Michael Elsworth, Manager, Fixed Income, Lonsec
Artificial intelligence is emerging as an increasingly important structural theme for global fixed income markets. While the investment case for AI has initially been dominated by equity-market valuations, the next phase of the AI investment cycle is increasingly being transmitted through credit markets.
The five major US hyperscalers — Microsoft, Amazon, Alphabet, Meta and Oracle — are undertaking an unprecedented expansion in data centres, computing infrastructure, networking and electricity capacity. J.P. Morgan estimates that capital expenditure by these five companies will reach US$697 billion in 2026, approximately US$173 billion higher than its estimate at the beginning of the year.
The financing implications are substantial. The BIS reports that hyperscaler gross corporate bond issuance exceeded US$100 billion in 2025, with most issuance at maturities beyond five years. By July 2026, Reuters reported that the five major hyperscalers had already issued approximately US$194 billion of bonds, with Goldman Sachs forecasting full-year issuance of around US$250 billion for 2026 and US$400 billion in 2027.
At the same time, AI infrastructure is increasingly being financed through private credit, joint ventures and special-purpose vehicles, obligations that can be economically similar to corporate debt but remain largely outside hyperscaler balance sheets.
For global fixed income investors, this creates important implications:
- AI is becoming a fixed income story.
- A gradual change in the composition and risk characteristics of broad benchmarks such as the Bloomberg Global Aggregate Bond Index.
- The rise of off-balance-sheet AI financing.
1. AI is becoming a fixed income story
The defining feature of the current AI cycle is its extraordinary capital intensity. Unlike earlier software-led technology cycles, generative AI requires substantial physical infrastructure. Data centres need advanced processors, networking equipment, cooling systems, land and buildings, while their electricity requirements are creating additional investment in generation and transmission infrastructure.
J.P. Morgan estimates that the five largest US hyperscalers will spend approximately US$697 billion on capital expenditure in 2026. This represents a significant acceleration from the approximately US$524 billion implied for 2025. The BIS estimates that the five major hyperscalers will invest more than US$1 trillion during 2025 and 2026, with capital expenditure rising substantially relative to their existing asset bases. Historically, hyperscaler operating cash flows meant that debt was often used primarily for balance-sheet management rather than to fund the majority of investment. The financing equation is now changing. As AI-related capital expenditure rises, an increasing proportion is being funded through debt, leases, project-level financing and private credit. AI is therefore transitioning from an equity-market investment theme into a major source of fixed income supply.
2. Implications for the Bloomberg Global Aggregate Bond Index
The most visible effect is the increase in corporate bond issuance – hyperscaler debt issuance is changing the investment-grade bond market. AI-related borrowing is creating a substantial new pool of relatively high-quality corporate duration.
The Bloomberg Global Aggregate Bond Index is the most used benchmark by global fixed income managers and provides an important framework for understanding the long-term implications. The index is a broad measure of global investment-grade debt covering Treasury, government-related, corporate and securitised fixed-rate bonds across developed and emerging markets.
The Index contains over US$75 trillion of notional debt and over 30,000 securities. The Index added approximately US$7.4 trillion of debt during 2025, highlighting the enormous underlying growth of the global investment-grade bond market. Against this scale, even US$250–400 billion of annual hyperscaler issuance remains relatively modest.
However, the issue is not simply the absolute size of AI debt. If AI-related corporate borrowing continues to grow faster than the broader corporate bond market, the Index’s corporate component could gradually become more exposed to technology and AI infrastructure.
AI infrastructure typically has a long economic life, with data centres able to operate for decades, while the associated electricity and network infrastructure may have even longer useful lives. This naturally encourages long-term financing. The BIS notes that most hyperscaler issuance in 2025 was for maturities exceeding five years. This means fixed income markets are not simply absorbing more credit, they are absorbing more duration. This is particularly important for passive or index investors who do not independently decide that billions of long-term additional hyperscaler debt represents good value.
3. The rise of off-balance-sheet AI financing
Another significant development for credit investors is occurring outside the conventional corporate bond market. The BIS identifies an increasing use of special-purpose vehicles and joint ventures to finance AI infrastructure. Under these arrangements, a data centre may be owned by a dedicated vehicle funded by private equity and private credit. The hyperscaler may then enter into a long-term lease, capacity agreement or purchase commitment. Economically, the hyperscaler has effectively committed to the infrastructure without necessarily issuing the debt directly on its balance sheet.
A recent example is the reported US$35 billion private-credit transaction supporting Anthropic’s AI infrastructure, involving an SPV and debt backed by chip lease payments. The transaction demonstrates how AI financing is beginning to move beyond conventional corporate bonds into asset-backed and structured-credit markets.
What does this mean?
The AI revolution is beginning to reshape global fixed income markets.
The extraordinary capital requirements associated with data centres, chips, electricity and networking infrastructure are generating substantial new bond supply while simultaneously accelerating the growth of private credit and structured financing. Hyperscaler bond issuance has already moved sharply higher, while off-balance-sheet financing is creating additional exposures that may not be immediately apparent from traditional balance-sheet analysis.
For the Bloomberg Global Aggregate Index, the effect is likely to be evolutionary rather than transformational in the near term. However, if AI-related debt continues to grow faster than the broader investment-grade market, the Index could gradually acquire greater exposure to technology, data centres, utilities and other AI infrastructure. Its corporate sector could become more technology-intensive, more duration-sensitive and more interconnected with private-credit and infrastructure markets.
For fixed income investors, the AI story is increasingly a story about duration, credit supply, leverage and benchmark evolution — and these effects are likely to persist well beyond the current investment cycle.
Important Information: This article has been produced by Lonsec Research Pty Ltd ABN 11 151 658 561, AFSL No. 421445 (Lonsec). Generation Development Group Limited ABN 90 087 334 370 is the parent company of Lonsec Research.
Warning: The content presented in this article is for informational purposes only and is not a prospectus, product disclosure statement or any other form of disclosure document within the meaning of the Corporations Act 2001. Any express or implied rating or advice presented in this article is limited to “general advice” (as defined in the Corporations Act 2001 (Cth)), without taking into account the investors’ objectives, financial situation and needs. Prior to acting upon any information presented in this article, advisers must do their own investigation and analysis of the appropriateness of the information for their investor, having regard to their objectives, financial situation and needs.
Disclaimer: Lonsec Research have provided this article for exclusive use of its clients. It is not intended for use by a retail client or member of the public and should not be used or relied upon by any other person. No representation, warranty or undertaking is given or made in relation to the accuracy or completeness of the information presented in this article. Lonsec Research assumes no obligation to update this article following publication. Except for any liability which cannot be excluded, Lonsec Research, their directors, officers, employees and agents disclaim all liability for any error or inaccuracy in, misstatement or omission from, this article or any loss or damage suffered by the viewer or any other person as a consequence of relying upon it.
