Blackstone Defends AI Debt Play as CoreWeave's Stack Hits $50 Billion

CoreWeave's debt has swelled to roughly $50 billion, and shares of the neocloud have shed more than 20% in recent days—but Blackstone's AI chief says fears of an AI-finance meltdown miss the point.

Speaking at TechCrunch Disrupt last week, Jas Khaira, who leads Blackstone N1 (BXN1), the firm's San Francisco–based AI hub, pushed back against critics who compare today's AI lending boom to the 2008 mortgage crisis. "This doesn't feel like something where you've got 2008 banks that are shaky credits," Khaira told the audience.

Blackstone, with $1.3 trillion in assets under management, has emerged as the dominant non-hyperscaler in AI compute. "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.

Where Blackstone has put its chips

The firm has wired that thesis into a string of deals. BXN1 committed $5 billion to Crux AI, a neocloud co-launched with Google to commercialize Google's custom TPUs. It also backed Etched, a four-year-old chip startup. Earlier this year, Blackstone co-led a $10 billion credit facility for Australian GPU provider Firmus alongside Coatue.

The template for this lending model came in 2023. CoreWeave, fresh off a ChatGPT-fueled demand surge, walked into Blackstone with just $30 million in revenue and "minimal profitability," Khaira said. Blackstone wrote a $2.3 billion loan, secured by Nvidia GPUs and anchored to contracts with investment-grade customers such as Microsoft and Meta. That structure—customer credit plus hardware collateral—became the playbook.

"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.

What changed for the neoclouds?

The cracks showed up this fall. Interest rates ticked higher in September, hotter-than-expected inflation hinted at more tightening, and a wave of AI-infrastructure debt hit the market at once. "The cost of capital to price in debt for a Meta or a Microsoft has widened out," Khaira said. Lenders now demand fatter spreads, even from tech titans with pristine balance sheets.

CoreWeave's debt has since ballooned to about $50 billion. On CNBC, Altimeter Capital founder Brad Gerstner warned: "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."

The market voted. Firmus pulled its planned Australian Securities Exchange listing, which had targeted a roughly $30 billion 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.

What does this change for lenders and operators?

Khaira's argument: panic is misplaced because the underlying credit isn't mortgage paper or subprime—it is hyperscaler contracts. Most data centers being built today serve Google, Microsoft, Meta, and Amazon, he said, so the debt financing them is ultimately backed "by the most creditworthy companies in the world." In his framing, the recent repricing reflects risk reassessment, not a coming collapse.

For investors, the practical signal is that the era of GPU-only, customer-contract collateral is closing. Spreads have widened, leverage is harder to stack, and IPO windows shut when valuations outrun the market's appetite. CoreWeave's stock slide and Firmus's pulled listing are the cleanest evidence.

For the neoclouds themselves, the calculus is more urgent. Higher borrowing costs compress the gap between hardware depreciation and customer revenue, especially for shops running Nvidia-only stacks with shorter useful lives. Firms with diversified architectures—Crux's TPUs, Etched's purpose-built silicon—gain a structural edge in financing talks.

Is the AI buildout itself slowing?

Khaira doesn't think so. Demand still outstrips supply, data center construction continues, and Blackstone is widening the bet. This year the firm launched "Ode with Anthropic," a $1.5 billion joint venture with Anthropic and Hellman & Friedman that embeds forward-deployed engineers inside companies to help them adopt AI. The first stop: Blackstone's own 280 portfolio companies.

Asked whether those portfolio companies are actually seeing returns, Khaira conceded the picture is mixed. "There is definitely ROI, but it is uneven," he said, describing adoption as a "sawtooth-type pattern" that can look like oversupply before usage catches up.

For practitioners, that's the takeaway: the money is still flowing, but the price of that money has reset. Neoclouds, their lenders, and the hyperscalers buying their capacity all face a steeper climb through 2026. Blackstone—the shop that built the playbook—is now the one telling the market the model still holds.