AI's Impact on Fixed Income

The growth of artificial intelligence has had significant impacts across the economy and investment markets. While investors often hear about its effects on equity markets—where a relatively small number of AI-related companies are driving returns and index concentration—many overlook the impact AI is having on fixed income markets.

AI is highly capital intensive, with substantial upfront costs related to data centers, semiconductors, energy, and other infrastructure. Estimates for newly developed data centers are approximately $35 billion per GWh (gigawatt hour), and roughly 200 GWh of new capacity is planned across approximately 3,300 data centers that are either under construction or have been announced.¹ As firms undertake these multiyear buildouts, many are turning to fixed income markets to fund these projects. Issuance by hyperscalers (Google, Meta, Microsoft, Amazon, and Oracle) has been significant over the past year, totaling roughly $120 billion, and we expect annual supply to increase to more than $200 billion in 2026. While at first glance this additional issuance will increase the technology sector’s footprint in fixed income indices, it also carries broader implications for fixed income investors.

First, the rise in AI-related debt issuance is increasing issuers’ interest burdens, placing pressure on free cash flow and pushing leverage higher. In some cases, this has already resulted in spread widening, or cheaper valuations, as markets incorporate the pro forma impact on fundamentals. For example, Oracle has experienced spread widening tied to increased leverage expectations, as well as higher lease commitments and concerns over customer concentration. Separately, hyperscaler issuance has been skewed toward longer maturities—often 10 years or longer—affecting relative value across individual issuers’ credit curves.

Technology is not the only sector affected by the capital demands of AI investment. Within corporate credit, utilities and power generators have also seen increased issuance, as data centers require substantial and reliable energy. Estimates suggest that approximately 30% of the total cost of a new data center is related to energy infrastructure.¹ Beyond corporates, we expect growing volumes of securitized and infrastructure debt tied to AI development. These financings are often structured around specific projects, such as Meta’s Louisiana data center. Importantly, not all companies are funding these ventures in the same way. In Meta’s case, much of the financing is expected to be off balance sheet, with Blue Owl issuing private debt to support its commitment to the project. These structural nuances are particularly important for investors evaluating the growing volume of AI-related issuance.

AI-related debt is likely to continue growing as a share of fixed income indices due to the reinvestment cycle—the useful life of chips is typically four to six years—and continued investment to improve models (scaling laws). While many of these investments have the potential to drive future revenue growth, a loss of confidence among credit investors could materially impact future development.

In our view, relative value should be a key consideration when evaluating these opportunities, particularly because data-center-specific issuance carries idiosyncratic risks related to both the evolution of AI and the economic significance of individual projects. Given current valuations and the expected increase in AI-related debt issuance in the years ahead, we believe more attractive opportunities to add value exist elsewhere in fixed income.