Blackstone’s top AI dealmaker doesn’t think the debt behind the AI buildout will become the next credit crisis. Speaking at TechCrunch Disrupt, Jas Khaira, who runs the firm’s San Francisco AI hub Blackstone N1 (BXN1), dismissed fears that borrowing for data centers and GPU clouds could bring down the financial system, according to TechCrunch AI. “This doesn’t feel like something where you’ve got 2008 banks that are shaky credits,” he said.
Few investors are more exposed to this question. Blackstone manages $1.3 trillion in assets, and Khaira says it’s “the largest owner of compute outside of the hyperscalers today.” So his view carries weight, and it also comes with a clear stake in the outcome.
Blackstone wrote the GPU lending playbook
Much of today’s neocloud financing started with a Blackstone deal. Neoclouds are cloud providers that rent out AI GPU capacity. In 2023, CoreWeave had only “30 million of revenue and minimal profitability,” according to Khaira, yet it asked Blackstone for a $2.3 billion loan.
Blackstone agreed for two reasons:
- Collateral: CoreWeave pledged its large stock of Nvidia GPUs.
- Customers: It held contracts with blue-chip buyers like Microsoft.
That structure became the template for billions in follow-on debt across the sector. Blackstone has since co-led a $10 billion credit facility for Australian GPU provider Firmus alongside Coatue, committed $5 billion to Crux AI (a neocloud built with Google to commercialize its TPUs), and backed chip startup Etched.
Khaira’s pricing logic is simple: “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.”
Why the jitters are showing up now
Khaira admitted the mood has shifted in recent weeks. He pointed to three pressures:
- Rising rates: Rates ticked up in September, and hotter-than-expected inflation suggests borrowing costs could keep climbing.
- Supply glut: A flood of AI infrastructure debt is hitting the market all at once.
- Wider spreads: Even Meta and Microsoft now pay bigger premiums to borrow.
The market has noticed. CoreWeave’s debt has grown to about $50 billion, and its shares dropped more than 20% in recent days. Firmus scrapped its planned ASX listing at a target valuation of about $30 billion. “Ultimately, the market wasn’t there at that valuation,” Khaira said.
Altimeter Capital founder Brad Gerstner went further on CNBC:
“The collateral that CoreWeave is putting up which is really only GPUs is not good enough anymore.”
Two readings of the same numbers
Khaira’s argument is that most data centers are built for hyperscalers, so the debt behind them is ultimately backed “by the most creditworthy companies in the world.” Demand is still there too. AI labs can’t get enough compute.
What stands out to me is how much that argument depends on the type of borrower. It’s strongest for data centers leased long-term to Microsoft or Meta. It’s weakest for neoclouds whose main collateral is hardware that loses value every time Nvidia ships a new generation. Gerstner is aiming at that second group.
In fairness to Khaira, his early CoreWeave bet looks smart so far, and his 2008 comparison has a point: this risk isn’t sitting on shaky bank balance sheets the way mortgage debt did. But “not 2008” is a low bar. A repricing that squeezes neocloud valuations and cuts off cheap financing can do real damage without becoming a systemic crisis.
The adoption side: ROI is “uneven”
Blackstone is also pushing on demand. With Anthropic and Hellman & Friedman, it formed Ode with Anthropic, a $1.5 billion joint venture that places forward-deployed engineers inside companies, starting with Blackstone’s 280 portfolio companies. Asked whether those companies are seeing gains, Khaira was candid: “there is definitely ROI, but it is uneven.” Coming from someone with every reason to talk it up, that’s a revealing admission.
What to do with this
- Compute buyers: Don’t treat every GPU provider as equally stable. Check your neocloud’s debt load and customer concentration, and keep a backup provider.
- Investors: Don’t lump all “AI debt” together. Hyperscaler-backed data center debt and GPU-collateralized neocloud debt are very different risks.
- Enterprises: Uneven ROI means you should pick narrow, measurable use cases first. Blackstone itself is paying for hands-on engineers instead of just buying licenses, and that’s worth copying.
The next few quarters will test Khaira’s thesis. If rates keep rising, expect neocloud valuations, not hyperscaler balance sheets, to show the strain first. Full details of the interview are available at TechCrunch AI.