The AI Boom Runs on a Money Loop. Here Is What That Means for Your Portfolio After the Worst Week Since April

Stocks just had their roughest stretch in months. For the week that ended Friday, November 6, 2026, the S&P 500 lost about 2.7 percent, which several market recaps called its worst week since April 2025. The Nasdaq Composite fell roughly 3.4 percent, its fourth straight weekly loss, and the Dow held up better with a smaller slip. Chip stocks took the hardest hit, with the widely watched semiconductor group dropping about 12 percent on the week.

Here is the strange part. Earnings were mostly fine. The economy is still growing. So what unsettled people? Two related worries about how the artificial intelligence buildout is being paid for. Neither is brand new. One was sharpened by investor Michael Burry back in November 2025, and the other has been building all year as the dollar figures climbed. A more immediate jolt came in October, when OpenAI told investors that its revenue was lower than had been widely reported. If you own an index fund, a target date fund in your 401(k), or just a few big tech names, these arguments now touch your money whether you follow them or not. Let me walk you through both.

Worry one: the money goes in a circle

The first concern has a nickname that sounds like a board game: circular financing. The basic idea is that a small group of companies are funding each other, so the same dollars can show up more than once as both an investment and as sales.

A simplified version looks like this. Nvidia, which makes the chips, takes a stake in OpenAI, which builds the AI models. OpenAI uses money to rent computing power from cloud providers like Oracle. To deliver that computing power, Oracle buys more chips from Nvidia. Cash leaves Nvidia as an investment and can come back later as revenue, having passed through the other companies along the way. Analysts at Yahoo Finance and elsewhere have estimated the interlocking commitments between Nvidia, OpenAI, and Oracle alone at roughly $745 billion, with broader tallies across the industry running far higher.

Why does that make investors nervous? Because a lot of it leans on one company that is still losing money. OpenAI has reportedly lined up well over a trillion dollars in long term computing commitments, while generating about $50 billion in annualized revenue as of late September 2026. That $50 billion is a net figure OpenAI disclosed on October 8, 2026, and it came in below an earlier, higher number that had counted partners' gross sales. The arrangement depends on OpenAI continuing to raise fresh capital faster than its losses pile up. Pablo Hernández de Cos, General Manager of the Bank for International Settlements, described these arrangements in a September 2026 speech as difficult to observe, because the cross shareholdings and purchase promises link chipmakers, cloud providers, and AI firms in tangled ways.

There is a serious rebuttal, and it deserves equal airtime. Nvidia's chief executive, Jensen Huang, pushed back at a September conference, saying the business is not really circular because the company puts in a little money and a lot comes back. His point is that real customers are paying real bills, and Nvidia still generates billions of dollars in free cash flow each quarter, which is actual cash, not an accounting trick. Supporters also note that vendor financing, where a supplier helps fund its own customers, is an old and sometimes perfectly healthy practice. The honest answer is that nobody knows yet whether this reflects durable demand or a loop that flatters everyone's numbers until it slows.

Worry two: how fast do the chips wear out?

The second argument is quieter and more technical, but it goes straight to reported profits. It was pushed into the spotlight in November 2025 by Michael Burry, the investor made famous by The Big Short, and it is about a single accounting choice called useful life.

When a company buys expensive equipment, it does not count the whole cost as an expense in year one. It spreads that cost over the number of years it expects the gear to stay useful, a process called depreciation. Stretch the assumed life longer, and the yearly expense gets smaller, which makes reported profit look bigger. Here is the catch Burry raised: the big cloud companies have generally assumed their AI servers will last five to six years, while he argues the real economic life of cutting edge chips is closer to two or three years because newer, faster models keep arriving.

Burry estimated in late 2025 that extending those useful lives could understate depreciation by about $176 billion across the industry from 2026 through 2028, which would inflate reported earnings along the way. By his math at the time, that could overstate Oracle's 2028 profit by roughly 27 percent and Meta's by about 21 percent. He called the practice one of the more common accounting problems of the modern era. To learn how depreciation and other items flow through the numbers, our guide on how to read company financial statements is a good place to start.

This claim is hotly disputed, and that matters. As CNBC noted, it is genuinely hard to prove, because accounting rules give companies real discretion in estimating useful lives. Plenty of respected voices argue the opposite case: that older data center hardware has historically stayed in service for years rather than being tossed out, and that the longer life assumptions are reasonable. This is a debate about judgment, not a confirmed error, and it is worth holding both ideas at once. It is also a background concern that predates this particular week, not a freshly discovered problem.

Why the spending number is so big

Both worries sit on top of a simple fact: the sums involved are enormous and still climbing. For 2026, Amazon has guided to roughly $220 billion in capital spending, Alphabet to $195 billion to $205 billion, Meta to $135 billion to $145 billion after it narrowed its range on its October 21 earnings release, and Microsoft to about $170 billion. That adds up to somewhere around $720 billion to $740 billion, but the companies do not all count spending the same way. Amazon emphasizes cash capital expenditure, Meta includes principal payments on finance leases, and Microsoft counts finance leases when they begin, so the total blends slightly different measures. Even allowing for that, it is a jump of roughly 78 percent from about $410 billion in 2025. When spending grows that fast, the market starts asking the uncomfortable question Burry keeps repeating: when does it end, and what happens to profits when it does?

What this means for a regular investor

You do not need a view on OpenAI's balance sheet to be affected by this. A handful of these companies now make up an unusually large slice of the major U.S. indexes, so when they wobble, broad funds wobble too. That is the heart of concentration risk, and it is the practical takeaway here. If your portfolio has quietly become a bet on a few AI names, that happened because they grew, not because you chose it.

A few steady habits tend to help in weeks like this one. Know what you actually own by looking at your funds' top holdings. Decide in advance how much of your money you are comfortable having tied to one theme. And remember that selling into a scary headline often does more damage than the headline itself, because timing the exit and the re entry are two separate guesses you have to get right.

The calendar will keep the pressure on. The October inflation report is due Tuesday, November 10, earnings from Cisco and Applied Materials land after the close on Thursday, November 12, and Nvidia itself reports on November 17, with the Federal Reserve's next rate decision set for December 9. None of these will settle the big debate on their own. But together they will give investors their next read on whether the AI money loop is spinning on real demand or just spinning. For now, the smartest move for most people is the boring one: understand what you own, size your risk on purpose, and let the giants argue it out.

This article is general information, not personalized investment advice. Figures reflect reporting and company guidance as of early November 2026 and will change.