Morgan Stanley: AI capital expenditure may reach 10 to 15 trillions USD in the next 10 years

According to Jinse Finance APP, Morgan Stanley CEO Ted Pick stated at the International Financial Leaders Investment Summit that capital expenditures in the artificial intelligence (AI) sector could reach $10-15 trillion in the next 10 years, cementing AI's role as the most significant economic transformation since the birth of the internet. He estimates that leading large-scale technology companies could deploy $800 billion to $1 trillion annually in data center and computing investments over the next 7-12 years. Pick described the current deployment as still in its formative stages, saying, "We're probably in about the second inning of a nine-inning game."

Pick believes that compared to stock valuations, the total cost of $10-15 trillion, while large, is not unreasonable versus the S&P 500's market capitalization of roughly $70 trillion, with $2-3 trillion already implemented or in use. He estimates that productivity gains from AI could increase the United States' real GDP by 50-70 basis points annually.

However, PIMCO CEO Emmanuel Roman warned that AI-related financing could rise to about $10 trillion in the next 5 years, eventually reaching excessive levels. He estimates that annual financing needs to support AI industry capital expenditures could reach around $2 trillion, while the US bond market is roughly $12 trillion in size.

Emmanuel pointed out that, ultimately, the scale of financing "will become too large," and "sooner or later, there will be a pullback and then problems will emerge." He added that rising US interest rates, war in the Middle East, and a eurozone crisis could all trigger a turning point. State Street Chairman and CEO O’Hanley commented that it is currently "very hard" to call AI a bubble, since market demand is supporting substantial investment: "These data centers are not being built at random — there are genuine orders behind them, and in most cases, there are real companies backing them." He believes the real question is whether AI will eventually become so efficient that all the capacity under construction today will no longer be necessary.