High yield: Navigating the AI issuance wave

Artificial intelligence (AI) continues to reshape the global economy, and its contribution to US growth is increasingly evident. In 2026, AI‑driven investment is expected to remain a central pillar of economic expansion, with important implications for credit markets and investors. 

While investment‑grade (IG) issuers are likely to finance most of the AI buildout, sub-IG markets will still play a role, contributing an estimated $200 billion in funding.[1] For high yield investors, this introduces a new wave of issuers, structures and pricing dynamics. We believe it is essential investors understand this fast-evolving dynamic to identify the potential opportunities and risks that AI may have on the high yield universe. 

The four key risks for high yield investors

• Sustainable demand

All investing involves uncertainty, but the challenge is particularly acute for high yield investors exposed to the AI ecosystem. AI models have advanced rapidly over the past two years, and visibility on their long‑term trajectory remains limited. This raises questions about whether today’s data‑centre infrastructure will remain essential as models become more efficient, and what sources of demand will support these assets when companies refinance.

Such uncertainty encourages investors to move away from traditional buy‑and‑hold strategies. As a result, issuers may need to offer more compelling pricing to attract and retain capital willing to underwrite AI‑related risk.

• Off‑balance‑sheet financing

Morgan Stanley estimates that technology firms could rely on up to $800 billion of off‑balance‑sheet financing – echoing patterns seen before the global financial crisis. [2] While this naturally raises concerns, we believe today’s environment differs in several important ways.

Alternative financing structures are generally disclosed, providing greater transparency. Additionally, much of the capital is directed towards meeting current demand for data centres, power capacity and connectivity—rather than building speculatively for the future potential demand as was the case during the Telecom/fibre build of early 2000s.

• Lack of multipurpose sites

Training facilities develop AI models, while inference facilities deploy them for real‑time use. Training sites require substantial upfront investment but have limited repurposing potential. Although alternative uses exist—high‑performance computing, scientific simulations or graphics rendering—these applications typically require less compute power, increasing the risk of future stranded assets.

• Construction risk and market volatility

Pricing construction risk remains a key challenge. Markets have reacted sharply to issuers experiencing delays, even when long‑term cash flows appear secure once facilities become operational. This dynamic underscores the importance of execution discipline in early‑stage AI infrastructure projects.

Issuer innovation to mitigate risks

These risks are amplified by the sector’s rapid technological cycle, heavy upfront capital requirements and uncertain long‑term cash‑flow visibility. Consequently, issuers operating in the AI ecosystem have adopted structures that diverge from the standard features of the high‑yield market, reflecting investors’ preference for shorter exposure, stronger credit and recovery protection.

• Shorter‑duration structures 

Typical high‑yield bonds carry maturities of around five to eight years. In contrast, most AI‑linked high yield issuance has maturities under five years, limiting long‑term exposure to demand uncertainty. 

• Amortising debt 

Mandatory annual paydowns of 5–7% after two years help reduce leverage and improve credit profiles over time. This structure is unusual in high yield, where bonds typically do not amortise and the debt is paid at the end of the bond’s life.

• Lease guarantees 

Some structures include guarantees from highly rated counterparties such as Google, reducing counterparty risk once projects are operational.

When considering how to position our high yield portfolios, we begin with assessing the backdrop for the companies. In this case, recognising among other things the weight of capital flows, the economic heft of the companies ultimately directing this build-out and the strategic national importance of the AI revolution. We then use our range approach to assess the upside and downside risks case by case. Position sizing is managed by combining the natural volatility of the AI theme along with the downside “drop risks” for each security. 

We see opportunity in the data centre subsector overall but will manage exposure carefully around the ever-changing broader market narrative.

Outlook

While AI dominates market narratives, technology currently represents just 7% of the high yield index.[3] Even if this share doubles to 14% over the next five years, the market will remain broadly diversified, reducing the impact it has on overall index performance.

It is also worth highlighting a distinction between public high yield markets and leveraged loans/direct lending. Leveraged loans have more than double the technology sector exposure compared to public high yield, underscoring a notable difference for high yield investors—especially as AI-related issuance accelerates in private credit and loan markets.

AI’s influence, however, extends beyond the technology sector. Potential job displacement shifts in income distribution, and changes in consumer behaviour could affect the remaining 93% of the index, with implications for corporate revenues and cash flows. At the same time, we believe AI offers meaningful opportunities: enhanced productivity, lower operating costs and new business models that could strengthen credit fundamentals across multiple industries. 

When considering how to position our high yield portfolios, we begin with assessing the backdrop for the companies. In this case, recognising among other things the weight of capital flows, the economic heft of the companies ultimately directing this build-out and the strategic national importance of the AI revolution. We see opportunity in the data centre subsector overall but will manage exposure carefully around the ever-changing broader market narrative.

 
[1] JP Morgan, “AI Capex – Financing The Investment Cycles”, 10 November 2025.
[2] Bloomberg, “Meta, AI Starting Trend for Billions in Off-Balance Sheet Debt”, 31 October 2025. 
[3] JP Morgan, “AI Capex – Financing The Investment Cycles”, 10 November 2025.