AI bubble: how can you tell if we are at the top?

The analysis: nobody can date a top

On October 7, 2026, in Singapore, Ray Dalio called AI a classic bubble, close to its breaking point. Yet the Bridgewater founder has been warning since June, and the market kept rising. He admits it himself: bubbles don't burst on a fixed schedule.

The beginner's trap is hunting for the day of the top. Nobody can do that. What you can do is spot the conditions that make a burst likely.

For Dalio, technology doesn't decide the timing. Liquidity does. A bubble bursts when investors have to sell to raise cash (debt, taxes, withdrawals). And when everyone sells at once, prices collapse.

Hence the idea of a checklist. Not a crystal ball, but a dashboard.

The historical fact: August 5, 2024, when the carry trade broke

The yen carry trade means borrowing in yen almost for free to buy higher-yielding assets such as US stocks.

On July 31, 2024, the Bank of Japan raised rates. The yen jumped, and those who had borrowed in yen had to sell. On August 5, 2024, the Nikkei lost 12.4% in a single session, and the Nasdaq dropped right after.

It wasn't a one-off. In March 2020, the basis trade (hedge funds buying Treasuries and selling the futures, with huge leverage) had already triggered emergency selling.

The lesson never changes: leverage turns a tremor into an avalanche.

The anecdote: Howard Marks's three ideas

Howard Marks, co-founder of Oaktree Capital, is famous for his "bubble.com" memo of January 2000, sent just before the dot-com crash. In his book The Most Important Thing, three ideas fit the AI bubble perfectly.

"What the wise man does in the beginning, the fool does in the end."

So be early: cut risk while everyone is euphoric.

"Being too far ahead of your time is indistinguishable from being wrong."

But not too early: selling at the first warning sign can mean missing years of gains.

"Never forget the six-foot-tall man who drowned crossing the stream that was five feet deep on average."

Always keep a margin of error.

The checklist: six boxes to watch

• The US 10-year yield nears 6%. Higher rates lower what investors will pay for a stock.
• The Fed hikes again on October 28. A second hike would confirm that easy money is over.
• The yen surges. That's what breaks the carry trade, as in August 2024.
• The Treasury market seizes up. That's what triggers forced selling in the basis trade.
• Oracle's bonds slide. With that much debt, it's the canary in the coal mine.
• A budget standoff after the midterms. More deficit means more Treasuries to issue, so even higher rates.

The rule is simple: if one box lights up, it's noise. If several light up at once, watch out. These risks feed each other: rates rise, bonds fall, leveraged players sell, AI stocks crack, and the Nasdaq drags down the S&P 500.

There is a safety net: in 2020 and again in 2023, the Fed stepped in. A scenario is never a certainty.

This topic is also covered on video in the "AI Bubble" series on the Axone-Capital YouTube channel.

The lesson for your portfolio

Lesson one: don't hunt for the date, watch the conditions.

Lesson two: a great company isn't necessarily a great buy at the wrong price. AI can change the world and its stocks can still fall.

Lesson three: a bubble is a market risk. If it bursts, it takes the whole index with it. Spreading your money across several tech stocks won't protect you.

Meanwhile, keep your margin of error. This content is educational and is not investment advice.

Frequently asked questions

Can you predict when the AI bubble will burst?

No. Even Ray Dalio says bubbles don't burst on a fixed schedule. What you can do is track the signals that make a burst more likely, such as rising long-term rates.

Why do interest rates matter so much for AI stocks?

Because part of the AI build-out is funded with debt. Higher rates also shrink the multiple investors pay for earnings: with a 10-year at 6%, a Fed model suggests 16 times earnings instead of 19 to 20.

What does "being too early is the same as being wrong" mean?

It's a Howard Marks idea: selling far too soon ends up looking like a mistake, because you miss the rally. Be ahead of the crowd, but not too far.