On the surface, the AI revolution runs on silicon chips and brilliant code. But dig a little deeper, and you’ll find its true fuel source: a web of deals so intricate, it looks less like a technological marvel and more like a trillion-dollar IOU written on a napkin between Nvidia and OpenAI. Here’s how the money-go-round works, and why it might spin out of control.

 Deals:

N.Y. Stock Exchange Parent to Buy Stake in Polymarket, as Prediction Markets Go Mainstream

Industry News:

Gold price tops $4,000 for first time

World:

America’s monetary policy risks getting too loose

Let’s talk about a number so big it basically breaks your brain: $1 trillion.

In just a few weeks, the AI world has seen deals for computing power that could easily rocket past that figure. For context, that’s more than the entire economic output of Switzerland.

This ocean of cash is being spent on a technology that, for all its world-bending promises, is still figuring out how to turn a consistent profit.

At the center of it all are two names you know: Nvidia and OpenAI. Their relationship has become the blueprint for a bizarre new business model that’s raising some very uncomfortable questions.

It’s a financial feedback loop so slick, you have to respect the hustle.

Imagine you own the only factory making pickaxes during a gold rush. Business is insane. But you want more.

So, you find a promising prospector and offer him a $100m loan. The only catch? He has to spend every penny on your pickaxes.

His purchase shows up as your revenue. Your stock soars. Wall Street calls you a genius for backing the next big thing.

But did you actually create new demand? Or did you just use your own cash to fund your own sales?

That, in a nutshell, is the central critique of the AI money machine. Nvidia, the undisputed king of AI chips, is acting as both the primary supplier and one of the biggest financiers for the entire industry.

It started with a monster deal: Nvidia agreed to invest as much as $100B in OpenAI. In return, OpenAI committed to filling its new data centers with millions of Nvidia’s chips.

This isn't a one-off. It's a web of capital propping up a multi-trillion-dollar boom, and it’s getting more tangled by the day.

Let’s meet the players.

This isn't a simple story of buyers and sellers. It's a high-stakes poker game where everyone at the table seems to be bankrolling each other.

Nvidia: The House Always Wins. As the dealer in the AI casino, Nvidia has ridden the wave to a staggering $4.5T market cap. Its strategy is brutally effective: make sure the whole world runs on its tech. To speed things up, Nvidia has become a prolific venture capitalist, investing in 52 different AI companies in 2024 alone. Many of these startups take Nvidia’s cash and… immediately spend it on Nvidia chips.

Exhibit A: Its relationship with CoreWeave, a cloud provider whose entire business is renting out Nvidia GPUs. Nvidia was an early investor. But the masterstroke is a $6.3B “backstop” agreement, promising to buy any of CoreWeave's unused cloud capacity through 2032. This guarantee lets CoreWeave borrow billions to buy… you guessed it, more Nvidia chips. Nvidia isn’t just selling the pickaxes; it’s providing the financing and the insurance policy for the miners.

OpenAI: The Ambitious Prodigy. As the creator of ChatGPT, OpenAI is the face of the revolution. It’s also burning cash at an astonishing rate. To feed its GPU hunger, it’s also cut a multi-billion dollar deal with Nvidia’s rival, AMD, that includes a warrant to buy up to 10% of the company. It also inked a $300B deal with Oracle for cloud infrastructure. The kicker? Oracle is spending billions on Nvidia chips to fulfill that deal, sending money right back up the chain to one of OpenAI's biggest backers.

Elon Musk's xAI: The Wild Card. Never one to miss a party, Musk is raising a cool $20B for his AI startup. And who’s a key investor? Nvidia, which is reportedly chipping in up to $2B. A huge chunk of the funding is being raised via a special vehicle that will take on $12.5B in debt specifically to buy Nvidia GPUs. It’s a clever way to isolate risk while guaranteeing a massive sale for the chip king.

For anyone who remembers the dot-com bubble, the echoes are deafening. The breathless hype, the insane valuations for unprofitable companies… it’s all here.

Back then, the mantra was "get big fast." Companies like Pets.com burned through millions in Super Bowl ads to build "brand awareness," only to go bankrupt nine months after an $82.5m IPO.

The ghost in that machine was a "marketing/finance feedback loop."

VCs would give Startup A (Pets.com) $50m. Startup A would spend $10m on banner ads on Startup B's website (say, Yahoo). Startup B would book that $10m as revenue, boosting its stock. Money just moved from one VC-funded pocket to another, creating the illusion of a booming economy.

When the Fed raised interest rates, the cheap money vanished, and the whole thing collapsed. The NASDAQ plummeted 78%, wiping out trillions.

The difference today, bulls argue, is that the AI boom is building something tangible. Instead of banner ads, they're buying GPUs.

But this might be a distinction without a difference. In the late ‘90s, telecom companies, convinced of infinite demand, spent $500B laying fiber optic cable. This created a massive glut of capacity that bankrupted many of them when the bubble burst.

The AI boom has simply substituted tangible, but potentially idle, hardware for intangible hype.

Of course, those are the four most dangerous words in investing. But there’s a case to be made.

Unlike the profitless startups of 1999, Nvidia is a cash-printing machine with insane profit margins. And unlike a hypothetical business plan on a napkin, ChatGPT is a real product used by hundreds of millions of people.

CoreWeave CEO Michael Intrator brushes off the circular financing concerns, pointing to real customers. "When Microsoft comes to us to buy infrastructure to deliver to its clients who are consuming 365 or Copilot," he says, "I don't care what the narrative is... They have end users that are consuming it."

The argument is that these deals are just priming the pump for a tidal wave of real, organic demand.

Viewed another way, this tangled web of deals might be a necessary evil—a new form of collaborative risk management. Building AGI could cost trillions, a sum too vast for any single company. This structure spreads the astronomical risk across the ecosystem.

In this view, the circularity isn't a bug; it's a feature.

So, is this a "virtuous cycle" building the future, or a trillion-dollar echo chamber waiting to implode?

The truth is probably somewhere in the middle.

The biggest risk is how tightly bound everyone is. In 1999, the failure of Pets.com didn’t kill Yahoo. Today, a major downturn for OpenAI would send catastrophic shockwaves through Nvidia, AMD, Oracle, and the entire ecosystem. The risk isn’t isolated; it’s systemic.

The AI boom is building cathedrals of computation at a breathtaking pace. The tech is real and the potential is immense.

But the financial architecture looks suspiciously like a sophisticated pile of IOUs. The only question that matters is whether these cathedrals are being built on a foundation of solid rock—or on quicksand.

Join 4M+ professionals who start their day with Morning Brew—the free newsletter that makes business news quick, clear, and actually enjoyable.

Each morning, it breaks down the biggest stories in business, tech, and finance with a touch of wit to keep things smart and interesting.

Try it yourself (for free)

Have a Question? Let's Hear It!

I love building this newsletter with you in mind, and my favorite part is hearing from you directly.

Have a burning question, a great idea, or a thought you'd like to share? Send it my way! I'll pick some of our favorite submissions to answer and feature in next week's edition.

>> Click Here to Share Your Thoughts <<

P.S. I read every single submission! Let me know if you'd prefer to remain anonymous.

See you next week,

If you enjoy my emails, move them to your primary inbox and let me know by sending a reply or clicking the poll below. I read every response.