Will AI Crowd Out Jobs? (paulkrugman.substack.com)

A new paper by Stijn Van Nieuwerburgh raises significant concerns about the impact of artificial intelligence (AI) on employment, particularly in light of the massive capital investments in data centers that power AI models like ChatGPT and Claude. Van Nieuwerburgh estimates that establishing a 200 MW AI training campus costs approximately $8.2 billion, which temporarily creates a sizable number of construction jobs but ultimately leads to only 50 to 150 permanent jobs once operational. In contrast, similar investments in non-AI sectors could support over 10,000 jobs, suggesting that AI development "crowds out" traditional job creation. The research highlights the potential worsening of income and wealth inequality in the U.S. as the capital-intensive nature of AI undermines labor demand. With hyperscalers expected to invest over $10 trillion in data centers by 2032, the predicted outcome is a loss of up to ten million jobs compared to what would have been created through traditional investments. While the actual unemployment might not rise by that number, the diminished demand for labor is likely to push wages down and exacerbate existing disparities in economic power, benefiting capital owners at the expense of workers. This paper calls for a reevaluation of the implications of AI on the labor market, underscoring the urgency of addressing these troubling trends.