There has been no shortage of money for AI-related stories. But after a flood of issuance, investors are becoming more cautious about how much hyperscaler debt they want to hold.
Amazon’s latest $25bn bond sale was the clearest sign yet that the hyperscaler funding boom has entered a more sober phase.
The company still raised a huge amount of debt and drew $62bn of peak demand. But final orders fell to almost $41bn after pricing was tightened, leaving the deal just 1.6 times covered.
For most borrowers, a $41bn orderbook would be a triumph. For Amazon, in this market, it looked underwhelming. Its previous dollar deal in March raised a record $37bn and finished with about $119bn of demand, or 3.2 times oversubscribed.
This first wave of AI-related bond issuance was met with enthusiasm bordering on frenzy.
Orderbooks were vast, price tightening was aggressive and issuers could rely on the force of the theme as much as the strength of their credit.
Alphabet, Oracle, Meta, Nvidia and Amazon all showed this year how deep the market was for high grade technology debt.
But supply on this scale has changed the terms of demand. One investor described it as a simple calculation: the more bonds hyperscalers print, the less urgency investors have to chase each new deal.
Another said buyers were increasingly willing to wait for cheaper paper in future deals, knowing these companies are likely to return.
Hyperscalers are no longer occasional visitors to the bond market. They are becoming frequent borrowers.
Data centres, chips, energy supply and cloud infrastructure require enormous capital investment, and even companies with vast cashflows are turning to debt markets to help fund that expansion.
The AI boom is still reshaping corporate funding. And of course, the biggest technology companies will remain central borrowers in the bond market.
But Amazon’s cooler reception shows the mood has changed. The market will still fund the AI buildout, but no longer on hype alone.