"Not that AI is overvalued, but that the entire system supporting AI valuations is built on borrowed money, borrowed time, and borrowers who are running out of both..."

Luis Flavio Nunes

Here's a puzzle for you: Nvidia (NVDA) reported $57.01 billion in quarterly revenue, beat expectations by billions, and guided for $65 billion next quarter. The stock jumped 5% at the open. Then it all evaporated. By the close, the Nasdaq (QQQ) was down 2.38%, the S&P 500 (SPY) lost 1.56%, and the VIX spiked 11.7% to its highest level since April.

Most people chalked it up to profit-taking. But the real story is more interesting, and more troubling. When margin debt surges 45% while consumer sentiment plummets to near-record lows, something has to give. Thursday, it started giving.

Two Towers of Debt

The AI boom rests on two pillars. One is Wall Street leverage, the borrowed money investors use to chase the next big thing. The other is Main Street spending, the consumer demand that ultimately pays for all those AI services. Both pillars are cracking at the same time.

Start with Wall Street. Margin debt, the money investors borrow from their brokers to buy stocks, hit $1.18 trillion in October. That's up $58 billion in a single month. Since April, margin debt has climbed 39%, the fastest five-month run since October 2021. Remember what happened after that surge? The S&P 500 dropped 25% over the following year.

That kind of leverage amplifies everything. When Nvidia opens up 5% and then reverses, margin calls start hitting. If you borrowed money to buy at $140, your broker doesn't care about your long-term thesis when the stock hits $130. They sell your position to cover the loan. Other investors get forced out at the same time, and systematic strategies pile on. Suddenly a great earnings report ends with a selloff driven by mechanical liquidation.

But margin debt alone doesn't usually break a market when companies are still posting $57 billion quarters. What makes this different is what's happening on Main Street.

The Consumer is Tapped Out

The University of Michigan Consumer Sentiment Index dropped to 50.3 in November, the lowest reading since June 2022 and down nearly 30% from a year ago. The current conditions index hit a record low. Around 71% of households now expect unemployment to rise over the next year. That's more than double the share from a year earlier.

Those aren't just vibes. U.S. credit card debt reached $1.233 trillion in the third quarter, the highest figure since the New York Fed started tracking this in 1999. The average household is carrying $9,326 in credit card balances at an interest rate of about 22.25%. Before the Fed started raising rates, typical credit card rates ran between 13% and 16%.

Delinquencies are climbing everywhere. In the poorest 10% of ZIP codes, the 90-day credit card delinquency rate jumped to 22.8% in early 2025 from 14.9% in late 2022. Even in the wealthiest 10% of ZIP codes, serious delinquencies have increased by more than 70% over the same period. When rich people start missing credit card payments at that rate, something fundamental has shifted.

The Phantom Debt Layer

Official credit statistics miss an entire category of household borrowing: buy-now-pay-later plans. These are the four-installment payment schemes you see at online checkout, offered by companies like Klarna, Affirm, and Afterpay.

Recent surveys show 42% of BNPL users made at least one late payment in 2025, up from 39% in 2024 and 34% in 2023. Most of these loans don't get reported to credit bureaus. A consumer can juggle several BNPL plans across different providers, and traditional lenders won't see the full picture. About 61% of BNPL borrowers fall into subprime or deep-subprime categories, meaning this hidden debt is concentrated among people already struggling with high-rate credit card balances.

Nobody knows exactly how large this shadow lending market is, but it's growing fast. And it's adding another layer of strain to households that official statistics say are already stretched thin.

The Holiday Season Warning Sign

PwC's 2025 Holiday Outlook survey shows consumers expect to spend $1,552 on average during the holidays, down 5% from 2024. That's the steepest projected decline since the pandemic. Gift spending is expected to fall 11%.

Generation Z is pulling back hardest, planning to cut holiday spending by 23%. A quarter of Gen Z respondents say their financial situation is worse than a year ago, up from 17% in the prior survey. Young consumers are supposed to be the early adopters of AI-powered everything. If they're cutting back this aggressively, what does that mean for the demand assumptions baked into tech company growth models?

Why This Threatens the AI Story

The bull case for AI stocks assumes data centers stay busy and demand for AI services keeps growing. But that demand ultimately comes from consumers and small businesses buying AI-enabled products and services.

If 71% of households expect higher unemployment, if credit defaults are rising across all income levels, and if holiday spending plans are falling off a cliff, then enterprise AI spending becomes fragile. Projects get delayed. Budgets get trimmed. Adoption slows down.

The hyperscalers backing Nvidia's growth are expected to spend more than $300 billion on capital expenditure in 2025, using about 70% of their combined cash flows. That bet only works if AI services generate solid returns. Early surveys suggest fewer than half of enterprise AI projects are actually profitable, and only a small share have scaled across entire organizations.

When the foundation is shaky, it doesn't matter how impressive the superstructure looks.

The Fed's Impossible Choice

Normally when consumers are struggling and delinquencies are rising, the Federal Reserve cuts rates. This time is messier. Headline inflation is around 3%, still above the 2% target. Year-ahead inflation expectations recently jumped to 4.7%. Market-implied odds of a December rate cut have collapsed from about 98% a month ago to roughly 35%.

Minutes from the latest Fed meeting showed wide disagreements among policymakers. Some want cuts to support employment. Others worry that moving too quickly would lock in higher inflation. Chair Jerome Powell is trying to hold that coalition together while both households and markets are screaming for relief.

As long as that debate remains unresolved, both sides of the leverage problem stay exposed. Consumers get no respite from 22% credit card rates, and investors running large margin balances can't count on the Fed riding to the rescue if markets stumble.

What Thursday Actually Meant

Nvidia's earnings didn't get less impressive over the course of a single trading day. What changed was how a highly leveraged market interpreted them.

Wall Street has taken on $1.18 trillion of margin debt to chase AI and growth stocks, with borrowing growing faster than underlying returns. Main Street is carrying $1.233 trillion in credit card balances at record rates, with sentiment near historic lows and delinquencies climbing. The Fed sits between them, afraid to cut too soon and afraid to hold too long.

Thursday's move was more than routine profit-taking. The jump in the VIX signaled that volatility markets are starting to price how sensitive this setup is to small shocks, with no guarantee of quick policy support.

The real leverage crisis isn't on Wall Street. It's on Main Street. Consumer spending drives corporate earnings, not the other way around. When the people who are supposed to buy AI-powered products and services are already tapped out, the revenue assumptions behind $300 billion in annual tech capex start looking shaky.

That's what Thursday told us. Not that AI is overvalued, but that the entire system supporting AI valuations is built on borrowed money, borrowed time, and borrowers who are running out of both. In an economy where investors and households are both heavily exposed to debt, and where the central bank isn't sure what to do about it, great earnings sometimes aren't enough.