Artificial intelligence is undoubtedly the undisputed protagonist of the financial and technological markets in recent years. However, while enthusiasm around its potential grows, signals also emerge that call for caution. The comparison with the dot-com bubble of the early millennium is increasingly frequent among analysts and investors, who watch with concern the concentration of value in the so-called Magnificent Seven: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla.
These giants today represent over a third of the S&P 500 index, a share much higher than the 15% held by the main technology stocks during the peak of the internet bubble in 2000. Such a concentration inevitably increases systemic risk.
It is not just a matter of capitalization. During the dot-com bubble era, the rush to invest in telecommunications infrastructure led to an excessive expansion of fiber optic networks, culminating in catastrophic failures when the promised demand did not materialize in the short term.
Today, history seems to repeat itself: the major AI companies are investing hundreds of billions of dollars in building new data centers, with an overall expenditure approaching trillions of dollars, figures once associated only with the GDP of large nations. The question everyone is asking is whether this investment rush is justified or if we are on the brink of a new crisis.
The demand for Artificial Intelligence (AI): beyond the consumer boom
The media attention is often focused on the mass adoption of tools like ChatGPT, which in the month of July alone exceeded five billion visits. However, the true economic impact of AI will be measured based on adoption by both consumers and businesses.
According to the published data by the National Bureau of Economic Research, about 40% of the U.S. population has used generative AI systems by the end of 2024, and 23% have employed them at least once for work in the week prior to the survey. The adoption of AI in the workplace is happening at a faster pace compared to that recorded for the personal computer or the internet in their respective early days, indicating that we are facing a general-purpose technology destined to profoundly transform the economy.
Yet, the path towards a tangible economic return is anything but simple. A study conducted by MIT on over 300 public AI initiatives, more than 50 companies, and hundreds of executives, revealed that 95% of businesses are still not obtaining returns from AI investments. Only 5% of the companies analyzed have been successful, thanks to three key factors: preferring the purchase of ready-made solutions over internal development, integrating AI directly into business units rather than central labs, and choosing tools compatible with existing workflows.
Despite the difficulty in turning AI into concrete value, 90% of companies are seriously considering purchasing AI solutions, confirming a widespread interest that follows the classic hype cycle of innovative technologies.
An emblematic example is that of Bank of America, the second largest bank in the United States, which has allocated four billion dollars to new technologies such as AI. The institution has developed a tool that helps bankers prepare for meetings with clients, retrieving information from different systems and drastically reducing preparation times.
Limits and prospects of current AI models
The expansion of AI usage fuels the debate on its real potential and the sustainability of the current development model. So far, progress has been driven by large language models, which improve with the increase in computing power and the amount of available data. However, some authoritative voices in the sector call for caution.
Richard Sutton, a pioneer in AI, had already observed in 2019 that general methods leveraging computational power surpass those based on human ingenuity and complex heuristics, defining this reality as “The Bitter Lesson”. Recently, Sutton criticized the excessive emphasis on scaling up, suggesting the need for a paradigm shift towards agents capable of learning continuously.
Even Gary Marcus, a well-known critic of the AI hype, has expressed doubts about the latest versions of ChatGPT, arguing that the development model based solely on scale is not the right path. According to Marcus, alternative approaches are needed, which might require even greater investments in research and development.
AI Bubble: between excessive optimism and risk of correction
The debate on the possible AI bubble is becoming increasingly heated, especially when figures like Sam Altman, one of the main architects of the current boom, warn about the risk of an overheated market. Altman and other investors point to sky-high valuations, capital chasing business models that are still untested, and the danger of building infrastructure at a pace exceeding real demand. The concern is not so much about the long-term potential of AI, but rather the inflated expectations that could set the stage for a sharp correction.
The real risk, according to many observers, is to fall into a binary vision, oscillating between irrational enthusiasm and the fear of an imminent bubble, without grasping the nuances of a complex phenomenon. The long-term potential of AI remains enormous, but markets rarely follow linear trajectories. A correction could temporarily slow growth, but at the same time strengthen investment discipline and push towards greater attention to the quality of models and real economic value.
Towards the future: discipline, research, and concrete value
The future of artificial intelligence will depend on the ability to overcome the current phase of hype, focusing on advanced research, improving the quality of models, and targeted investments aimed at generating measurable value for businesses and consumers. Only in this way will it be possible to avoid the mistakes of the past and fully exploit the opportunities offered by a technology destined to redefine our way of living and working.
