Tech companies are always joining forces to gain an edge...
But the AI world has taken it to a whole new level of entanglement.
In March, ChatGPT developer OpenAI signed an $11.9 billion deal with cloud-infrastructure company CoreWeave (CRWV).
In the announcement, CoreWeave said it would use that money to build data centers. And OpenAI would be its main customer.
In May, the two companies expanded that agreement by $4 billion. Then, they expanded it by another $6.5 billion in September.
Also in September, OpenAI announced a similar $300 billion deal for data-center services from tech giant Oracle (ORCL).
Oracle said it would borrow $18 billion to finance the build-out. Then the company upped its borrowing by another $38 billion.
Meanwhile, chip titan Nvidia (NVDA) said in September that it would buy $6.3 billion in data center services from CoreWeave.
That's a follow-on deal from an agreement the companies signed back in 2023, in which Nvidia agreed to pay for any of CoreWeave's unsold data-center capacity through 2032.
Then, Nvidia announced a $100 billion investment in OpenAI. In that deal, OpenAI said it would use the money to build more data centers... which will, of course, use Nvidia chips.
But things didn't stop there...
Right after that, OpenAI announced a deal with Nvidia's main competitor, Advanced Micro Devices (AMD).
OpenAI said it would use AMD chips in some of its new data centers. And in return, it would end up owning 10% of AMD.
It's enough to make your head spin. And it's a lot of money sloshing back and forth between these companies.
The structure of these deals is alarming, regardless of whether AI delivers on its promise. But no one usually gets hurt, unless there's debt involved.
In September and October, tech companies collectively borrowed $75 billion. That's more than double the annual average over the past 10 years.
Now, the AI business has gone from asset-light software to debt-heavy infrastructure in less than two years.
According to estimates, AI capital expenditures will consume about 94% of the tech sector's operating cash flows in 2025 and 2026. That's up from 76% in 2024.
Earlier this month, OpenAI CFO Sarah Friar got people's attention when she said that growing AI might be easier if the government would guarantee this debt.
She ended up walking back those comments. But the cat was already out of the bag...
The fact that anyone at OpenAI would even float the idea shows how risky this whole setup has become.
These companies are betting billions and billions of dollars on a future that isn't totally clear. And that raises the question of who might come up short when this great trade eventually unwinds.
If the CFO of OpenAI is nervous enough to mention government backstops, that tells you something about the risks ahead...
Watch for Massive Fallout If OpenAI Stumbles
There's nothing new about circular investment structures. The problem is that they end in massive losses.
An early example is the South Sea Company in Great Britain, which collapsed in 1720. The company lent money to investors so they could buy more South Sea Company stock.
This created a self-reinforcing cycle that inflated share prices. And when the bubble burst, the only way shareholders could repay their loans was by selling stock at fire-sale prices.
Another example is the margin lending boom here in the U.S. in the 1920s...
Banks lent investors up to 90% of stock purchase prices, using the stocks themselves as collateral.
Rising stock prices enabled more borrowing. And that pushed prices even higher.
When stock prices started to fall in 1929, lenders began to call those margin loans. Forced selling wiped out leveraged investors and sent the economy spiraling into the Great Depression.
Then there's the subprime mortgage crisis in 2008...
Banks made risky mortgage loans, bundled them into securities, and then sold them to investors. Many of those investors were other banks.
They then used those proceeds to make more risky loans. This all worked as long as housing prices went up.
But when housing prices fell, the global financial system nearly collapsed.
Today's AI deals follow the same pattern. Again, the question isn't whether someone gets burned... It's who.
The obvious weakest link in this chain is OpenAI. The company has made roughly $1.4 trillion in infrastructure commitments. But it only expects to have $13 billion in revenues this year.
Bulls will argue that AI will generate trillions of dollars in productivity gains – making these investments look smart. But even if that's true, the timing and structure of these deals matter.
So who gets hurt if OpenAI stumbles?
Nvidia would take a hit. But it wouldn't be a fatal blow. The company would lose its $100 billion investment in OpenAI, plus it would have to write down its CoreWeave obligations.
A $100 billion-plus hit would crater the stock and upset shareholders... but Nvidia has a viable business beyond the OpenAI ecosystem.
It supplies chips to fellow tech titans Microsoft (MSFT), Alphabet's (GOOGL) Google, Amazon (AMZN), and Meta Platforms (META). All of those companies are also building their own AI infrastructures.
AMD would lose a major customer after permanently diluting its shareholders by 10%. For AMD, that would be painful... but not fatal. And like Nvidia, the company also has opportunities outside the OpenAI ecosystem.
CoreWeave would be toast. Its entire business model hinges on OpenAI honoring those commitments. If OpenAI defaults, CoreWeave could face bankruptcy.
It has built capacity specifically for OpenAI. And that capacity might not have other buyers at comparable prices.
Oracle would be another major casualty. If OpenAI can't meet its $300 billion obligation, that leaves Oracle with more than $100 billion in debt and underutilized data centers.
That's a debt-to-commitment ratio that would make even the most aggressive Wall Street dealmakers nervous.
As of its most recent quarter, Oracle had $82 billion in long-term debt. And it's currently in the market with a $38 billion bond offering.
Should OpenAI fail, Oracle would likely face credit downgrades and have to restructure its debt – turning a growth bet into a balance-sheet crisis.
So maybe AI transforms everything and proves all the skeptics wrong. As I've discussed previously here at the Chaikin PowerFeed, AI is creating massive value... it's just a matter of identifying who will capture that value.
Or perhaps AI eventually follows the same path as previous bubbles – with a painful unwinding that wipes out billions of dollars in shareholder value.
To be clear, I'm not saying the AI boom is over... It has the potential to transform how we work and dramatically increase productivity across every industry.
But in either of those scenarios I mentioned, history shows that when the financing becomes circular, someone always ends up holding the bag.
The key is to make sure it's not you.
Good investing,
Joe Austin
Editor's note: There's no question that AI is still dominating the market narrative right now...
But while most investors are chasing the hot AI stocks, some of the richest and most powerful people in America are moving big money into a new part of the market. In fact, this big story has caused 28 stocks to double or more in the first 10 months of this year.
Tomorrow, our corporate affiliate Stansberry Research is hosting an urgent financial summit to cover it all – and how you can take advantage. It's completely free to attend – just register to save your spot and get more details here.
