The global artificial intelligence (AI) boom is spreading into a $3 trillion (about 4,200 trillion won) wave of data center investment. Morgan Stanley projected that global data center expenditure will approach $3 trillion by 2028, and analyzed that $1.5 trillion, or half, will be covered by private credit and other liability. This is fueling concerns that the tech stock rally could spill over into a credit-driven overheating.
According to the Guardian on the 2nd (local time), AI infrastructure investment is being led by global big tech such as Nvidia, Microsoft, Google, and Meta. Nvidia became the first in the world to surpass a market capitalization of $5 trillion, standing at the peak of the AI boom. Microsoft and Apple also both topped $4 trillion, and OpenAI was valued at $500 billion. The industry expects the value of Microsoft's equity stake in OpenAI to exceed $100 billion, and the possibility of a $1 trillion listing next year is also being discussed.
Buoyed by demand for AI infrastructure, Alphabet, Google's parent, surpassed $100 billion in quarterly revenue for the first time. Apple and Amazon also posted strong results as they expanded AI-related services. Yet despite these results and expectations, the revenue structure of the AI industry has not been firmly verified. There is an assessment that it is unclear how steadily corporations, governments, and consumers can generate revenue through AI services in practice.
Because of this, criticism is mounting in the market that "investment is outpacing revenue." This means that at a stage where a money-making structure has not yet been established, huge sums are being injected first on the back of expectations for future growth. Joe Tsai, Alibaba's chairman, warned that "data center projects are increasing that are driven by money without customer acquisition," adding that "a bubble is beginning."
U.S. company Meta raised $29 billion in the private credit market to expand its Louisiana data center. Morgan Stanley expects private credit to fill more than half of total investment funds. However, because private credit falls in the realm of "shadow banking," which is not subject to financial authorities' regulation, there are concerns that default risks could spread rapidly if the economy slows. Gil Luria, head of technology research at U.S. investment firm DA Davidson, said, "Some AI infrastructure investment is speculative asset pushed forward without real demand," adding, "If liability swells into the hundreds of billions of dollars, structural risks could arise across the global economy."
OpenAI's ChatGPT, the symbol of the AI boom, has 800 million weekly active users, fanning optimism. But research by the Massachusetts Institute of Technology (MIT) found that 95% of corporations investing in Generative AI projects have not yet generated revenue.
Nevertheless, massive funds are being poured into expanding AI infrastructure worldwide. Real estate consultancy JLL predicted that construction will begin this year alone on new data centers totaling about 10 gigawatts (GW). That is equivalent to one-third of the United Kingdom's total power demand. Goldman Sachs forecast that by 2030, global data center capacity will double from the current 59 GW, and analyzed that an additional $720 billion in power grid investment will be needed to handle it.
In line with this trend, warnings are also emerging about an AI bubble bursting. Harris Kupperman, founder of U.S. hedge fund Praetorian Capital, said, "The value of data centers will decline at twice the pace of the revenue generated from them," warning that "if the AI frenzy does not translate into long-term revenue, it could follow in the footsteps of the dot-com bubble."