AI investment boom faces a test of future returns

PwC projects global data-center spending could exceed $30 trillion by 2050, almost equal to the value of outstanding U.S. Treasuries[1]. PwC says the projected spending would dwarf railroad and dot-com-era spending even after inflation adjustments[1]. The report also notes that Anthropic’s planned spending is more than 100 times its 2025 revenue[1]. These figures describe the scale of the plans; their underlying assumptions include broad-based productivity gains and future profits, for which there is little evidence so far[1].

Bain & Company says productivity gains in existing markets would not justify current AI spending, so entirely new markets would need to emerge[1]. It identifies AI-guided robots and new materials for batteries and semiconductors as possible examples[1]. Bain estimates U.S. hyperscalers and other AI companies would need more than $4.2 trillion in new revenue over five years to fund the buildout[1]. JPMorgan estimated that U.S. productivity gains would need to reach 3% to 5% annually for 10 years to justify a valuation, compared with the CBO’s 1.75% baseline expectation[1].

Cambridge University economist Diane Coyle said productivity effects from past revolutionary technologies typically took 10 to 50 years to emerge[1]. Anthropic’s economics team modeled 2030 annual growth of 2.4% with modest AI impact, 5.4% with substantial impact and 15.4% with extreme impact, against a 2% non-AI baseline[1]. The team assigned no probabilities to those scenarios and said higher growth would mean more jobs lost[1]. JPMorgan has also cautioned that technology booms can end when infrastructure buildouts stop producing sufficient returns[1].

Anthropic CEO Dario Amodei described an AI future as potentially “a thing of transcendent beauty,” while OpenAI CEO Sam Altman said new wonders could arrive rapidly as models improve themselves and accelerate breakthroughs[1]. Google DeepMind strategy chief Jasjeet Sekhon called recursive self-improvement a key part of the investment thesis, while noting that it could deliver unprecedented productivity gains if achieved[1]. Amodei forecast that AI could eliminate half of entry-level white-collar jobs within five years[1]. Stanford researchers reported employment among workers aged 22 to 25 in AI-exposed industries was 19% lower than in jobs they considered hard for AI to replicate[1].