500 billion in 2026.

“Some compare the current wave to the railway boom, on radio in the 1920s or the Internet bubble of the 1990s,” summarizes John Plassard, Cité Management Private Bank, in a note published on Monday.

However, “unlike the Internet bubble, AI giants already display colossal cash flows and solid profitability, capable of absorbing these investments,” added the market specialist.

John Plassard points out more “a risk of excess capacity”, “as shown by episodes of overproduction in maritime transport or energy”.

Investments in large tech groups are accompanied by an improvement in their income statements. Thanks to the AI, Microsoft was able to display a 39% growth of its dematerialized computer service division (“Cloud”) in the last quarter published when Meta explained that AI allowed it to target advertising content more and improve the conversion rate. The company’s revenues jumped 22% in the last quarter.

Encouraging signals

UBS sees, more generally, encouraging clues. “While the growth of tech groups remains lower than that of CAPEX, the monetization of AI continues to show signs of improvement,” she said at the end of July. The profits by action of tech groups had, moreover, increased by 15% this year and by 12.5% in 2026, according to its projections.

“These enormous investments support profits because they represent income for someone, and they help stimulate productivity gains and increase beneficiary margins, not only for technological companies, but for all American companies,” Jeff Buchbinder, chief strategist at LPL financial, wrote by Barron’s on Monday.

“The amounts committed could propel the economy to a new productivity cycle … or create an oversized infrastructure bubble. In all cases, ignoring this movement would be a strategic error for any investor,” says John Plassard.

An important test took place at the start of the year, when several media reported that the Chinese start-up Deepseek had developed AI models with performance comparable to those of American groups but with much lower costs. The values related to the AI have tanned on a session but did not collapse. And market experts very quickly put into perspective the threat made up of Deepseek as well as its prowess.

Without a bubble bursting, however, phases of corrections cannot be excluded. This is what happened Tuesday at Wall Street, where the Nasdaq Composite fell 1.5%weighted by the folds of several “IA” shares such as Nvidia (-3.5%) or Palantir (-9.4%).

“Investors have reduced their exhibition to the sector, fearing that the spectacular increase started in April was too fast and too marked,” judges John Plassard.