AI-Linked Debt on Track to Nearly Double to $570 Billion in 2026, Morgan Stanley Says

AI-Linked Debt on Track to Nearly Double to $570 Billion in 2026, Morgan Stanley Says

The artificial intelligence boom has spent three years being told as a stock story. In 2026 it is becoming a bond story, and the numbers are large enough to reshape the investment-grade credit market that sits quietly inside millions of American retirement accounts.

Global debt issuance tied to AI is on track to more than double to nearly $570 billion this year, according to a Morgan Stanley forecast. By May 31, AI-related issuers had already sold close to $236 billion of debt worldwide, roughly four times the pace over the same stretch of 2025. If the trajectory holds, AI bonds will become the single largest investment-grade sector by issuance, edging out the banks and utilities that have long anchored the market.

From Cash Flows to Capital Markets

For most of the buildout, hyperscalers paid for their data centers the old-fashioned way: out of operating cash flow. That era is ending. The scale of planned spending has simply outrun what even the most profitable companies on earth can self-fund.

Microsoft is guiding to roughly $190 billion in capital expenditure. Google has signaled $175 billion to $185 billion. Amazon is pouring more than $20 billion into custom silicon alone. Anthropic has signed leases for more than a gigawatt of U.S. data center capacity, and the OpenAI-anchored Stargate joint venture is targeting an eventual 10 gigawatts. Collectively, hyperscaler capex is expected to clear $600 billion in 2026.

Cash flow cannot stretch that far, so the money is increasingly coming from somewhere else. Since late 2024, the five largest hyperscalers have tapped capital markets for more than $137 billion, a historic surge in tech-sector debt issuance. Morgan Stanley expects $250 billion to $300 billion of fresh issuance in 2026 from hyperscalers and their joint ventures alone. Add chipmakers, data-center project finance, and structured vehicles, and the bank's full-year figure approaches $570 billion.

"For bond investors, the AI buildout is no longer just a stock story," Morgan Stanley analysts wrote. "It is becoming the largest single position in the investment-grade credit market." By October 2025, total AI-tied debt had already ballooned to roughly $1.2 trillion.

The Deal That Defines the Era

No single transaction captures the new financial architecture better than the roughly $36 billion debt package Apollo Global Management and Blackstone have been shopping to fund Anthropic's compute.

The structure is intricate. A special-purpose vehicle borrows the money, uses it to buy Google's custom tensor processing units, and then leases that hardware to Anthropic for data centers in New York, Texas, Louisiana, and Indiana. The debt is sliced into roughly $6 billion of A1 notes, $25 billion of A2 notes, and $4.5 billion of B notes. Broadcom, which helps Google manufacture the chips, is providing a residual-value backstop on about $31 billion of the senior debt: if Anthropic stops paying and the used chips fetch too little on resale, Broadcom absorbs the shortfall.

The arrangement would rank among the largest private-credit deals ever and the biggest chip-financing transaction on record. Crucially, the leasing structure keeps the debt off Anthropic's own balance sheet entirely. It is a template the industry is racing to copy, and one that makes the true scale of AI leverage harder to see.

The Circle Tightens

That opacity is where the worry lives. Analysts increasingly describe the AI ecosystem as a web of circular financing, in which chipmakers, cloud providers, model labs, and their lenders all hold claims on one another's growth.

"The entire system is leveraged on itself," one credit strategist warned, describing capital that keeps recycling "within a closed circle" that "breeds tight interdependence and inflates headline growth." Nvidia invests in customers who buy Nvidia chips; hyperscalers backstop labs that rent hyperscaler capacity; private-credit firms fund the chips that secure their own loans. Round-tripping like this can flatter revenue on the way up and amplify losses on the way down.

The unresolved question underneath all of it is return on investment. Morgan Stanley and others now frame 2026 as the year the math gets tested. The most-cited data point is uncomfortable: Microsoft's AI services are running at roughly $37 billion in annual recurring revenue, against something closer to $97 billion in AI-related spending over the past four quarters. Most analysts still believe the demand is real and the revenue will catch up. But for the first time, much of the gap is being financed by bondholders rather than shareholders, and bondholders are paid back in fixed dollars regardless of whether the bet pays off.

"Investors are going to demand proof that these outlays accelerate revenue or margins," one analyst noted, summarizing the shift from faith to scrutiny.

What to Watch

The next two quarters will tell investors a great deal. Watch whether the Apollo-Blackstone Anthropic deal prices cleanly and how tight the spreads come; a smooth syndication would signal that institutional appetite for AI credit remains deep. Watch ratings agencies, which have so far kept hyperscalers comfortably investment-grade but are scrutinizing leverage and the off-balance-sheet leasing structures now proliferating. Watch free cash flow, which several of the biggest spenders may push negative this year for the first time.

And watch the ROI line. As long as AI revenue keeps climbing fast enough to make the spending look forward-looking rather than reckless, $570 billion in new debt is a sign of confidence. If that revenue stalls while the interest payments do not, the same number becomes the measure of how much is at stake.