Never has so much money been directed toward a new technology as is currently being invested in artificial intelligence (AI). The scale of investments already exceeds the amounts invested in previous technological revolutions, such as railways and the internet — and is beginning to raise questions about how long it will take for the investments to generate returns.
According to a PwC projection cited by Reuters, cumulative global spending on data centers could surpass US$ 30 trillion (about R$ 159 trillion) by 2050. The value approaches the stock of U.S. Treasury securities, and according to the consultancy, exceeds the investments made during the expansion cycles of railways and dot-com companies, even after adjusting for inflation.
Anthropic, one of the companies at the center of this race, plans to spend US$ 518 billion (about R$ 2.75 trillion) in the coming years, according to its prospectus for an initial public offering (IPO) analyzed by Reuters. The value is equivalent to more than 100 times the revenue recorded by the company in 2025.
The defenders of these investments bet that AI will have a greater economic impact than the transformations caused by the steam engine and industrialization.
The problem is that, for now, most of this bet depends on productivity gains and future profits, for which there are still few concrete evidence — and little historical precedent capable of guaranteeing that the results will appear at the necessary pace.
Productivity still does not meet expectations.
JPMorgan stated in August that broad productivity gains in the United States, the country leading the race for AI, “continue to be difficult to find”, raising doubts about the sustainability of sector company valuations.
A study by Bain & Company reached a similar conclusion. According to the consultancy, the productivity gains achieved in existing markets would not be sufficient to justify current investment levels. To close this gap, the emergence of entirely new markets would be necessary.
Among the possibilities are the use of AI-guided robots and the development of new materials for batteries and semiconductors.
The companies responsible for the AI infrastructure in the United States — including Google, Amazon and Microsoft — and other firms involved in the race would need to find more than $4.2 trillion (about R$22.3 trillion) in new revenue over the next five years to finance the expansion, according to Bain.
The question, according to the study, is whether applications capable of generating this money will arrive in time to pay for the infrastructure built.
JPMorgan also warned that historically, technological expansion cycles can end when infrastructure investments no longer provide sufficient returns.
The account to justify the investments
- Nvidia, the chip manufacturer, appears as an example of this equation;
- According to a JPMorgan estimate, U.S. productivity gains would need to be between 3% and 5% annually over the next decade to justify the company's valuation. This would represent a significant increase compared to the Congressional Budget Office's annual growth expectation of 1.75% for the same period;
- In the United States, investments in AI could reach approximately $9 trillion (about R$47.7 trillion) between 2025 and 2032, according to Columbia Business School economist Stijn Van Nieuwerburgh; The value would equate to 3.2% of the United States' Gross Domestic Product (GDP) every year.
OpenAI is one of the AI companies in the equation – Image: Vitor Miranda/Shutterstock
