Blackstone's Jas Khaira Says Fears Over AI Infrastructure Debt Are Overblown, Even as Borrowing Costs Rise

Key Points: How Blackstone Sees AI Debt Concerns

According to TechCrunch, Jas Khaira, a senior Blackstone figure, spoke on stage at TechCrunch Disrupt last week about how the private equity giant (a firm that invests in and buys companies with pooled capital) is positioning itself at the center of the AI boom. TechCrunch says Blackstone's approach includes making large loans to data centers and investing in frontier AI models and applications.

TechCrunch reports that some pessimists worry a large buildup of AI infrastructure debt could lead to a broad credit crisis, a concern Khaira played down. He said: "This doesn’t feel like something where you’ve got 2008 banks that are shaky credits". He noted that the vast majority of data centers are being built for hyperscalers (large cloud companies that run giant data centers), so the related debt is ultimately backed by "the most creditworthy companies in the world".

TechCrunch also notes that even as some infrastructure players come under strain, AI companies still cannot get enough computing resources, and that demand keeps driving data center construction.

Blackstone's AI Strategy and Neocloud Financing Model

TechCrunch reports that earlier this year Blackstone set up Blackstone N1 (BXN1), its AI hub in San Francisco, led by Khaira, who moved from New York for the role. Blackstone manages $1.3 trillion in assets. TechCrunch explains that neoclouds are cloud providers that rent out AI GPU infrastructure; GPUs (graphics processing units) are chips commonly used to train and run AI models. Khaira said: "We’re the largest owner of compute outside of the hyperscalers today", a claim covering Blackstone's investments in neoclouds and data centers.

According to the report, Blackstone pioneered a financing structure for lending large sums to venture-backed neoclouds (young companies funded by venture capital investors). It analyzes the contracts customers have already committed to, and the loans are secured by pledged GPUs as collateral, meaning assets the lender can claim if the loan is not repaid.

TechCrunch says Blackstone's first such deal came in 2023 with CoreWeave. Khaira said CoreWeave then had only "30 million of revenue and minimal profitability" yet asked for a $2.3 billion loan. CoreWeave borrowed against the value of its large collection of Nvidia GPUs and had contracts with blue-chip customers such as Microsoft. According to the report, the deal quickly became a financing template, and Blackstone and other credit firms followed with billions of dollars in further debt for CoreWeave and its rivals.

Khaira said Blackstone first looks at the nature of the customer. If the customer is an investment-grade name such as Microsoft or Meta (high credit rating, low default risk), the risk is priced relative to that customer. TechCrunch also mentions that earlier this year Blackstone co-led, with Coatue, a $10 billion credit facility (a pre-arranged pool of loans a borrower can draw on) for Firmus, an Australian GPU provider. Blackstone has also committed $5 billion to Crux AI, a neocloud formed with Google to commercialize Google's in-house AI chip, the TPU. In addition, Blackstone has backed Etched, a four-year-old chip startup.

What Pressures Are Showing Up in the Market

TechCrunch says Khaira acknowledged that in recent weeks the huge debts piled up by these neocloud operators have started to raise alarms. The report says CoreWeave's debt has swelled to roughly $50 billion. Brad Gerstner, founder of Altimeter Capital, said in a CNBC interview: "The collateral that CoreWeave is putting up which is really only GPUs is not good enough anymore."

Khaira said the unease in credit markets stems from several factors: interest rates rose in September, higher-than-expected inflation signals suggest borrowing costs could keep climbing, and a large volume of AI infrastructure debt is being issued at the same time. He also said credit spreads are widening, meaning lenders are charging a higher risk premium on top of already elevated benchmark rates. Even Meta and Microsoft now pay more to borrow.

TechCrunch also reports that Firmus canceled its initial public offering (IPO, a company's first listing on a stock exchange to raise capital). The listing was planned on the Australian Securities Exchange at a target valuation of about $30 billion. Khaira said "the market wasn’t there at that valuation", so the deal did not go ahead.

What It Means for General Readers

For general readers, the report shows the AI boom is not only a race between models and chips; it also involves how huge amounts of capital are borrowed and who bears the risk. Blackstone's executive believes the risk is manageable, but the same report also lists signs of rising borrowing costs and strain on operators, so both signals coexist.

On the application side, TechCrunch says Blackstone formed a $1.5 billion joint venture this year with Anthropic and Hellman & Friedman called "Ode with Anthropic". It supplies forward-deployed engineers, who work directly with companies to help them adopt AI and build it into their operations, starting with Blackstone's 280 portfolio companies (the businesses Blackstone owns or has invested in). Asked whether those companies are seeing real gains, Khaira said "there is definitely ROI, but it is uneven", and said technology adoption may follow a "sawtooth-type pattern" (rising and falling in a zigzag).

Browse the latest news in this topic

• AI news