Citadel Securities forecasts $500 billion AI chip debt binge by 2028

Tech's most cash-rich companies are becoming the credit market's biggest borrowers. Citadel Securities forecasts more than $500 billion in new debt by 2028 to fund the chips powering AI data centers.

"This has the potential to become one of the largest new sectors in investment-grade credit," Jeff Eason, head investment-grade desk analyst at Citadel Securities, said. "The scale is unprecedented relative to today's market."

The $500 billion forecast would equal more than 5 percent of the Bloomberg US high-grade index by 2028, Eason said. He expects most issuance to be shorter-dated, around three to five years, to match the lifespan of the chips, with a portion sold as 144A private offerings. Global markets have already absorbed roughly $570 billion of AI-related debt, much of it from hyperscalers including Amazon.com Inc., Microsoft Corp. and Alphabet Inc.'s Google. That figure would be a fraction of the $250 billion-plus Eason estimates chip-makers will issue in 2028 alone.

The financing wave could reshape investment-grade credit portfolios, forcing investors to trim technology, media and telecommunications exposure to make room for chip debt. "The result is more than a funding story," Eason said. "It could fundamentally change the composition of the investment-grade market."

The debt machine behind the AI buildout

Citadel estimates AI capital expenditure will climb to roughly $600 billion in 2026, up from about $400 billion in 2025. Morgan Stanley projects the private credit market for AI data centers could reach around $800 billion by 2028, part of a $1.5 trillion financing gap against a $2.9 trillion global data center capex outlook.

The velocity of recent issuance is telling. Hyperscalers pushed out about $75 billion in bonds and loans during September and October 2025 alone, all directed at AI data center construction. US data center debt hit $25.4 billion in 2025, a 112 percent year-over-year increase. Chipmakers are joining in too — MediaTek recently approved a $5 billion financing package to boost production capacity for AI data center chips.

Why cash-rich giants are suddenly debt-hungry

The buildout required for competitive AI infrastructure is so capital-intensive that even companies generating tens of billions in annual free cash flow cannot self-fund their way through it. Leading AI labs are increasingly leaning on debt structures and backstops to access the investment-grade market. Anthropic clinched a roughly $35 billion financing package earlier this year to purchase Google's custom TPU chips, one of the largest private credit transactions in history. Broadcom backstopped payments on the largest senior portions of that debt, allowing Wall Street banks to trade portions of the tranche. OpenAI is similarly burning cash to expand and relying on debt support from large institutions.

The return-on-investment question looms over the leveraged bet. Tech companies are borrowing on the assumption that AI revenues will materialize fast enough to service the debt. If those returns disappoint, credit stress could ripple through markets. Citadel, which launched high-grade credit in early 2024 and traded roughly $500 billion notional last year, is positioning for the wave. Investors will have to decide whether the AI buildout's financing needs create opportunity or risk in the investment-grade market.

This article is for informational purposes only and does not constitute investment advice.