Big Tech is borrowing hundreds of billions for AI, and bond investors are getting cautious

The companies that used to fund everything out of their couch cushions are now taking out loans. Hyperscalers, data center operators and chipmakers are borrowing hundreds of billions of dollars to pay for the AI buildout.

AI-related debt issuance for 2026 is estimated at between $450 billion and $570 billion. That is a striking number for an industry long famous for sitting on enormous cash piles.

The borrowers are familiar names: Amazon, Microsoft, Alphabet, Meta and Oracle, plus chipmakers like Nvidia and Broadcom. The shopping list includes data centers, servers, GPUs and the power capacity needed to keep all of it running.

The numbers behind the borrowing binge

Goldman Sachs projects that hyperscaler capital expenditures will land between $600 billion and $820 billion in 2026, with spending potentially topping $1 trillion in 2027.

The bond issuance is already piling up. Goldman Sachs has noted that gross corporate-bond issuance by hyperscalers in the early months of 2026 has already surpassed the total for all of 2025.

AI-related debt issuance in 2025 reached $100 billion to $121 billion. That figure was more than three times the average of prior years, and 2026 estimates sit several multiples above it.

Amazon completed a $54 billion bond sale in March 2026 as part of its push to fund AI infrastructure.

Leverage ratios for leading hyperscalers have reportedly risen from roughly 0.9x to 1.8x in just six months.

Aggregate lease commitments for hyperscalers now stand at around $1.5 trillion, up from $200 billion five years ago. Approximately $1 trillion of that total sits in leases that have not yet commenced. Those obligations are not reflected directly on balance sheets, which means headline debt figures may understate how much these companies have actually signed up to pay.

From cash rich to credit hungry

Oracle and certain chip manufacturers are experiencing credit pressure as their debt levels grow. The borrowing is not confined to the cloud giants, either: chipmakers and related entities are adding their own financing demand to the pile.

What this means for investors and the AI trade

The bond market is sending early signals of fatigue. Spreads on AI-linked bonds have widened, meaning investors are demanding more compensation to hold this debt.

Cover ratios on recent deals, which measure how many orders come in relative to the bonds on offer, have been declining.

With roughly $1 trillion in leases not yet commenced, the true level of indebtedness may be harder to measure than standard leverage ratios suggest.

For investors, a few indicators are worth tracking. Watch cover ratios on upcoming hyperscaler bond deals, the direction of spreads on AI-linked debt, and whether capex guidance keeps climbing toward the upper end of Goldman’s range.