🦔Nvidia announced agreements yesterday with the six biggest names in private capital, Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs, and KKR, to raise over $500 billion so its own customers can buy Nvidia chips. These are memorandums of understanding, with final terms still to come. Jensen Huang is calling GPUs a new investable asset class, like real estate or toll roads. The financing lets OpenAI and others buy hardware without putting it on their balance sheets. My Take I flagged Nvidia's customer financing back in July. This is that same idea, now at 60 times the size. Lucent and Nortel ran this exact play in the late 1990s, when they lent customers the money to buy equipment they couldn't otherwise afford. Both ended in the largest corporate bankruptcies of their day. The vendor lends to keep its own revenue up, the sales look strong for a while, and the risk stacks up on the buyer's side of the ledger. We have seen how the story ends. I don't buy the idea that a GPU is an asset like a toll road. A toll road still earns money in 30 years. A GPU is close to worthless in five, and Nvidia's own next chip crushes the resale price of the last one. So they want pension funds and insurers to lend half a trillion dollars against hardware that ages like fruit, and the six firms who put the deal together collect their fees up front no matter how it turns out. Huang keeps his sales up and hands the downside to whoever holds the paper. I have watched this movie twice now, and the vendor never ends up holding the bag. Everyone downstream does. Hedgie🤗

🦔Banks led by Morgan Stanley are planning to offload $15 billion of debt tied to a Google-backed data center in Hubbard, Texas, leased to Anthropic. The bond will carry a speculative-grade rating even with Google's backstop, because Google's guarantee only kicks in after the data center is fully built. Until then, bondholders take on construction delays and cost overruns. Banks that hold this paper want to sell it into the bond market to reduce their exposure. More than $50 billion of similar debt tied to Oracle data centers sits on bank balance sheets and needs buyers. My Take The originate-to-distribute model that Wall Street used with mortgage-backed securities in the mid-2000s is back for AI infrastructure. Banks write the debt, add credit enhancements, and sell it to pension funds and insurance companies that want the yield. The Anthropic bond carries a junk rating during construction and an investment-grade rating after Google's guarantee activates. Subprime mortgage bonds carried investment-grade ratings using similar rating logic while the underlying loans defaulted. The AI capex commitments underneath this debt are enormous. Hyperscaler capex is approaching $3 trillion through 2029. AI lab valuations reach into multiple trillions between OpenAI, Anthropic, and xAI. Data center bonds and neocloud debt add hundreds of billions more. Insurance companies owned by private equity buy this paper and stuff it into insurer balance sheets. Pension funds have raised infrastructure allocations to hold the same debt. The Magnificent 7 makes up over a third of the S&P 500, and their growth story depends on AI capex paying off. If AI capex growth slows even 20 to 30 percent, Google's guarantees, Nvidia's guarantees, and Broadcom's guarantees all get called at once, which pulls in the insurance companies and pension funds that hold the underlying bonds. The 2008 crisis and the dot-com bust each had a single anchor asset that repriced. AI infrastructure has several, and they are already tied together. Hedgie🤗

🦔AI companies are bulk-buying rare books, scanning them through high-speed machines that cut the spines off, and shredding the originals. A service called ISBNdb facilitates orders of up to a million books and keeps buyers anonymous. Pre-2022 books are premium because they're free of AI-generated text. A federal judge ruled the practice is fair use because eliminating the original means only one copy exists at a time. Anthropic hired the former head of Google Books partnerships to obtain "all the books in the world." My Take This got to me. A bookseller told 404 Media that rare books with almost no surviving copies are being fed into this pipeline. Books that survived wars, fires, and centuries of handling are being shredded so an AI can learn to write a better marketing email. ISBNdb's website literally says "'AI company destroys two million books' is not a headline that generates sympathy," and they still built an entire business around making it happen quietly. They offer NDAs as a feature. They coach clients to call it "digital preservation." I've covered AI companies scraping the internet, torrenting libraries, and stealing music. This is worse because it's irreversible. You can re-upload a website. You can reprint a bestseller. You can't replace the last three copies of an 18th-century botanical text once someone shreds them for training data. And the judge said it's legal. So it's going to accelerate. "We shred rare books and offer NDAs so nobody finds out" is a legitimate business model in 2026. What a timeline. Hedgie🤗

🦔Nvidia is stacking up more than $750 billion in AI deals at once, and Bloomberg says the way they're wired is spooking even people who own the stock. There's a $500 billion deal with SK Group in Korea. There's the OpenAI package, where Nvidia backstops up to $250 billion of OpenAI's debt and funds another $350 billion of its chip buys. They mostly run the same loop. Nvidia hands a company money, the company hands it back for chips. Jensen Huang says calling that circular is "ridiculous." I don't think it is. My Take Circular is a fancy word for a simple problem. Nvidia is paying part of the bill for the chips it sells, so some of its blockbuster revenue is just its own cash going out one door and coming back the other. That makes demand look bigger than it is. It also welds Nvidia to OpenAI, a customer that still loses money on everything it does. While OpenAI keeps growing, none of this matters. If it wobbles, Nvidia eats it twice, once on the debt it guaranteed and once on the chip orders that stop. So I don't buy Huang's "ridiculous." A company that has to lend its customers the money to buy its own product is telling you those customers can't pay on their own yet. That's not the top falling out tomorrow, and Nvidia does sell to plenty of real buyers with real cash. But this is exactly the kind of thing that looks genius on the way up and obvious on the way down. I'd rather flag it now while the stock is still near records than explain it to you after a bad quarter. Hedgie🤗

🦔Yesterday's post on $1.65 trillion in hidden AI debt got a lot of attention. Thank you for reading and for the comments and pushback, I really do appreciate both. The most common response was "these aren't debt, they're future obligations." Technically true. But they're binding commitments to spend money that hasn't been earned yet. You can't walk away from them. When those data centers go live, the obligations move from footnotes to the balance sheet all at once. The second was "institutional investors already know." Some do. Most retail investors in index funds and bond ETFs don't read 10-Q footnotes. Goldman's trading desk didn't say "signs of panic" because everyone had it figured out. My Take AI servers last 18 to 36 months. The bonds financing them run 5 to 20 years. Some companies are issuing 40-year bonds to build data centers full of GPUs that will be obsolete before the first coupon payment is due. If AI revenue doesn't cover the bond payments and four or five rounds of GPU replacements, the economics fall apart. To the "they can afford it" crowd... Amazon generates $148.5 billion in operating cash flow and is guiding $200 billion in capex. Microsoft generates $130 billion and is guiding $190 billion. Alphabet and Meta are roughly break-even. Pivotal Research projects Alphabet's free cash flow will drop 90% this year. Amazon's will go negative. The AI buildout is draining the cash that used to fund buybacks, acquisitions, and the cushion that justified their credit ratings. If they could fund it comfortably from operations, they wouldn't be issuing hundreds of billions in bonds for the first time in their histories. Hedgie🤗