The AI financial bubble risks becoming the largest in modern economic history: 8 times larger than the subprime mortgage crisis of 2008 and even 17 times larger than the dot-com crisis of the late 1990s. This is the gist of a note, with a decidedly pessimistic flavour, written by the expert Julien Garran, partner of MacroStrategy. And Garran wasn’t the only one to express some concern about AI. Economists and the most authoritative financial institutions are divided between those who speak of speculative excess and those who believe that it is a phase of physiological growth of a technology destined to profoundly change global productivity. In this in-depth analysis we will examine what those who see an imminent risk, such as the aforementioned Garran and the CEO of, say JP Morgan Jamie Dimon, and what the most cautious analysts, such as those at Goldman Sachs, indicate instead.
AI bubble: the opinion of economists
Returning to the analysis company’s study MacroStrategy Partnershipwho calculated the size of the so-called “AI bubble”, let’s see some of the macroeconomic indicators he combined to arrive at saying that the AI bubble risks being the largest in history. To understand the significance of this calculation we must return to the thought of the Swedish economist Knut Wicksell, who lived in the 19th century, who argued that capital was allocated efficiently only when the average cost of debt for companies exceeded nominal GDP growth by approximately 2 percentage points. Today, after years of low interest rates and expansionary monetary policies, this proportion has altered, generating a “Wicksellian deficit”: a measure of the share of capital invested inefficiently. According to Julien Garran, this “misallocated” part of GDP includes the huge flow of money to AI, including real estate, NFTs and venture capital in the calculation. This is how we arrive at the number we entered with: an AI bubble 17 times larger than the dot-com bubble.
Garran’s analysis goes further, identifying a technological limitation that he says is already visible in large language models, or LLMs, such as ChatGPT. According to Garran, the costs of training these systems grow exponentially while the performance improvements decline rapidly. GPT-3, for example, would have cost around $50 million; the subsequent GPT-4 cost 10 times as much; the GPT-5 model, the latest made available by OpenAI, with an estimated investment of 5 billion dollars, would not have shown progress commensurate with the increasingly large investments.
This is why many people express concerns similar to those of Garran and believe that the current euphoria about AI is reminiscent of that which existed between 1998 and 2000 for the so-called dot-coms, when it was enough to add “.com” to the name of a company to make its value skyrocket on the stock market. Today the same dynamic seems to repeat itself with the term “AI”. The director of IMF Kristalina Georgieva warned of the risk of “overvaluation of technology assets” driven by excessive optimism about the potential of future productivity. Georgieva declared:
Spurred by optimism about the productivity-enhancing potential of artificial intelligence, global stock prices are rising. If a sharp correction occurs, tighter financial conditions could dampen global growth.
Even the Bank of England expressed concern about the level of market concentration: the top five companies in the S&P 500 index now represent around 30% of the overall value, the highest share in the last 50 years. According to data compiled by analyst Howard Silverblatt, seven companies – Alphabet (the holding company that controls Google), Amazon, Apple, Meta, Microsoft, NVIDIA and Tesla – alone generate more than half of the earnings of the entire index. In other words, much of US economic growth depends on very few players, almost all of whom are involved in the development or use of artificial intelligence, and if something were to go wrong in the AI market, the damage to the global economy could be anything but negligible.
The CEO of JP MorganJamie Dimon, warned that this concentration could result in a «significant correction» of the stock market over the next two years. Dimon does not doubt the reality and long-term value of AI, but believes that «most of the people involved will not succeed», just as happened in the Internet market at the beginning of 2000.
Could the AI bubble burst? Goldman Sachs’ position
On the other hand, Goldman Sachs, another well-known investment bank in the sector, adopts a more balanced position. While he recognizes signs of overvaluation, he does not believe that a bubble burst is around the corner.
In this regard, in fact, Peter Oppenheimer, member of the board of directors of Goldman Sachs, declared:
There are elements of investor behavior and market prices currently that rhyme with previous bubbles.
Despite this veiled optimism, Oppenheimer also added:
While it appears we are not yet in a bubble, high levels of market concentration and increased competition in the AI space suggest that investors should continue to focus on diversification.
AI: between innovation and speculation
What emerges, beyond estimates and future predictions, is that the artificial intelligence sector today finds itself at the crossroads between innovation and speculation. The record numbers, stock market valuations and unprecedented capital inflows signal extraordinary confidence in the sector’s potential, but also an intrinsic vulnerability: if earnings expectations are not met, the correction could be abrupt. The key could therefore be to learn to distinguish between AI as a transformative technology (which will continue to evolve and find concrete applications) and AI as a financial phenomenon, where enthusiasm risks inflating numbers and expectations well beyond what is the economic reality.
