Emmanuel Roman, chief executive of bond giant Pimco, warned on Tuesday that financing for the artificial intelligence buildout could swell to about $10 trillion over the next five years — and eventually become more than markets can absorb. Roman told the Global Financial Leaders' Investment Summit in Hong Kong the industry would need to raise roughly $2 trillion a year, against a US bond market of about $12 trillion.
"At some point in time, it's going to be too much," Roman said. "At some point in time, there's going to be a downturn, and then things are going to break." He named rising US interest rates, a war in the Middle East or a eurozone crisis as possible triggers, and said that for portfolio construction today, taking on too much risk is the bigger danger — not being too cautious.
Oracle's $20 billion raise brings the warning to life
Roman's warning landed a day after Oracle showed exactly how fast the bill is growing. On Monday, Oracle filed to raise $20 billion through bond and share sales to fund its AI and cloud infrastructure buildout: about $10 billion from selling 86.2 million shares at $116 each, an 11.2 percent discount to Friday's close, plus about $10 billion in notes maturing between 2029 and 2046.
The market's verdict was swift. Oracle shares fell 8.8 percent to $119.22 on Monday, leaving them down about 39 percent this year. The sting was sharper because only last week Oracle had pledged to raise no more than $20 billion in additional AI financing through fiscal 2030 — then used the entire allowance at once. The company had $125.34 billion in notes payable and other borrowings at August 31, against fiscal 2026 revenue of $67.36 billion, and says capital spending will peak this fiscal year and next.
Blackstone: the buildout needs more equity, not just debt
Sharing Roman's panel, Blackstone president Jon Gray argued the buildout will need more equity financing as lenders grow increasingly exposed to the companies driving it. The sheer capital required is creating concentration risks for debt investors, both across the sector and at individual companies, he said — and debt holders do not share in the upside if the bets pay off. "You're playing for repayment and coupon," Gray said.
The numbers behind the debate are staggering. About $550 billion of AI-related debt has been issued so far this year, according to summit coverage, while Goldman Sachs estimated in a Monday report that global data-centre demand will require about $8.5 trillion of capital spending over the next five years. Morgan Stanley chief Ted Pick has put the AI buildout's capital needs at $10 trillion to $15 trillion over the next decade.
Not everyone at the summit is worried
Roman's caution was not the consensus. Goldman Sachs chief David Solomon told Bloomberg Television that deal demand remains robust, State Street chief Ron O'Hanley said it is hard to see AI as a bubble yet, and White House economic adviser Kevin Hassett called AI a well-financed ecosystem. Bridgewater co-chief investment officer Bob Prince said inference supply could rise 30-fold by 2028, requiring about $2 trillion in annual revenue to absorb — while insisting Bridgewater remains heavily invested and "super bullish" on the technology.
Nobody in Hong Kong was doubting demand for AI infrastructure. The argument is about how it gets paid for — and whether a financing stack built this fast, this big, can hold if the economy turns. Oracle's $20 billion morning suggests the industry will keep testing the answer, one raise at a time.
