Topline
The AI buildout could require so much borrowing that Wall Street’s bond market may not be able to finance it, creating a roughly $1 trillion gap that could be filled by private credit, Apollo’s chief economist said Friday.
The world’s largest companies have pledged hundreds of billions of dollars to match growing demand for AI.
AFP via Getty Images
Key Facts
Torsten Slok, Apollo’s chief economist, wrote Friday that AI-related borrowing accounts for more than 40% of new long-term, investment-grade corporate debt, and that Apollo estimates the “AI ecosystem” could support more than $2 trillion in debt.
Investment-grade markets may be able to absorb less than $1 trillion of that amount through 2030 because of “concentration and ratings constraints,” Slok wrote, indicating investors may already have significant exposure to companies driving the AI buildout and that firms can only take on so much additional debt before ratings agencies view them as riskier borrowers.
The gap, totaling more than $1 trillion, could be filled by private lenders and other financing backed by infrastructure, equipment and individual AI projects, Slok said.
Slok noted the private financing deals could offer lenders better protection by tying the debt to specific assets or contractual guarantees, unlike traditional unsecured corporate bonds.
big number
$220 billion. That’s the highest capital spending projection among “Magnificent Seven” companies, with Amazon taking the top spot ahead of Alphabet ($205 billion) and Microsoft ($175 billion). Amazon, Alphabet, Microsoft and Meta combine for $738 billion in projected spending this fiscal year.
what to watch for
How companies escalate plans to build more AI infrastructure. Nvidia and OpenAI have reportedly discussed a data center near Columbus, Ohio, that could cost more than $500 billion, with OpenAI discussing a chip purchase valued up to $350 billion. The project could become the largest such project by power capacity in the world by far: A proposed 10-gigawatt site would dwarf the power output of other planned projects, including a $20 billion, 3.2-gigawatt center in Effingham County, Georgia, announced by OpenAI in July.
key background
AI infrastructure has increasingly become a larger source of corporate borrowing, as mega-cap tech firms and data center operators raise billions of dollars to back new projects. Traditional bond markets have limitations on how much debt they can absorb, however, largely because investors can only take on so much exposure to the same companies or industry. Private lenders face fewer restrictions and can instead make loans backed by assets or revenue, such as data centers or their future income.