The AI Infrastructure Boom: A Double-Edged Sword?
The AI revolution is in full swing, and with it comes a surge in investment and spending on infrastructure. The tech giants and hyperscalers are leading the charge, partnering with Wall Street firms to mobilize an astonishing $500 billion in third-party capital. This unprecedented buildout is fueled by a combination of traditional financing methods and innovative structures, such as bond markets, joint ventures, and leases.
But this boom is not without its risks. The recent collapse of the Situational Awareness hedge fund highlights the dangers of leveraged AI trades. With a portfolio heavily concentrated in AI-related stocks, the fund was unable to meet margin calls during a tech sell-off, leading to a dramatic loss of assets. This event raises questions about the visibility and potential unwind of leverage in the AI sector.
The scale of borrowing by hyperscalers is staggering. Goldman Sachs analysts estimate a combined lease commitment of $1.5 trillion for data centers, R&D facilities, offices, and equipment, a fivefold increase from just five years ago. This includes $1 trillion in 'uncommenced' lease commitments, which are not yet reflected in financial statements but will result in future payments. As these obligations are recognized and payments come due, the true leverage and liquidity needs may become more apparent, potentially underestimating the scale of the buildout.
The AI capex cycle, adjusted for inflation, is on track to be the largest investment cycle since the 19th-century railway construction, according to Lotfi Karoui, a multi-asset credit strategist at PIMCO. However, the ultimate scale of the buildout remains deeply uncertain. Consensus forecasts predict hyperscaler capital spending to surpass $1 trillion per year from 2027 onward, with no clear signs of moderation. This scale of borrowing is prompting hyperscalers to diversify their debt issuance beyond dollar-denominated paper, tapping into various markets.
The relative outperformance of euro-denominated bond spreads issued by Amazon and Alphabet compared to their U.S. counterparts hints at potential 'demand fatigue' in the dollar market. This could lead to higher spreads in the U.S. due to the underperformance of larger AI-exposed issuers. The question arises: Is AI leverage becoming a bigger market risk?
While the Situational Awareness collapse demonstrated the vulnerability of crowded, leveraged AI trades, it is essential to consider the broader context. Sahil Mahtani, director of the investment institute at Ninety One, argues that elevated earnings expectations are a more immediate concern than leverage. He believes that the main risk the AI trade poses to markets is the expectation of high and rising earnings in the years ahead, which is more of an expectations problem than a leverage problem.
Equity concentration in tech-heavy markets, especially the U.S., is historically high, and this concentration can act like leverage by amplifying market moves when heavily weighted stocks fall. However, Mahtani also notes that the impact of the Situational Awareness debacle has been largely contained, as it coincided with the unwind of leveraged ETF structures in East Asia, which were primarily launched in the first half of this year.
In conclusion, the AI infrastructure boom is a double-edged sword. While it drives innovation and investment, it also brings risks, particularly in the realm of leverage. As the sector continues to evolve, it is crucial to carefully manage these risks and ensure that the eventual returns from AI infrastructure can justify the vast sums being spent. The future of AI depends on striking the right balance between innovation and stability.
