Amazon's $25B AI-infrastructure bond sale

Amazon got its money. The story is what it cost to get it. When Amazon sold its biggest-ever bond earlier in 2026, it was inundated with orders on the back of AI-boom optimism. This time the reception was measurably cooler: peak demand for the eight-part offering reached roughly $62 billion, about half the orders the March deal drew, and the book ultimately pared back to around $41 billion ^[2]. That works out to roughly 1.6 times the deal size, well short of the roughly 4-times oversubscription that has been average for US high-grade sales in 2026 ^[1].

The gap matters because bond demand is a real-time referendum on how much AI debt investors are willing to keep buying. Amazon remains a top-rated, cash-generative borrower, so this is not a credit scare. It is an appetite problem. With this sale, Amazon's total bond issuance has climbed to more than $100 billion over the past year ^[2], and even a balance sheet that strong is starting to test how much paper the market will absorb without demanding a richer price.

The most telling signal was not in the new deal at all - it was in the secondary market. To make room for Amazon's new notes, investors dumped existing tech and hyperscaler bonds, pushing Amazon's outstanding paper about 7 to 10 basis points wider and SpaceX bonds about 9 to 13 basis points wider ^[3]. When buyers have to sell what they hold to fund what is coming, it is a sign that portfolios are already saturated with this kind of risk.

John Lloyd of Janus Henderson framed the rotation as housekeeping rather than panic - investors sell one position to free capital for another, the way someone sells one house before buying the next ^[3]. But the same firm is underweight tech and hyperscalers precisely because of the relentless supply. That combination - orderly today, but structurally cautious - is the quiet warning underneath an otherwise successful sale.

Amazon's raise is one brick in a wall that has been going up all year. Its 2026 capital-expenditure guidance sits near $200 billion, up sharply from about $131 billion in 2025 ^[1], and combined AI spending across Alphabet, Microsoft, Meta and Amazon is expected to top $700 billion this year ^[1]. Cash flow alone no longer covers that ambition, so the biggest names have turned to the bond market at a pace without precedent: the five largest hyperscalers issued about $121 billion of US corporate bonds in 2025, versus an average near $28 billion a year over 2020 to 2024 ^[4].

And the forecasts keep climbing. BofA Global Research raised its 2026 hyperscaler new-debt estimate to $175 billion from $140 billion after Amazon's deal ^[4], while global AI-linked debt issuance has already reached roughly $335 billion in 2026, more than double 2025 levels ^[2]. The wave has grown exotic as well as large - Alphabet sold the first 100-year tech bond in decades, and Oracle stacked a $30 billion February deal on top of an $18 billion one from the prior autumn ^[6]. Each new jumbo deal makes the next one harder to place.

The bears are not arguing that Amazon will miss a coupon. They are questioning the shape of the financing. RBC BlueBay's Andrzej Skiba has warned that spreads will stay under pressure unless hyperscalers show more large-scale equity raising to fund capex rather than leaning almost entirely on debt ^[3]. The structure of this very deal sharpens that concern: the longest tranche is a note due in 2066 priced around 1.45 percentage points over Treasuries ^[5], ultra-long debt funding hardware that turns over far faster than 40 years. A separate skeptical read has cast the broader boom as AI increasingly financed by borrowing rather than earnings ^[6].

Retail investors picked up the same thread from a different angle. Community reaction skewed cynical and consumer-facing, with the loudest voices treating the raise as a cost that will eventually land on customers rather than a signal of financial strength, while a more analytical minority defended debt as a rational, tax-efficient way to match long-dated infrastructure to long-dated funding. That split - Wall Street asking whether the funding mix is sustainable and Main Street asking who ultimately pays - is the tension worth tracking as the next round of hyperscaler deals arrives.