Never before has so much money flowed into a new technology as it is now with artificial intelligence, surpassing the amounts invested in railways or the internet when those technological revolutions attracted capital.
اضافة اعلان
Global spending on data centers alone could exceed $30 trillion by 2050, according to projections by PwC, bringing it close to the value of outstanding US Treasury bonds. The company said this spending would be “far greater” than what was spent during the railway and internet booms, even after adjusting for inflation.
Meanwhile, Anthropic, one of the major companies competing in the AI race, plans to spend $518 billion over the coming years, according to an initial public offering prospectus seen by Reuters. That figure is more than 100 times its revenue in 2025. The company’s backers argue that AI will bring a transformation greater than that produced by the emergence of steam engines and industrial manufacturing.
But behind the staggering forecasts, massive spending by AI companies and extremely high valuations lie assumptions about broad productivity gains and future profits, without sufficient evidence so far or historical precedents confirming that they can be achieved, economists say.
JPMorgan wrote in August that broad-based productivity gains in the United States, which is leading the AI race, “remain elusive,” raising questions about the sustainability of AI companies’ valuations.
A study by Bain & Company concluded that productivity gains in existing markets would not be enough to justify current spending, and that “entirely new markets must emerge to close the funding gap.” These could include AI-powered robots or the development of new materials for batteries and semiconductors.
Bain said major US cloud computing companies, including Google, Amazon and Microsoft, along with other companies participating in the AI race, need to generate more than $4.2 trillion in additional revenue over the next five years to finance infrastructure construction.
“The question is whether applications will emerge in time to cover their cost,” the study, published last month, said.
Few doubt AI’s ability to transform everything from office work to research laboratories, much as previous revolutions cut travel times from days to hours or connected the world with the touch of a keyboard.
But what appears more certain are the calculations required to ensure a return on investment, or the deadlines for repaying loans. This is prompting economists to examine the impact of the AI boom on the global economy beyond the fluctuations of investment cycles.
JPMorgan wrote: “Historical precedents suggest that technology-driven booms often end when infrastructure expansion no longer generates sufficient returns.”
Using Nvidia as an example, the US company whose chips underpin the AI revolution, the bank estimated that justifying its valuation would require US productivity to grow by between 3% and 5% annually over the next decade.
That would represent a significant increase from the Congressional Budget Office’s baseline forecast of 1.75% annual growth over the same period.
For the United States alone, which accounts for roughly three-quarters of total global AI investment according to some estimates, investment could reach about $9 trillion between 2025 and 2032. That is equivalent to spending 3.2% of US GDP annually, according to Stein van Nieuwerburgh, an economist at Columbia Business School.
Van Nieuwerburgh estimates that the US AI sector would need to generate annual revenue of about $3.55 trillion by 2032 to achieve a 10% return on investment. Its current revenue represents only a small fraction of that amount.
Moreover, the heavy reliance on leveraged debt to finance much of AI infrastructure means that “a relatively modest decline in demand, delays, or a decline in asset values could result in much larger losses,” according to van Nieuwerburgh, who wrote this in a conference paper revised in October.
AI and the “pace of new wonders”
The staggering figures have not stopped leaders of US AI companies from speaking with extraordinary enthusiasm about the changes ahead.
Anthropic CEO Dario Amodei said the future of AI could be “something of indescribable beauty.” OpenAI CEO Sam Altman said that “the pace of achieving new wonders will be enormous” as models learn to improve themselves and accelerate discoveries.
Jasjit Singh, Google DeepMind’s chief strategy officer, said during a summit at the University of California, Berkeley, in August that this self-improvement, known as “iterative self-improvement,” is “a key part of the investment thesis.” If achieved, he said, it could deliver unprecedented productivity gains, according to Reuters.
Although iterative self-improvement could lead to enormous and accelerating advances in AI, it has also raised concerns about existential risks to humanity.
Even so, productivity improvements may remain slower than the timelines required by corporate finance departments.
Diane Coyle, an economist at the University of Cambridge, said the impact of previous transformative technologies on productivity typically took between 10 and 50 years to emerge.
Anthropic’s economics team modeled different scenarios for the amount of additional growth AI could generate in 2030. Assuming baseline growth of 2% without AI, the team estimated growth would reach 2.4% in a limited-impact scenario, 5.4% in a high-impact scenario and 15.4% in the most extreme scenario.
The company said higher growth would mean greater job losses, without assigning probabilities to any of these outcomes.
Amodei predicted last year that AI would eliminate half of entry-level white-collar jobs within five years. But some researchers say its impact so far appears to have been limited to making it harder for people seeking office jobs to find employment.
Studies in the United States and Britain have pointed to slower early-career hiring for office jobs involving tasks that AI performs well, despite overall hiring remaining strong.
Researchers at Stanford University said in August that employment among workers aged 22 to 25 in AI-exposed sectors such as accounting and legal assistance had fallen 19% compared with jobs that are harder for AI to replicate, such as cleaning and construction.
Even if the promised transformation takes longer than the figures associated with AI companies suggest, its genuine economic benefits are likely to remain, just as trains continued operating after the Panic of 1873 bankrupted major railway investors, and the internet did not disappear after the dot-com bubble burst in the 1990s.
Coyle said: “History is our ally in trying to understand what is happening. As long as the infrastructure needed to support future productivity gains remains in place, things will be fine.”
Resource: Al-Ghad.
