AI is still growing. Investors now want proof it pays.

AI isn’t slowing down. Big tech has spent about $497 billion on data centres and chips in just the first nine months of 2026, and plans to spend more in 2027. Nvidia earned over $120 billion last fiscal year, and Nasdaq-100 earnings are up 19% from a year ago.

And yet, last week, the AI trade cracked.

Markets aren’t questioning whether AI is real. They’re questioning whether the huge sums being poured into it will earn enough to justify the prices investors have paid. As one US fund manager put it: “The market is moving from expectations about AI to the price of AI.”

1. What Cracked

On 5 November, US chip stocks fell 12.2%, their worst day since the Covid crash of March 2020. The Nasdaq fell 5.5% and entered a correction. Nvidia alone lost about $448 billion in value in a single day.

By the next morning, the selling had spread across Asia. Korea’s Kospi and Japan’s Nikkei each fell 6%, and Korean trading was paused soon after the open. Both markets lean heavily on chipmakers. China was the exception. Its AI rally had already cracked in October, and Hong Kong tech stocks still ended the week up 8.6%.

The trigger was political. After the Democrats won both chambers of Congress in the 3 November midterms, investors began to fear tighter rules on data centres and the electricity they use.

But the election was the spark, not the fuel. US chip stocks, Japan’s Nikkei and China’s CSI 300 all peaked on the same day, 22 June, and have been losing air ever since. Doubts about AI returns had been building for months. Washington simply gave investors a reason to act on them.

2. The $500 Billion Question: Why Chipmakers Took the Hardest Hit

This is the gap at the centre of the story. In the first nine months of 2026, Amazon, Alphabet, Microsoft and Meta spent about $497 billion on data centres and chips, more than in all of 2025. Yet Microsoft’s AI business earns about $37 billion a year, and Sequoia estimates spending outruns AI revenue by about $600 billion a year.

India has seen this before. Telecom companies spent heavily on spectrum and towers, mobile data took off, yet many never earned back what they spent.

Nvidia sits on the other side of that gap. It doesn’t spend on AI. It sells the chips. Its data centre revenue jumped 117% to $89 billion last quarter, which is proof that demand is real. But every dollar is someone else’s spending, and the big four are now spending more on capex than their operations bring in.

That’s why chipmakers fell hardest:

• Few buyers. If one big customer slows down, orders drop fast. Microsoft has already cut its 2026 spending outlook to about $170 billion.
• Borrowed money. The cost of insuring CoreWeave’s debt against default hit a record on 5 November.
• Biggest run-up. Stocks that rise fastest fall fastest.

On 5 November, Microsoft fell 3.6% while AMD fell 17.3%. The market isn’t punishing the use of AI. It’s repricing the companies that depend on AI spending continuing whether or not it pays off.

3. Why India Rallied While AI Markets Fell

Here’s the twist. In the same week, the Nifty rose 8.58% and the Sensex 8.82%, their best week since February 2021. Foreign investors bought over ₹35,000 crore of Indian shares in five days.

India rallied because it sits outside the AI spending story. With few AI chip or data-centre stocks, it drew money leaving Korea, Japan and Taiwan. A 16% drop in Brent crude and solid results from Reliance, Bharti Airtel and M&M helped too.

But India isn’t fully insulated. IT was the weakest sector, up just 1.14%. If US companies squeeze tech budgets to fund AI, Indian IT could feel it next.

The rally is also fragile. It was led mostly by banks, and domestic institutions were selling while foreigners bought. Money that arrives fast can leave fast.

4. Bubble or Correction?

There are worrying signs, but today’s AI leaders are very different from the dot-com companies of 2000.

On price and profits, today looks healthier than 2000. On concentration, it looks worse.

The case for a bubble: 57% of big investors in a Deutsche Bank survey see an AI valuation crash as the top market risk. OpenAI is valued at about $852 billion, double Coca-Cola, without ever making a profit.

The case for a correction: The leaders are earning real money, and big tech mostly fills its own data centres, so there’s less risk of empty capacity than in the 1999 fibre-cable boom.

As Capital Economics puts it, the share-price bubble may be deflating while the business of AI keeps growing. Whether this becomes a crash depends on one thing: whether AI spending starts producing real returns.

5. What Investors Should Watch

1. Check your hidden AI exposure. Many international funds, US index funds and global tech funds hold the same few AI names. You may own far more AI than you think.
2. Don’t chase the rebound or panic in the fall. Sharp swings are normal when a theme gets this crowded. Decisions made in weeks like this are usually the ones investors regret.
3. Don’t read India’s rally as a guarantee. It was helped by cheaper oil and foreign money seeking a safe spot. Both can change quickly.

What we’re watching next: big tech’s 2027 spending plans and any sign that AI revenue is catching up, new US rules on data centres, oil prices and foreign flows into India, and Indian IT companies’ commentary on US client budgets.

AI is still a real and growing business. The question is no longer whether AI matters, but what price is fair to pay for it.

Is your portfolio too heavy on one theme, or built to handle a correction like this? That depends on what you own, your goals and your time horizon. A newsletter can’t answer that, but a review can.