Is artificial intelligence a bubble?

ChatGPT, Open AI, AI slop, Italian brainrot, profile pictures ghiblized, chatbot assistants… Artificial intelligence is everywhere on everyone’s lips and will conquer the world. A hot topic that we have been following since 2023, observing the mainstream trends and listening to the insiders’ rumours. In the midst of the party someone is talking quietly about a speculative bubble, but many have not paid attention because the noise is deafening. What lies behind this mass euphoria? Let’s find out together.

Small disclaimer: this article speaks of generative AI and not of AI understood as machine learning. The latter is a technology that has been known since the 1950s. However, for the sake of brevity, we will simply speak of AI, going along with the generic name used in the mainstream debate.

AI for the masses, the tip of the iceberg

If you look at the 2023 Will Smith eating spaghetti first video and compare it to the 2025 version, the image improvement made by generative AI is incredible. Some would say ‘in just two years we got here!’, while others might ask ‘in two years we only got here?’.

Apart from the obvious improvement, in the eyes of an attentive observer, in the mainstream AI still seems to have little concrete value. Image generators have ‘gifted us’ with the AI slop in its infinite shapes (have you come across the fake animal rescues on Facebook? I hope not), the Italian brainrot and other more or less undesirable ‘side effects’, such as the increasingly popular Youtube thumbnails, the blog article headers and other Instagram graphics, which are all too often impersonal and now sadly recognisable.

These are certainly phenomena that have already affected the lives of many, but without any real revolutionary thrust. On the contrary, one could reflect on their possible damaging effect on the reliability of information and the distortion of perception by the less attentive (did someone say scams on old people?). This is not to mention the unresolved ethical problems related to the copyright of the creative material involved in the process. You all have heard of the Studio Ghibli case, haven’t you? You can find some reflections in this video by Matteo Flora, but there are many others.

If we want to stop at the surface, the impression is that it all boils down to a fading wave of entertainment for the masses, which will pollute the Internet like an oil slick and which will soon – hopefully – exhaust its strength, as it has done to so many other memes and trends that have kept us company over the past fifteen years. Broadly speaking, this is evident from the speech of American Youtuber The Art Mentor. His recent video AI art’s been dying for 2 years + I predicted it in this regard is worth watching.

Another impression is that this technology could lead us to a new approach to the production of creative content. A hasty, homologated and industrial approach, with more exaggerated and amplified (but less creative) workflows than social networks have already accustomed us to for many years.

For those who frequent platforms such as Facebook, Instagram and TikTok, daily exposure to such a quantity of mediocre content, not at all game changing and of little value to companies that enjoy a solid brand identity (at best good for a few social posts), almost makes it seem as if the AI has already fired all its bullets. Blanks, at that. And here a doubt arises, a little voice whispering in our ear: is all this enthusiasm legitimate or are we dazzled by the flash of a straw fire?

Is this really all there is? Is the AI a speculative bubble or is there substance? As the title suggests, this is but the tip of the iceberg of a more articulated reality. The world is vast, there is room for markets based on creative products of different qualitative value, and let’s not forget that content, however questionable or incomprehensible, still becomes a source of income for some. Moreover, there are not only images, videos and texts.

To understand whether this is a bubble, one has to go deeper. On the business side, there are many other applications that could help people. But will these solutions really have a profound impact? And how pervasive will they be?

The AI case according to some experts from the financial world

Rather than starting with our own personal opinion, below are some articles that explore the feelings of a segment of the financial world. Of course, the Internet provides many other points of view, but for reasons of time we have had to limit ourselves to those we consider most useful in counterbalancing mainstream enthusiasm. Therefore, we do not consider this brief summary exhaustive of the subject and invite you to document further.

In a article from July 2024, Vahid Karaahmetovic reports that, according to the investment fund Jefferies, the huge investments in AI contrast with the projected profits for 2025, already painting a problematic financial picture in the eyes of experts. These predict more rigorous investor demands on companies to account for their returns by mid-2025. The time is ripe to see what will happen from June onwards. Jefferies also pointed out the presence of various obstacles: lack of promising business models and strategies, as well as the high electricity costs dictated by the servers employed by AI.

In August 2024, a article by Riccardo Cantadori reported that the world-renowned investment fund Elliott Management had recently emphasised the overvaluation of this technology, in relation to the capitalisations involving Nvidia at that time (processors are essential to make AI perform). Some financial indicators, he writes, were not giving good signals. Cantadori specifies that, 'as in the case of dot-coms, the uncertainties worrying investors concern not the usefulness of artificial intelligence, but how to quantify its economic potential’.

In October 2024, Malcolm Hawker shared a more optimistic article. A reflection on the comparison between the dot-com bubble, which the author experienced, and the AI situation, with their obvious similarities and differences. Certainly, the consequences of a negative outcome will not be as well disastrous. However, the actual ability of companies to build large-scale AI solutions remains the question with the greatest short-term risks.

Although some of the articles cited are already eight months older, by January 2025 the situation did not seem to have improved much. According to John Rau of Forbes this is not a bubble, but he wonders whether the investment needed today to play the game will really yield a return substantial enough to win the bet. The large energy consumption, the scarcity of new data for training, the absence of a must-have app for the masses and future legislation to regulate AI and protect copyright are all aspects that could undermine its future. The journalist predicts the failure of countless startups and investors while only a small group of four or five players who will survive in the long run will succeed.

The AI bubble according to the Reddit community

Other interesting insights come from the Reddit platform, where it is known to find a slice of users from the tech world, closer to the realities where AI is studied for real than those who observe these phenomena from the outside.

Around September 2024, user ThinkBigger01 polled the Reddit community. With a few exceptions, the common sentiment emerging from the group’s comments is that this is not a bubble, although this technology has not yet expressed its full potential. There is talk of overestimation by people: from believing that the AI can do without human input to achieve appreciable results, as user Otaehryn writes, to overestimating the ubiquity it will be able to achieve in all sectors, according to IntegerSpins. Adding that, on the contrary, the most obvious and profitable impact will involve only certain niches.

dSolver and redhtbassplyr031 speak of the possible bubble of LLMs (Large Language Models), such as ChatGPT, but shifting attention to other uses of the AI that could bring different benefits, ignored by the general public today. iceland00 and others speak of reducing costs of scale for companies, as in the case of large call centres. RedditMapz speaks of similarities with the dot com, augmented reality and blockchain, referring to many companies attempting to make money from this technology and which will inevitably disappear as the market collapses, in favour of a few winners.

JPMorgansStache argues that many signs point in the direction of a recession, starting with the value of the hypothetical AI market. This, to be profitable, would have to be exponentially higher than the value of the enormous funding received by, for example, Open AI. A value in the billions, which the user finds hard to believe.

What’s Dreamonkey opinion

The general perception is that, on the ‘surface’, there is a lot of buzz. To varying degrees, the topic interests many of the companies we have come into contact with and who frequent the tech world. There are events, discussions and chats at the coffee machine. We ourselves have given space to some experiments at the office and, for the ClubD of Unindustria, our member Paolo Caleffi has coordinated the organisation of some events dedicated to the topic. Dal Laboratorio alla Fabbrica: l’Intelligenza Artificiale nell’industria reale, Super Calcolo e Intelligenza Artificiale: la nuova rivoluzione industriale passa da qui, Casi d’uso di AI applicata ai processi aziendali, Casi d’uso di AI applicata ai processi aziendali - 2° incontro.

The ground seems fertile for sowing but, on closer inspection, other signs herald a much more limited change in the Italian SME fabric. Or at least an evolution that is slow in coming and has been typical of other technologies and investments that have never really taken off, except in limited cases.

In nine years of experience, we have often seen tech phenomena go from the stars to the stables. Augmented reality, cryptocurrencies and blockchain are three examples of technologies that have been the talk of the town but which, on balance, have not left a lasting imprint, disrupting businesses and changing the habits of the world’s population. Least of all the Italian population or the corporate world. Apart from all the in-depth investigations that could be done to better analyse the numbers, the impression remains of an echo of great unfulfilled promises.

Even Industry 4.0 investments, so promising for the upgrading of production companies, have not had all the hoped-for effects. Especially when it comes to the conscious and profitable use of the data collected by machines. This is a subject we are familiar with and often talk about in our industry.

Other digital trends, such as websites with parallax effect (remember them?) and Corporate Memphis graphics (or Globohomo) have unfortunately plagued us for years but, once the fever dream was over, we have gradually got rid of them, at least in part…

Despite the scepticism, in our opinion, artificial intelligence still has some advantages over the compared cases. Yet, beyond all the hype and chatter, one of the problems we have noticed is the poor number of projects that go beyond the simple chatbot assistant. We ourselves continue to question where AI is able to bring concrete value with interesting, out-of-the-box ideas. The companies around us seem to face the same difficulties.

Another aspect, too often ignored by companies that have jumped on the hype train, is that many processes that today one would like to optimise by means of AI are in fact optimisable with the much more classic and tried-and-tested digital automation solutions, with software of various kinds. Both paths involve substantial investment and the entrepreneurial will to restructure one’s business processes but, while “classic” software guarantees deterministic output, AI (we repeat, especially generative AI), by its very nature, does not. The accuracy of machine learning is much higher, but this would deserve a separate article.

The subject is technical and would deserve space in a separate article. Suffice it to say that those wishing to adopt AI must put in long and costly research and development sessions, studded with problems related to training and testing databases, which may be insufficient or poor quality. In short, the investment could end up in failure.

Conclusions

Perhaps it is not a bubble as such but, in our opinion, the signs are not entirely heartening. We cannot yet say for sure if and when artificial intelligence will reach its potential but, as we await the financial implications of the second half of 2025, we suggest a cautious attitude. There is a long way to go and, like any technology, if it does not find a way to return on investment and generate profit, it will be doomed to fall into disuse.

If it finds sustainable business strategies and a solution to its energy problems – and it is not a foregone conclusion that it will – perhaps it will find its markets. It will, however, have to dribble out other current and emerging drawbacks, such as new restrictive laws, copyright lawsuits and data training exhaustion. These also include the so-called AI data poisoning, that is the malicious corruption of databases used for artificial intelligence training, in order to boycott it (on which, however, we have not conducted in-depth investigations and do not know whether the phenomenon is objectively relevant).

On the Italian front, once the more or less successful experimentation has passed, we believe it is likely that many R&D teams will return to other things and, gradually, SMEs will lose interest or adopt solutions from other big corporations. Some will experiment on their own, but many AI tools will drop from above, as was the case in the past with other software for which in-house development was not deemed effective or advantageous.

If you are interested in finding out more about this new technology, contact us. We can offer you interesting insights and recommend the most suitable software solutions for your projects. We also offer AI consulting, integrations with LLM, RAG systems and other solutions. Who knows, maybe your problem can be solved even without AI!