Investments in artificial intelligence (AI) are unprecedented, surpassing previous technological booms like railways and the internet. A report from PwC predicts that global spending on data centers could exceed $30 trillion by 2050, nearly equal to the total amount of US Treasuries. Companies like Anthropic plan to invest $518 billion in the coming years, which far exceeds its projected revenue, based on the belief that AI will revolutionize industries more than earlier inventions did.
However, behind these impressive projections and high valuations are uncertain assumptions about productivity gains and profits, with economists noting there is little evidence to support these expectations. JP Morgan pointed out that broad-based productivity improvements in the US, a leader in AI, are still “elusive,” raising doubts about the long-term sustainability of current investment levels. A study by Bain & Company highlighted that existing markets alone cannot justify current spending and new market opportunities need to be created to bridge the funding gap. US hyperscaler companies like Google, Amazon, and Microsoft need to generate over $4.2 trillion in new revenue in the next five years to support ongoing AI infrastructure development.
The urgency of developing applications that can justify this investment raises further questions, as previous technology booms often ended when infrastructure investments failed to deliver enough returns. JP Morgan’s analysis estimated that for Nvidia, central to the AI movement, US productivity has to increase by 3% to 5% annually over the next decade, significantly above the current expectation of 1.75% growth. For the US, which constitutes a large portion of global AI investments, predicted spending could reach about $9 trillion from 2025 to 2032, representing 3.2% of yearly GDP.
Anthropic’s team suggested AI should generate around $3.55 trillion annually by 2032 to yield a 10% return, yet current earnings are far below that. The use of debt for AI investment means that even small drops in demand or value can result in significant losses. Despite these financial concerns, leaders in the AI field speak optimistically about its capabilities. Dario Amodei from Anthropic describes an AI future as potentially “a thing of transcendent beauty,” and Sam Altman from OpenAI forecasts immense progress through advanced AI models that improve themselves.
The concept of recursive self-improvement in AI offers hope for exceptional productivity gains but also raises serious concerns about risks to society. Past technological revolutions typically took decades for their productivity benefits to materialize fully. For instance, economists have noted that the benefits from technologies like railroads and the internet unfolded gradually, requiring patience for any significant economic impact.
Models predicting AI’s growth assume varying scenarios of impact by 2030, ranging from mild to extreme productivity increases. However, projections indicate higher growth could displace many jobs, particularly entry-level white-collar positions. Research shows that employment in AI-affected industries for younger workers has dropped significantly, even with overall job numbers holding steady.
Overall, even if the AI transformation takes longer to achieve, the infrastructure developed will likely offer lasting economic benefits. As history has shown, technologies continue to function and evolve even after market flucuations, suggesting a promising future for AI.
With information from Reuters
