Amazon went back to the debt window on Tuesday, and it brought a big order. The company launched an eight-tranche investment-grade bond offering seeking at least $25 billion, its second jumbo raise of 2026, as it races to fund an artificial intelligence buildout that will consume roughly $200 billion in capital spending this year, according to Reuters. The Amazon bond sale could grow if investor demand allows, Bloomberg reported, citing people familiar with the matter.
Maturities on the offering run from 3 to 40 years, with the longest tranche due in 2066 carrying initial price guidance of about 145 basis points over Treasuries, per TradingKey. Barclays, Goldman Sachs, J.P. Morgan, and Morgan Stanley are running the books. One detail stood out to credit desks: Amazon told its underwriters this will be its last debt issuance of 2026, CNBC reported, a signal that the company wants to draw a line under a year of heavy borrowing.
What Is Amazon Raising $25 Billion For?
On paper, the filing lists general corporate purposes, the standard catch-all that can cover debt repayment, acquisitions, and capital expenditures. Nobody on Wall Street is confused about where the money is headed. Amazon said in February it plans to spend approximately $200 billion this year on data centers, chips, and other AI-related equipment, a staggering jump from $131 billion in 2025.
That $69 billion year-over-year increase is larger than the entire annual capital budget of almost any company on earth. Amazon Web Services sits at the center of the plan. The cloud unit needs land, power, buildings, and silicon at a scale that has begun to outrun even Amazon’s formidable cash generation, which is precisely why the bond market keeps seeing the company’s name.
How the Amazon Bond Sale Is Structured
Eight tranches of floating-rate and fixed-rate notes cover the full curve, from 3-year paper for conservative buyers to a 40-year bond for pension funds and insurers hunting long duration. Pricing the 2066 maturity at roughly 145 basis points over Treasuries offers a meaningful pickup for a borrower of Amazon’s credit quality.
That structure echoes Amazon’s March deal, an 11-part offering that targeted $37 billion and was heavily oversubscribed. Demand for Big Tech paper has been deep all year. Investors treat hyperscaler bonds as a rare combination of scale, liquidity, and balance-sheet strength, and recent deals from the sector have drawn order books several times the offered size.
“It’s fertile ground right now in capital markets, and you’re also in the first half of the year,” said George Catrambone, head of fixed income, Americas, at asset manager DWS, in comments carried by Yahoo Finance, describing the window that issuers like Amazon are rushing to use.
The AI Arms Race Is Being Fought With Bonds
Amazon’s raise is one salvo in a financing war. Big Tech, including Amazon, Alphabet, Microsoft, and Meta, is expected to spend more than $700 billion on AI this year, according to Reuters. UBS thinks even that undershoots it, projecting aggregate hyperscaler capital spending could top $770 billion in 2026, roughly 23% above the bank’s earlier forecast.
Funding sources have shifted dramatically. Alphabet announced an upsized equity sale of about $85 billion last month after raising some $32 billion in the dollar and euro bond markets earlier this year. Meta sold $25 billion of investment-grade bonds this year, following a $30 billion deal in October that ranked as its largest ever. Oracle has issued $25 billion of its own. Morgan Stanley forecasts global AI-related bond issuance will approach $570 billion this year, a number that would have sounded absurd three years ago. International Daily Finance covered the early phase of this shift in Big Tech’s AI spending and investor payouts.
All that borrowing is rewriting the corporate bond market itself. The five major AI hyperscalers issued $121 billion in U.S. corporate bonds last year, against an average of $28 billion per year between 2020 and 2024. BofA Global Research just raised its forecast for hyperscaler debt issuance in 2026 to $175 billion from $140 billion.
“The increase in net supply is largely a non-financial story, and the biggest upside risk is AI hyperscaler capex, which could require more jumbo public deals than typical,” Barclays analysts wrote, adding that total U.S. investment-grade issuance could exceed $2 trillion in 2026, beyond even the post-COVID records of 2020.
Why Is Amazon Borrowing Instead of Using Cash?
For two decades, Silicon Valley’s giants funded themselves. Cash flowed in faster than engineers could spend it, and debt was something other industries needed. That unspoken contract with shareholders, fund growth from operations and return the rest, has been shattered by AI economics.
Building frontier AI infrastructure requires spending commitments measured in hundreds of billions of dollars over multiple years, with revenue that arrives later and less predictably. Even for Amazon, which generates enormous operating cash flow, matching that spending curve to internal cash generation alone would mean starving other priorities. Amazon’s LTL freight expansion is a reminder that the company still funds big ambitions well outside AI. Debt at investment-grade spreads is cheap relative to the opportunity, and CEO Andy Jassy has framed the stakes in unambiguous terms, calling AI a “once-in-a-lifetime opportunity” that requires big bets.
Markets have mostly rewarded the aggression. Amazon shares ticked up about 0.6% in premarket trading after the bond news broke, a muted but telling reaction. A $25 billion debt announcement that moves the stock higher means investors still believe the spending earns its keep.
What Skeptics Are Watching
Not everyone is comfortable. Some equity investors have begun asking whether hundreds of billions of dollars in AI investment will generate returns on the timeline the spending implies. The combined 2026 capital expenditures of Google, Microsoft, Amazon, and Meta are projected to reach roughly $725 billion, per TradingKey, and the market’s patience with long payback periods is wearing thin.
Bond desks have their own worries. A flood of hyperscaler supply pressures spreads across the investment-grade universe, and long-dated tech paper concentrates AI-outcome risk in portfolios that bought it for safety. If AI monetization disappoints in 2027 or 2028, today’s 40-year bonds will still be outstanding in 2066. Bond buyers are effectively underwriting a technology bet with a four-decade tail, and the long end of the Treasury market has already shown this year how unforgiving duration can be when sentiment turns.
How Amazon’s Debt Stacks Up Against Its Rivals
Amazon’s two 2026 deals, the $37 billion March target and this $25 billion-plus offering, put it at the front of the hyperscaler borrowing pack. Meta’s combined $55 billion across its last two deals runs close behind. Alphabet has leaned harder on equity, and Microsoft has so far borrowed least aggressively among the four.
Here is how the AI infrastructure spending is being financed across the group:
| Borrower | Recent debt raised | Detail |
|---|---|---|
| Amazon | $62 billion-plus across two 2026 deals | $37 billion March target plus at least $25 billion in July |
| Meta | $55 billion across its last two deals | $25 billion this year after a record $30 billion in October |
| Alphabet | About $32 billion in bonds this year | Paired with an upsized equity sale of roughly $85 billion |
| Oracle | $25 billion this year | Its own entry in the hyperscaler debt wave |
Behind the borrowing sits a simple competitive logic. Cloud market share is being reset by AI workloads, and capacity is the constraint. Every quarter a hyperscaler delays a data center is a quarter of AI revenue conceded to rivals, including upstarts: the market showed what it thinks of challengers when SpaceX passed Amazon in market capitalization within days of listing. Amazon is borrowing now, at spreads investors will happily fund, to make sure the capacity question is never the reason it loses a customer.
Should Bond Investors Buy Amazon’s New Debt?
For bond buyers, the Amazon bond sale offers scale and a rare final-supply pledge. Amazon saying it will not issue again in 2026 removes an overhang and could tighten secondary spreads on the new paper. For equity holders, the calculus is the one Jassy has argued all year: the cost of overbuilding AI capacity is measured in interest expense, while the cost of underbuilding is measured in lost markets.
One number will settle the question at pricing: the final spread on the 2066 tranche against its 145 basis point guidance. March’s $37 billion deal was heavily oversubscribed, and a repeat would tell credit desks that appetite for four-decade AI risk has not cooled, no matter how loudly the equity crowd frets.
