At the lows on Thursday, CoreWeave was down more than 25% since the beginning of the week.

Had the losses held, the poster child for “subprime AI,” as one bear ungenerously describes the largest neocloud by revenue, would be one bad session away from notching its worst weekly performance as a public company.

Sentiment was dastardly following the midweek mini-panic that landed CoreWeave above the fold across the digital mediasphere. Elon Musk’s one-word assessment of the plunge in the company’s bonds — “Yikes” — didn’t help.

The bearish pile-on was such that CoreWeave risked succumbing to a self-fulfilling prophecy. In a testament to the perceived peril, the rumor mill was rife with speculation that Nvidia, which has a 10% stake, might have to rescue the company.

On Thursday, Jensen Huang weighed in. CoreWeave “doesn’t need” a backstop, he told Bloomberg. “Their cost of capital is actually decreasing.”

That was enough to stabilize the stock, but who knows for how long. A quick scroll through finance-focused social media suggests the peanut gallery’s still very much inclined to perpetuate a doom spiral. Indeed, to read some of the more abrasive color you’d think CoreWeave had already collapsed.

To be clear: This isn’t a dip that I’d buy personally. But as I mentioned in a comment on Wednesday evening, I doubt seriously that Nvidia and CoreWeave’s hyper-scaler clients would sit idly by in the event deteriorating sentiment pushes the company to the brink, which is possible when your business model’s as sensitive as CoreWeave’s is to the vagaries of capital markets’ manic cycles.

The figure above shows you the stock price (inverted) with the cost to insure against a default. At nearly 1200bps, CoreWeave’s CDS implied better-than-even odds of a credit event by 2030 when, according to a Gartner analysis, eight in 10 neoclouds may be out of business.

CoreWeave co-founder Brannin McBee didn’t do the company any favors when he told the FT that CoreWeave’s dream is to secure an IG rating so it can — and I’m quoting McBee directly —  borrow “unlimited” sums of money from insurers. “Like, all the money in the world sits in the insurance tranche,” he said.

That was an own-goal. When the main concern around your company is its debt burden and your capacity to service that debt, telling the world’s foremost financial media outlet that if only you could, you’d leverage yourself to infinity, probably isn’t the best idea.

I’m in no position to dispute McBee’s contention that CoreWeave would be “immensely cash-flow positive” if it throttled back its expansion plans, but as the linked FT article obliquely notes (i.e., says without saying it), CoreWeave’s near-term IG aspirations look far-fetched set against consensus projections for negative free cash flow out to 2028.

The figure above’s worth highlighting: Put open interest on CoreWeave hit a record this month.

Not everyone’s as skeptical. If you ask Deutsche Bank — which knows a thing or two about lending to controversial borrowers — the plunge in the stock’s a buying opportunity.

“While more capital will be needed, this can happen at a steady pace and is backed by robust customer backlog,” the bank’s Brad Zelnick said, in a note, adding that CoreWeave’s “strong customer base, established record of execution, backlog conversion and leading infrastructure performance” support its issuer profile “relative to neocloud peers.”

Zelnick did concede the obvious. “There’s no doubt CoreWeave’s future is tied to the success of AI and demand for the underlying compute infrastructure supporting it,” he wrote, in the same note. “Thus its shares should reflect changes in sentiment here.”

Right, and that’s really the crux of the issue. The very near-term risk for CoreWeave is a negative sentiment spiral. Conceptually speaking, all leveraged business models are confidence games, which means confidence crises pose an existential risk.

Late Wednesday, when the shares lay smoldering in the ashes of that session’s mini-crash, The Information sneered at Michael Intrator’s characterization of the price action as “irrational.”

“He sold CoreWeave stock every week between April and the end of September while his fellow co-founders have also been selling regularly,” Martin Peers, the site’s editor-at-large, wrote. “Intrator could win some brownie points by pledging to stop selling his stock for a while.”