AI Boom Crowds Bond Market

The insatiable appetite for compute in artificial intelligence necessitates substantial investment in data centers, chips, and other expensive infrastructure, according to Westin.

This demand is reflected in the market, where six major hyperscalers have seen their share prices rise by an average of 15% year to date, with debt investors now playing a central role in financing this expansion.

The tech sector's share of investment-grade corporate bonds has dramatically increased from 5% in 2024 to nearly 20%, driven by hyperscaler borrowing.

Loren Moran projected that hyperscaler borrowing in the investment-grade market could reach approximately $300 billion through 2026, contributing to an overall 35% year-over-year increase in U.S. investment-grade issuance, expected to hit $2.25 trillion for the year.

While initial demand for hyperscaler debt was robust due to strong double-A credit ratings, market sentiment is now shifting as the growing volume prompts buyers to demand greater compensation to lend to these issuers, as noted by Loren Moran.

This shift is exacerbated by global competition for capital from sovereign balance sheets and increased M&A activity, creating market indigestion and raising concerns about the sustainability of current borrowing capacity into the next year.

Hyperscalers now constitute 5% of the debt index, a significant rise from 1-2% just a few years ago, according to Scott Schulte, indicating a shift in market concentration.

Unlike banks that issue shorter-term debt, hyperscalers commonly prefer long maturities of 20, 30, 40, or even 100 years, which is significantly extending the overall duration of the debt index and is projected to make the AI trade overtake U.S. money center banks in concentration risk within 18 to 24 months.

Data center financing introduces distinct risk profiles compared to typical investment-grade markets, a nuance highlighted by Loren Moran, as companies increasingly utilize special purpose vehicles to keep this debt off their main balance sheets.

These new financing structures introduce risks such as construction delays, permitting challenges, and electrification issues, with recent implied force majeure instances signaling vulnerabilities within the AI ecosystem that require investors to consider holistic factors like power availability and transmission.

Fixed income markets do not treat all hyperscalers equally, despite many issuing significant debt, showing a fair amount of dispersion within the sector regarding both absolute levels and how the market perceives their credit curves, Westin observed.

Credit investors prioritize companies with clear financial policies and predictable cash flow certainty, leading to differentiation between highly leveraged balance sheets, such as Oracle, and less leveraged ones, like Microsoft.

Investors are becoming increasingly sensitive to risk, particularly in high-yield names, with recent weeks showing a significant widening of spreads in certain sectors, noted Subadra Rajappa.

Despite these higher borrowing costs, the attractiveness of overall yields continues to underpin demand, although investors are becoming more discerning in their choices.

AI infrastructure borrowing, due to its substantial volume, may be crowding out other participants in the market, although the specific impact depends on the type of investor, according to Subadra Rajappa.

Unlike other corporate borrowers who might delay issuance to assess yield levels, hyperscalers operate with a 'time agnostic' mandate, committed to significant issuance this year and next, with the market currently absorbing this supply effectively.

A broader dynamic at play is the escalating magnitude of debt and deficits in developed markets, extending beyond the U.S. to countries like Japan, the U.K., France, and Germany, where increased sovereign issuance is driven by defense spending and infrastructure needs, Subadra Rajappa explained.

This rise in government bond debt, traditionally a source of investor demand, is now creating significant competition for capital and effectively crowding out the broader corporate bond market as investors weigh higher yields.

Subadra Rajappa expressed concern about how large companies will recoup the massive capital expenditures being invested in AI, not just this year but over the next several years.

Additionally, political headwinds regarding data center development could emerge as midterm elections approach, potentially leading to community and power-related resistance that could slow AI deployment and progress.