Tech giants are using the same playbook that nearly crashed the economy in 2008. The difference this time? They’re not hiding it. Here’s what you need to know before it matters.
THE STORY
Financial analysts are sounding serious alarm bells about how tech companies are financing the AI infrastructure race. A comprehensive NPR investigation reveals that major corporations are repeating patterns that haven’t ended well in the past: circular revenue deals where companies essentially fund each other’s purchases, special purpose vehicles that keep massive debt off balance sheets, and an unprecedented $121 billion in new debt accumulated by hyperscaler companies in just the past year.
Read the full story: https://www.npr.org/2025/11/23/nx-s1-5615410/ai-bubble-nvidia-openai-revenue-bust-data-centers
THE TAKEAWAY
Here’s what’s happening: Nvidia is giving OpenAI $100 billion so OpenAI can buy Nvidia’s chips. Google, Microsoft, and Amazon are doing similar things with their own ecosystem partners. None of this is illegal. None of it is necessarily hidden. But it creates a fundamental problem: we can’t tell what the real demand for AI actually is because much of the spending is circular. Companies funding their own customers to maintain the appearance of booming sales.
That matters because the entire financial structure supporting this boom is built on one assumption: that AI will generate massive, immediate revenue. If that doesn’t happen because growth slows, or if AI’s real-world applications turn out to be narrower than expected, the debt becomes a problem. Not a “startup failure” problem. A “systemic financial system” problem.
The most troubling part? This is literally what happened before 2008. During the housing bubble, complex financial instruments obscured real risk. Special-purpose vehicles hid debt. Circular arrangements made bad assets look valuable. Regulators and investors couldn’t see the risk until the whole system nearly collapsed.
The difference now: we know what to look for. We’ve studied it. We understand the pattern. And it’s happening again, largely in plain sight.
The hope: We’re not there yet. Analysts like Gil Luria at D.A. Davidson are publicly flagging this. Investors like Michael Burry are actively betting against companies they believe are hiding real demand problems. The conversation is happening before the crash, not after. And it’s an actual conversation this time. While Michael Burry was outspoken before the housing crisis, it made people realize he might be worth listening to.
Read this post and more by becoming a subscriber to my Substack!
WHAT YOU SHOULD DO
If you’re an investor or managing a portfolio:
- Don’t ignore the circular deal structure. When a major tech company starts heavy “partnership” financing with capital firms, ask what real demand looks like without that financing.
- Watch cash flow projections. Morgan Stanley estimates Big Tech will spend $3 trillion on AI infrastructure through 2028, but their own cash flows can only cover about half of that. Where’s the rest coming from? If it’s debt-dependent, that’s a risk signal.
- Consider your exposure to companies that are betting everything on immediate AI monetization. The companies most exposed to a demand slowdown are those that can’t pivot if returns disappoint.
If you’re a general reader:
- Start asking questions when you see headlines about massive tech deals. “Who’s funding it? Is it their own cash or borrowed money? Who benefits if growth slows?” These are the questions that matter.
- Pay attention to what regulators say about this. Regulatory scrutiny of tech infrastructure financing is likely coming. That will move markets.
- Recognize that a potential AI correction wouldn’t just affect tech stocks… it could ripple through pension funds, insurance companies, and retirement accounts that have invested in these deals. Your financial security might depend on this staying stable.
Most importantly: Demand transparency. We learned in 2008 that complexity in finance is usually a red flag. If you can’t understand how a $27 billion deal is structured and why, that’s a problem worth asking about.
Read what I have to say about AI on my Substack.