The rapid expansion of artificial intelligence infrastructure in the United States could create new financial risks as companies increasingly rely on external financing to fund massive data-center projects, according to research by Stijn Van Nieuwerburgh, a finance and real estate professor at Columbia Business School. Van Nieuwerburgh estimates that the artificial intelligence buildout could require about 3.6% of United States gross domestic product each year through 2032, translating into more than $10 trillion in investment. The scale would exceed the annual economic resources used during major infrastructure expansions such as railroads, interstate highways and telecommunications. Companies including Amazon, Meta Platforms and Alphabet’s Google initially funded much of the expansion through accumulated corporate cash. However, the growing investment requirement is pushing the sector toward outside financing involving technology companies, banks, private credit lenders and real estate firms. The research estimates that around 183 gigawatts of new data-center capacity could be developed over the next seven years, compared with approximately 57 gigawatts currently installed. Van Nieuwerburgh said the increasing use of complex financing structures and special-purpose entities is making the financial system more interconnected. The study also highlights uncertainty surrounding future artificial intelligence revenues. It estimates that the industry could need approximately $3.7 trillion in annual revenue by 2032 to generate the expected return on investment. With current combined annual revenue estimates for OpenAI and Anthropic at around $100 billion, the research indicates that revenues would need to expand at roughly 80% annually.
Read more here
