🥊 Core bob and weave

Coreweave had a wild week, with anxieties about its debt spurred by a rather candid comment by one of its founders — and stoked by Elon Musk — spiraling into a broader moment of anxiety around just how much money the AI revolution is requiring investors to put up. A full-court press from CoreWeave’s boosters appeared to reset expectations, but where does this all leave us as we assess the state of the AI trade?

First, a recap: 

• CoreWeave shares sold off as much as 15% on Wednesday after a remark from a cofounder about the company’s attempts to get an investment-grade rating that would allow it to borrow from insurers rattled markets, with the executive describing the capital as “unlimited, like, that’s all the money in the world that sits in the insurance tranche.”

• That put a bit of a spotlight on the company’s debt situation, as the 2031 CoreWeave bonds yield 15%, which is very very high for a company in its position.

• Well, amid the renewed scrutiny the cost of protecting that debt against default for five years rose to 1,194 basis points, 74.5 basis points in a day, swaps pricing in a 60% chance of default within five years.

• “Yikes,” wrote Elon Musk in a post. 

Wow, one might think, even Elon Musk is worried about the AI bubble? No. No he is not. Don’t be ridiculous. 

• Elon Musk simply owns a competitor: SpaceX is rapidly attempting to build out large AI data centers that can both reduce its reliance on — and ultimately compete with — neoclouds like CoreWeave. 

• In fact, as Grok and the other products of SpaceX’s AI lab continue to lag rival products out of Anthropic, OpenAI, Alphabet, Meta and so on, the core shareholder value of SpaceX’s AI business will be the neocloud element that can eventually sell compute to those winning teams. 

• Remember, SpaceX cut a $1 billion a month deal to sell compute in September, which of course leads to the question of “why does SpaceX have excess compute?” 

• Elon’s “yikes” is, if anything, more kicking a rival when they’re down rather than any concession that the technology he has tied his fortune to has issues. 

So, where does that leave us? Well, CoreWeave stock finished up 5% Thursday, still down 17% over the five-day, with its CEO waving the week away saying “there will be periods of time when the market misunderstands and acts irrationally.”

The Takeaway

The unique cultures of Silicon Valley and Wall Street have been in general alignment that the AI boom is going to make a lot of people a lot of money. But before we get cocky, it’s worth remembering two axioms held as truth in either camp that apply very well :

• From the West coast: the future is already here — it's just not very evenly distributed.

• From the East coast: markets can remain irrational longer than you can remain solvent.

The AI boom’s benefits to shareholders will not necessarily be evenly distributed. If, for instance, there are one or two models that get a lot of pickup or achieve milestones faster than rivals can achieve, even companies that get lots of money in investment will not be worth a whole lot. If the world runs all its AI out of data centers, the neoclouds win; if, on the other hand, we don’t need all these buildings and you can eventually run a solid model on a phone, then the neoclouds will bite it. 

Heck, maybe free Chinese open source models can do 97% of the work of AI, and the American labs become boutique R&D firms arguing with academics over whether they actually solved theoretical math problems. 

In short: There are lots of worlds in which both LLMs are popular, useful technologies integrated into the American economy, and also some of the companies getting high valuations don’t exist anymore. 

Sure, which companies win and which companies lose that future could be decided by breakthroughs in AI model development or a phalanx of Nvidia chips powered by fusion power or a data center deployed to space, or any of these big and exciting reasons that a company might win that investors like to imagine. But for lots of them, their victory or their death will be decided by the rather boring parts of the balance sheet describing how much money they owe creditors in 2032. 

Whether this week was actually the axis upon which the AI trade turns is a bit premature. But it is the shape of things to come, a taste of what the future will at some point bring, at least when everyone flips their cards on the table and the winnings get divided. Just because your business model was the beating heart of the AI industry in 2026 doesn’t prevent you from being the bloated, doomed appendix of the AI industry in 2032. Not every buzzy well-funded company’s gonna make it. Just ask Character.ai. If anything, this week showed that thing that kills them isn’t even guaranteed to be interesting, either. 

— Walt Hickey

• 🏈 CFB: This weekend has a few big college football games, but none bigger than No. 4 Ohio State taking on reigning national champions and No. 5 Indiana University. OSU is the 62% favorite to win. 

• 🏈 NFL: This weekend’s slate of NFL games is looking pretty good, with the Steelers 58% favored over the Tampa Bay Buccaneers, the Green Bay Packers having a slight 54% edge  in a conference showdown against the Dallas Cowboys, and a 50-50 matchup between the Chicago Bears and the Atlanta Falcons. 

*Event contracts are offered through Robinhood Derivatives, LLC — probabilities referenced or sourced from KalshiEx LLC or ForecastEx LLC.

July saw the 100,000,000th Tamogatchi sale, which comes as owner Bandai tries to revive the brand of digital pets on the toy’s 30th anniversary this coming November and make in-roads across Asia.