Blog Post
$1.5 trillion in AI debt — and that’s just the start
Whatever the broader economic impact of AI turns out to be, its impact on credit market issuance is already profound. The infrastructure buildout is happening now, at a scale that has few precedents in history, and it needs to be funded today.
Morgan Stanley pegs the total cost of the buildout at roughly $3 trillion between 2025 and 2028, with about half requiring external financing. To frame the external financing figure alone: $1.5 trillion is roughly the size of the entire U.S. high-yield market, or the entire U.S. leveraged loan market. And the further out into the future we look, the higher the numbers: McKinsey runs the total capex number to $7 trillion by 2030.
The investment-grade market is absorbing most of it, for now. Hyperscalers like Amazon, Meta and Oracle, long debt-averse, now dominate high-grade credit issuance. AI-related IG supply could reach $300 billion in 2026 alone, according to the Dallas Fed, and we would bet that estimate will end up being too conservative.
The non-investment-grade market is increasingly in the mix, though. Octus estimates that high-yield and unrated AI infrastructure issuers have raised more than $107 billion in committed and funded debt as of early May, split between neoclouds ($68.7 billion) and AI-native data center owners ($38.7 billion).
This may prove to be an opening act. On the basis of publicly disclosed capacity targets, the seven largest non-IG issuers may need to raise as much as $435 billion over the next four years, pushing the total to well over half a trillion.
The sector is generating novel structures, a new class of borrowers, and spread levels wide enough to suggest the market is still working out how to price the risk. Where things ultimately settle is an open question, but one thing is clear: there will be a lot more debt to go around.
This publication has been prepared by Octus Intelligence, Inc. or one of its affiliates (collectively, "Octus") and is being provided to the recipient in connection with a subscription to one or more Octus products. Recipient’s use of the Octus platform is subject to Octus Terms of Use or the user agreement pursuant to which the recipient has access to the platform (the “Applicable Terms”). The recipient of this publication may not redistribute or republish any portion of the information contained herein other than with Octus express written consent or in accordance with the Applicable Terms. The information in this publication is for general informational purposes only and should not be construed as legal, investment, accounting or other professional advice on any subject matter or as a substitute for such advice. The recipient of this publication must comply with all applicable laws, including laws regarding the purchase and sale of securities. Octus obtains information from a wide variety of sources, which it believes to be reliable, but Octus does not make any representation, warranty, or certification as to the materiality or public availability of the information in this publication or that such information is accurate, complete, comprehensive or fit for a particular purpose. Recipients must make their own decisions about investment strategies or securities mentioned in this publication. Octus and its officers, directors, partners and employees expressly disclaim all liability relating to or arising from actions taken or not taken based on any or all of the information contained in this publication. © 2026 Octus. All rights reserved. Octus(TM) and the Octus logo are trademarks of Octus Intelligence, Inc.