Nearly half of the S&P 500 now rides on AI. The last two times one theme got this big, the S&P 500 fell 49% and 25%.

The S&P 500 holds about 500 companies, but not equally. The biggest ones carry the most weight, and the biggest ones today are almost all AI companies. This week showed what that means: the average stock in the index went up, and the index itself still fell 2%.

The S&P 500 weights each company by its size. When a handful of companies grow far faster than the rest, they take over the index. As of October 16, the ten biggest positions in the largest S&P 500 fund were all AI-linked companies, and together they were 38.5% of it.

Add the chipmakers, the data-centre builders and the AI software firms further down the list, and about half of the index now rides on one theme. Goldman Sachs puts the AI-linked share at about 45%. Our own count of the holdings comes to about 51%. Anyone who owns AI stocks, an index fund, or both is holding the same trade.

The Nasdaq ended a four-week winning streak with its worst week since June. Chip stocks fell almost 10%. AI software lost about $260 billion in value. CoreWeave, a company that rents out Nvidia chips, lost about a quarter of its value in five days, and its bonds sold off until they yielded more than 15%.

Meanwhile utilities, energy and banks rose, and the typical S&P 500 company ended the week slightly up. If the index held every company equally, it would have ended the week up too. Because it holds the AI giants most, it lost 2% in a week, ten days after setting a record.

2000. The internet was real, and it changed everything. Tech had grown to about a third of the S&P 500 when the market peaked in March 2000. Over the next two and a half years the Nasdaq fell 78% and the S&P 500 fell 49%.

2022. After two years of the biggest tech stocks carrying the market, the top ten were about 30% of the index. When rates rose, the S&P 500 fell 25% and the Nasdaq-100 fell 33%.

AI companies are earning real money, and this is not a prediction that AI is a bubble. It is a reminder that when one trade is half of the market, a bad week for that trade is a bad week for the market, and a bad year is a bad year.

Every big fall started out looking like a dip. In March 2000 and January 2022, the first weeks looked a lot like this one: a record, a sharp drop in the leaders, a bounce that brought buyers back. Most dips recover within weeks. Some don't. From the inside, the first days of both look the same.

So the question that matters is not whether AI is a bubble. It is whether the market has already turned Risk OFF, and what to do with your portfolio if it has.

So, are we already in a Risk OFF market? That is the one question most investors only answer in hindsight, after the damage is done.