as Nvidia It has partnered with Wall Street firms to mobilize more than $500 billion in third-party capital for AI infrastructure, but the increasingly complex financing that underpins the boom is coming under increased scrutiny.
Hyperscalers and their financial backers are turning to bond markets, joint ventures, leases, and other structures to finance unprecedented infrastructure developments. At the same time, leverage is increasing investors’ AI exposure as hedge funds and other investors use prime broker borrowing and derivatives to magnify the returns on their boom bets.
But revelations after Situational Awareness, an AI-focused hedge fund, suffered losses on leveraged stock investments have raised concerns about how much money is being borrowed, where the money is, how visible it is, and how quickly it will be unloaded, as market participants debate whether the returns matched the scale of its spending.
Nvidia.
How much is being spent on AI infrastructure?
Nvidia’s plan is to develop a platform for AI infrastructure. apollo, black stone, black rock, brookfield, KKR and goldman sachs It may include structures similar to private assets or asset-based financing. Nvidia CEO Jensen Huang told CNBC on Monday that Nvidia’s chips are now “investable infrastructure assets.”
Some tech giants use joint ventures and other leasing vehicles to borrow money for AI data center spending, but the debt doesn’t show up on the balance sheet until the lease begins.
Analysts at Goldman Sachs estimate hyperscalers have $1.5 trillion in leases for data centers, R&D facilities, offices and equipment, up from about $200 billion five years ago. This includes about $1 trillion in “uncommitted” leases that are not yet shown in financial statements but are tied to future payments.
“This could lead to an underestimation of leverage and future liquidity needs, as these obligations are ultimately recognized and contractually due,” Goldman analysts said in an Aug. 6 note.
This surge in bond issuance (including less visible forms of leverage) has focused attention on whether the ultimate returns from AI infrastructure are worth the huge outlay.
Lotfi Karoui, multi-asset credit strategist at PIMCO, said the AI capital investment cycle is on track to become the largest inflation-adjusted investment cycle since railroad construction in the 19th century.
However, in a PIMCO comment dated Aug. 11, he said the ultimate size of the buildup remains “highly uncertain” and emphasized the consensus forecast that hyperscaler capital spending alone will exceed $1 trillion annually from 2027 onwards, with “no clear signs of moderation.”
Karoui said the sheer size of hyperscalar borrowings has led issuers to tap into the euro, pound, yen, Swiss franc and Canadian dollar markets and diversify their bond issuances beyond dollar-denominated paper.
He added that spreads on euro-denominated bonds have outperformed relatively. Amazon and alphabetcompared to their US equivalents, potentially indicating “demand fatigue” in the dollar market relative to euro-denominated paper.
He warned that a wave of AI-related issuance in the US could push spreads higher due to the underperformance of a few large issuers exposed to AI.
Is the use of AI becoming a bigger market risk?
The breakdown in situational awareness showed how vulnerable crowded and leveraged AI trading is to the sharp declines and rallies the sector has seen recently.
The fund was unable to meet a series of margin calls from lenders due to its highly concentrated portfolio, which included names such as: SK Hynix and core weave — Hit by the recent sell-off in tech stocks, assets fell from $45 billion to about $10 billion.
Then Citadel, Ken Griffin’s leading multi-strategy hedge fund, stepped in and bought publicly traded positions in Situational Awareness at a discount. After that, SK Hynix and Coreweave rebounded.
JPMorgan CEO Jamie Dimon recently told CNBC’s Leslie Picker that margin debt is “pretty high,” adding that it increases the risk of increased volatility.
Sahil Mahatani, director of investment research at Ninety One, told CNBC that rising earnings expectations, rather than leverage, are an immediate concern.
SK Hynix.
He said the expectation that “revenues will rise at a high level in the coming years” is the “key risk” that AI trading poses to the market. “This is not a leverage issue, but primarily an expectation issue,” Mahtani said in an email.
He said stock concentration is “historically high” in tech markets, particularly in the U.S., and while it may not be financial leverage, concentration can act like leverage by amplifying market movements when heavily weighted stocks decline.
“The major stock indexes are so concentrated that no one thinks anything can derail them,” Mahtani added, but he cautioned that if companies move from buybacks to additional issuances, support for share prices could disappear, just as AI-related valuations are already under pressure.
Mr Mahtani said the situational awareness problems reflected poor risk management, but that their wider impact had been largely contained.
“That bull market coincided with an unwinding of leveraged ETF structures, primarily in East Asia. Many of these structures, especially single-stock structures, were only launched in the first half of the year. In that sense, this is a relatively subdued case study.”
A spokesperson for the Alternative Investment Management Association, the global trade association for the hedge fund and alternatives industry, said leverage is a “core tool” used by hedge funds to enhance returns and provide market liquidity.
“The key question is not whether hedge funds use leverage, but whether that use poses a significant threat to financial stability. The available evidence does not support treating hedge fund leverage as an inherent systemic risk,” a spokesperson told CNBC.
They said two previous leverage-related failures, the Archegos Capital Management collapse in 2021 and the UK debt-driven investment gold market stress in 2022, involved different structures and investors.
“It is important not to conflate events in very different markets. While Archegos was a family office rather than a hedge fund, the 2022 Gilt case focused on leveraged LDI strategies used by pension funds. We see no reason to expect that the Situational Awareness case itself will trigger a new review of the rules governing hedge fund leverage.”
