Is the AI bubble about to burst? Not if the leaders of the largest technological companies should believe.
Gathered around a table in the White House, several of the CEOs of the most powerful technologies in the world have praised President Donald Trump for his leadership, saying that his AI approach will position the United States as a dominant player in the emerging sector.
Between them, Mark Zuckerberg, Tim Cook, Sergey Brin, Bill Gates, Sam Altman and others manage companies that lead the current AI revolution – and if they are to be believed, the American AI sector is at the dawn of grandeur.
The co -founder of Google, Brin, said: “It is a real incredible inflection point at the moment in the AI and the fact that your administration supports our companies instead of fighting with them – it is extremely important. It is a global race and I think that we are at dawn where these models of AI are about to become deeply useful … So we are very grateful for the support of your administration.”
The boss of Oracle Safra Catz accepted, adding that “the AI will change everything” while the CEO of Alphabet Sundar Pichai added: “The moment of the AI is one of the most transformative moments that we have ever seen or will see in our lives.”
But elsewhere this weekend, evidence suggests that the AI of media threshing cycle could lose steam.
Ralent of rates
A study by the Apollo Academy published this weekend has shown that AI adoption rates in large companies are starting to slow down.
The study is based on data from the American census office, which is published bihebdomedary and survey 1.2 million companies. A question added to the survey since 2023 is whether a company has used AI tools, such as automatic learning, natural language treatment, virtual agents or voice recognition, to help produce goods or services in the past two weeks.
From November 2023, the figure increased in American companies of all sizes, but perhaps without surprise, the fastest increase concerned companies with more than 250 employees. But according to the latest figures, this number dropped in June and July for large companies and has also flattened in small businesses.
The data measure business integrations, not the use of AI tools at all levels, but this suggests that the adoption of large downhill implementations slows down.
This is saved by RAMP expenditure data, published earlier this year. The Ramp AI index suggests that the dazzling increase in commercial AI expenses could show signs of deceleration.
The figures show that the penetration of AI among American companies had reached 41.7% in April, but the growth trajectory has flattened since the end of last year. It should be noted that RAMP index data is drawn from its customers, so it may not be reflected throughout the United States or the world markets.
Torsten Sløk, chief economist of the Apollo Academy, underlines that the slowdown in signals lends itself to large companies, while Arpit Gupta, an associate professor of finance in Nyu Stern, suggests that “billions of billions of people in IA Capex should probably be reconsidered”.
For Andy Ward, more than International to Absolute Security, reports highlight a growing concern that companies are starting to question the return on investment of AI projects.
He says: “AI can transform detection and response, but if it is deployed without robust resilience strategies, real -time visibility and clear governance, this may add more vulnerabilities than it does.
“Our research shows that more than a third (34%) of CISOs have already prohibited certain IA tools such as Deepseek entirely, motivated by fears of privacy violations and the loss of control. While attackers exploit AI to reduce the gap between vulnerability and exploitation, our defenses must evolve with an equal emergency. Now is time for security managers.
Back on investment problems
David Tyler, founder of the Ovre technology, believes that MIT data – which showed that 95% of AI generative projects have shown no value – suggests that the media threshing cycle is about to burst.
He claims that investors were “blinded by the promises of the braking machine” and “continuously increased the amount of money thrown into AI”. The main players report losing $ 2.40 for every $ 1 they earn; And recent economic modeling of data centers has shown that they would need 3.5 billion users, each paying $ 35 per month, just to break even, explains Tyler. “This is equivalent to 64% of all people with internet access worldwide.
“Thus, the economy does not accumulate, the majority of projects provide no value, and Chatgpt 5 has been criticized as being completely the” game changer “promised before its release. Which contributed to the increase in general attention to the Bulle of AI.”
However, he says that AI can always play an important role in the future if companies move expectations.
“We must realize that AI has specific uses, but there is not a single model that will revolutionize the whole world,” adds Tyler. “We will always need humans to be part of the process somewhere at the bottom of the line, and we will better focus on the small gains that we can make and the specific processes that we can improve with AI, rather than plowing billions of dollars in an inaccessible magic solution” which will be all things that will be all things to all men “.
A recent Qlik survey highlights the problem of return on investment. Some key dishes of the report have shown:
- Only 11% say that most of their AI initiatives have provided tangible gains
- Only 51% AI assessment using KPIs linked directly to the performance of the company
- Almost a quarter (23%) say that the majority of their AI use cases are still stuck in the experimental phase
- HR (37%) and finance (30%) are considered the departments where AI has the least tangible advantages
James Fisher, Director of Strategy at Qlik, says that this data does not mean that AI is all the media threshing.
“It is because many companies do not have the data ready to fuel these projects with their full potential. As we all know, the quality of the data used for AI projects determines whether it flows or swimming, ”he explains.
“Rather than noting AI as a media threshing that does not provide trade advantages, we must take a step back and obtain our data and analysis foundations in order, define clear use cases and evolve the adoption of AI as projects show their value.”
Adoption problems
There are also sectoral evidence suggesting a slowdown. According to a recent study by STL Partners Telecoms consultants, a total of 62% of new Genai telecommunications projects aim to monetize AI through business or consumption solutions.
However, the telecommunications market experienced a slowdown in new announcements, noted STL. Between June and August 2025, the number of Genai projects announced by telecommunications operators increased by only 12% – much less than the 41% increase in the previous period of three months.
According to Digitate Ugo Orsi’s customer director, Genai is currently in the stadium where he demonstrates his capacities for businesses, and that is why there was so much media threshing, which could have the impression of losing momentum.
“What we have is an adoption problem, not a problem of capacity: that is to say that companies do not know how to unlock the full potential of the AI or worse generation, others do not have the courage to start the trip,” explains Orsi.
“The essential is that the” media “perceived at the beginning was fueled by the” curiosity “of companies to understand the art of the possible. Now we are in a different time: adoption that takes place at a much slower pace. Gen Ai / AI is always incredible technology, and companies need time to adopt. »»
The founder and CEO of Squirro, Dorian Selz, agrees that we are going from a step where many companies launched pilots, to test the capacities of the Genai, in a new phase, focused on the adoption focused on the value.
“The idea that the AI of media threshing cycle is about to burst the situation as a whole,” says Selz. “Yes, the enthusiasm of AI pilots will be strongly cut. Many of these pilots have been launched without clear profitability analysis.
“Companies go beyond experimentation and focus more on the return on investment, integration and risk management, which take time and discipline. And it’s more difficult than many think it. “
The impact of the regulations – or the lack of clear frameworks – could also slow down adoption, add Selz, because many major organizations retain the scaling of AI projects because of the concerns about the responsibility, conformity, confidentiality of data and ethical use.
Selz adds: “We continue to see a high demand for AI solutions that provide measurable results, especially when governance and transparency are built from the start. Far from bursting, the AI cycle enters a more mature phase which favors more sustainable growth and responsible for short -term excitement. ”
This point of view was supported by Glean Chef of Solutions Engineering Emea Gavin Guinane, recognizes that many organizations had rushed into AI pilots in the last 18 months – with only a fraction which took them on a scale.
“It is not a bubble glow – it is a necessary break because the leaders focus on the construction of sustainable foundations for long -term success,” says Guinane.
“Companies discover that the transformation of AI does not concern flashy demos. This is confidence, security and adaptability. Systems must provide reliable results, protect sensitive information and work transparently between the tools that employees already use. Without these fundamentals, adoption will be holding up.
“The real story is not a retirement, but of the deadline. Companies go beyond media threshing cycles and invest in AI which can evolve in a responsible manner, integrate deeply and offer a measurable impact. Those who focus on these bases will be those that will transform pilots today into sustainable transformation. ”
In the second part, we will examine why some figures in the industry believe that the threshing media cycle does not have fun and AI is ready for even more in -depth growth in the years to come
