Blackstone Defends AI Infrastructure Debt as Market Risks Rise

Blackstone executive Jas Khaira is pushing back against concerns that mounting debt from artificial intelligence infrastructure investments could trigger a broader financial crisis. Speaking at TechCrunch Disrupt last week, Khaira argued that the industry's financing risks remain manageable because major technology companies underpin much of the demand for computing capacity and data centers.

Blackstone, which manages $1.3 trillion in assets, is financing the AI expansion through data center lending, investments in advanced AI models, and technology companies. Earlier this year, the investment firm established Blackstone N1 (BXN1), a San Francisco-based unit led by Khaira to coordinate its AI initiatives.

Khaira said Blackstone has become a major investor in computing infrastructure, including specialized cloud providers known as neoclouds, which rent out computing capacity powered by graphics processing units (GPUs).

"We're the largest owner of compute outside of the hyperscalers today, whether it's investments in neo clouds, investments in data centers, and being able to use that compute strategically to support companies," Khaira said.

The firm's investments include a $5 billion commitment to Crux AI, a neocloud launched in partnership with Google to commercialize its custom tensor processing units (TPUs). Blackstone has also backed Etched, a chip startup that has attracted investor attention for its rapid development.

Blackstone has developed financing arrangements that evaluate customer contracts alongside GPUs pledged as collateral. The approach helped the firm extend substantial credit to companies building computing infrastructure.

Its first deal of this kind came in 2023, when CoreWeave sought a $2.3 billion loan after demand surged following the launch of ChatGPT. Khaira described the company's financial position at the time as having "30 million of revenue and minimal profitability," highlighting the role of its computing assets and customer agreements in securing financing.

CoreWeave had a large inventory of Nvidia GPUs, and Nvidia was also an investor. Contracts with major customers, including Microsoft, provided additional support for the financing arrangement.

Khaira explained that customer creditworthiness remains central to Blackstone's lending decisions.

"One of the primary things that we're looking to is the nature of the customer. And if the customer is an investment-grade customer — let's say it's Microsoft or let's say it's Meta — the risk gets priced relative to that customer," Khaira said.

Blackstone subsequently expanded its lending to other infrastructure providers. Earlier this year, the firm co-led a $10 billion credit facility for Australian GPU provider Firmus alongside Coatue.

However, rising borrowing costs and growing debt loads have intensified scrutiny of the sector. CoreWeave's debt has reached approximately $50 billion, while concerns about the value of its GPU collateral have raised questions about lenders' exposure.

Altimeter Capital founder Brad Gerstner questioned whether the assets still provide sufficient security for new loans.

"The collateral that CoreWeave is putting up which is really only GPUs is not good enough anymore. They used that in the past to get cheap debt," Gerstner said in a recent CNBC interview.

Khaira attributed the growing concern among credit investors to higher interest rates, persistent inflation pressures, and a wave of debt offerings tied to AI infrastructure expansion.

"We're also seeing higher credit spreads," Khaira said.

He added that borrowing costs have increased even for major technology companies with strong financial positions.

"The cost of capital to price in debt for a Meta or a Microsoft has widened out," Khaira said.

The pressure has also affected company valuations and fundraising plans. CoreWeave shares fell more than 20% in recent days, while Firmus abandoned a planned initial public offering on the Australian Securities Exchange that targeted a valuation of approximately $30 billion.

Khaira attributed Firmus' canceled listing to market conditions that did not support its proposed valuation.

"Ultimately, the market wasn't there at that valuation, and so I think as a result, that deal didn't get done," Khaira said.

He said investors are reassessing the value of infrastructure providers as financing conditions tighten.

"The market is thinking what is the right price for neoclouds, what is the right price for these businesses in this moment when interest rates are moving up," Khaira said.

Despite the financial pressures, Khaira said demand for computing power remains strong as AI developers continue expanding their models and applications. He rejected comparisons with the 2008 financial crisis, arguing that the current infrastructure financing market has a different underlying credit profile.

"This doesn't feel like something where you've got 2008 banks that are shaky credits," Khaira said.

He pointed to the concentration of data center construction around major cloud providers, known as hyperscalers, as a source of financial support for the sector.

Khaira said the debt financing much of this infrastructure is ultimately supported "by the most creditworthy companies in the world."

Blackstone is also investing in AI adoption beyond physical infrastructure. This year, it formed a $1.5 billion joint venture with Anthropic and Hellman & Friedman called Ode with Anthropic, which supplies engineers to help businesses integrate AI into their operations, beginning with Blackstone's 280 portfolio companies.

Asked whether companies owned by Blackstone were seeing measurable returns from AI, Khaira said the results varied across businesses.

"There is definitely ROI, but it is uneven," Khaira said.

He described AI adoption as a process that can advance unevenly, potentially leaving infrastructure capacity ahead of immediate demand while businesses gradually integrate the technology.

"Sawtooth-type pattern," Khaira said, describing the pattern of adoption that can make computing capacity appear excessive before implementation catches up.