Amazon makes market history with largest corporate bond sale ever
Amazon set a new record for the largest corporate bond sale in capital markets history, raising US$53.8bn-equivalent as it gathers the money it needs to acquire a significant stake in OpenAI – while also embarking on the biggest corporate capital expenditure programme ever seen.
The deal, spanning 19 tranches and two currencies, took place over two days, with the US$37bn US dollar leg pricing on Tuesday and the €14.5bn euro component a day later, together beating the previous US$49bn record for the largest corporate bond set by Verizon Communications that has stood since 2013.
It was all the more spectacular given the market backdrop, with the war in the Middle East entirely closing the primary US bond market the day before the first tranche launched as oil prices surged by 30%. Even so, the deal attracted US$158bn-equivalent of demand across both legs, in a clear sign of investor appetite for hyperscaler exposure.
“Notwithstanding just how volatile markets are right now, for the right companies – the truly great companies and credit stories – there is still a tremendous amount of capital available,” said John Servidea, global co-head of investment-grade debt capital markets at JP Morgan.
Amazon is one of the key players in the AI investment boom. It plans to spend US$200bn this year building data centres, developing its own chips and training its Nova model – more than any of its hyperscaler peers – in what will be the largest corporate investment programme ever seen.
At the same time, it will also invest as much as US$50bn in OpenAI, which is both a rival and a major customer for its Amazon Web Services business. The huge spending programme is set to strain the Seattle-based company’s finances, pushing its free cashflow into negative territory.
That will make it reliant on debt, in a reversal of its recent strategy of paying down the money it borrowed during the 2010s, when it spent aggressively to cement its market dominance. The latest deal takes borrowings to US$122bn, up from US$58bn at the end of October.
Unusual sequencing
JP Morgan, one of three leads on Amazon’s most recent bond deal in November, which raised a relatively modest US$15bn, was in the driving seat this time. It was the only global coordinator on both legs alongside no fewer than 26 other banks coming in as joint bookrunners and co-managers across both pieces.
Like hyperscaler rival Alphabet, which raised US$31.5bn-equivalent in the US dollar, sterling and Swiss franc markets last month, Amazon also opted to tap more than one currency, supplementing the US dollar offering with euros and making its debut in the currency.
But unlike the Alphabet deal, which launched in sterling and Swiss francs first before adding the US dollar leg once the US woke up, Amazon took the opposite approach. Both legs were announced together around 7:30am New York time on Tuesday, too late to price the euro part of the deal that day.
According to another banker on the deal, the less-than-ideal sequencing – which effectively added an extra day to the execution – was dictated by the volatile market situation and the view that launching the US dollar leg first would give leads a better opportunity to set the tone. It also followed a stabilisation in markets overnight in Asia and Europe.
Servidea said the prospect of beating Verizon's record was not a motivating factor – and nor was size for the sake of it.
Given the scale of hyperscaler public market issuance this year, which he said could be as high as US$400bn, issuers are conscious that they need to be thoughtful about how they come to market.
“They all have very large, publicly stated capital needs – and they know that efficiency matters when you're talking about numbers that large,” he said. “It’s never just about size; it’s about how do I calibrate the balance of raising large amounts of capital over time in the most efficient way.”
Lucrative fees
Indeed, banks are emerging as major beneficiaries of the rush to finance the enormous sums being spent on the AI race. The four major publicly listed hyperscalers – Alphabet, Amazon, Meta Platforms and Microsoft – plan to spend US$630bn on capital expenditure this year, 50% more than last year.
Debt markets are helping finance those ambitions. Last month, Alphabet raised US$31.5bn-equivalent in a bond sale across three currencies while Oracle raised US$25bn in US dollars. Such deals can be highly lucrative: on the US dollar leg of the latest Amazon transaction, banks were paid US$84.7m in underwriting fees and commissions. While the euro leg fees had not yet been published at the time of going to press, they are expected to be north of €50m.
“When you think back to some of the estimates for hyperscaler issuance for this year, one could argue that we’ve undershot year to date,” said Servidea. “But my view is that, after this deal and those from Oracle and Alphabet, we’re now back on track with what the market was expecting.”
At the same time, privately held rivals such as OpenAI, Anthropic and xAI plan to spend tens of billions of dollars this year training and developing their own models. The three companies have raised US$160bn from equity investors this year – and are moving ahead with IPOs.
Some of those equity deals are also triggering a cascade of financing in public markets – as evidenced by the Amazon bond sale – as companies raise the money to invest. SoftBank Group is in talks with banks to borrow as much as US$40bn in the syndicated loan market to finance its equity investment in OpenAI.
“What a time to be alive,” said one San Francisco-based banker. “Just a few years ago none of these companies needed to raise any money because they were making such huge profits, but now their capital needs have shot through the roof. They almost can’t raise funds quickly enough.”
For details of the US dollar and euro tranches of the deal, see separate stories.
