AI bubble bursts: Buy the Nasdaq meltdown?

The case for buying the dip For long-term investors, the correction could offer an opportunity to buy companies whose share prices have fallen faster than their underlying earnings prospects. “US companies have posted strong earnings growth in Q3,” says Lakshmi Iyer, Group President–Investments and MD & CEO of Bajaj Alternate Investment Management. “There could be near-term volatile market moves. It would give investors an opportunity to cherry-pick stocks.”

The argument is for selectivity, not indiscriminate buying. Investors need to separate businesses with durable earnings and competitive advantages from those whose valuations depended largely on cheap capital and enthusiasm for the AI theme. Yet even the strongest companies face a growing test. Amazon, Alphabet, Microsoft and Meta are together expected to spend as much as $828 billion on capital expenditure this year, much of it on AI infrastructure.

Their strong cash flows give them considerable capacity to fund the investment. But the scale of spending raises a more fundamental question: how quickly will the infrastructure translate into revenue and returns? The bubble is in the investment Ritesh Jain, founder of Pinetree Macro, sees the AI trade entering a more protracted and damaging phase: the unwinding of a bubble. “We do have some world-class companies stuck in the whirlwind of the AI bubble and its apparent burst,” he says.

“When money was freely available, people went bullish on the theme. Now the major red flag is hyperscaler capex. Markets are questioning these heavy capex numbers, sometimes alongside negative cash flow, and asking whether the business has a moat.” Jain does not expect a single dramatic crash. He sees a longer period of repricing, as investors scrutinise the funding and economics of companies built around the AI boom.

“Equity markets started borrowing debt, and fixed-income spreads started blowing out.” A central flaw, in his view, is the assumption that a vast total addressable market, or TAM, automatically makes every company competing for it a good investment. “When the TAM is higher and everybody has a space. But when the size of companies becomes larger than the TAM, the bubble forms. And now, everyone is fighting with everyone else.”

The risk is that an enormous potential market attracts so much capital and competition that the returns on investment fall short of expectations. “Chips have a four-to-five-year life,” Jain says. “Then you look at the capex numbers and high valuations on such depreciating assets. This is the worst kind of bubble — the biggest misallocation in history.” He also sees a possible shift in where investors look for the next technology opportunity as US-China competition intensifies.

“An area where the US could get aggressive now is space,” he says. “Across the AI narrative, even China is giving very tough competition. One could say SpaceX could be a beneficiary as it is in an area less explored right now, where the TAM is yet to build.” A historic misallocation of capital? Devina Mehra, founder, chairperson and managing director of First Global, is more emphatic. She believes AI could eventually be remembered as a classic financial bubble — not because the technology lacks promise, but because the investment required may never generate commensurate economic returns.

“When the next book on financial bubbles is written, there will be a chapter on AI,” she says. The scale of the buildout lends weight to the concern. Research presented at a Brookings Institution conference estimates that AI infrastructure will require investment equivalent to about 3.6% of annual US GDP through 2032 — more than $10 trillion. That compares with 2.2% of GDP absorbed by railroads in the late 19th century and roughly 1% a year during the telecom and internet expansion.

Mehra also flags the cost and terms of financing. “Anthropic is already borrowing dollars at 10%,” she says. “And look at the SpaceX IPO. You are giving away corporate governance and voting rights while getting nothing in return. These are all signs of excess.” Anthropic reported a net loss of $80.7 billion in the first half of 2026, although $72.4 billion of that was an accounting charge linked to convertible instruments rather than an operating cash loss.

Not all AI stocks will be caught in the same selloff For investors, the distinction between the companies building the AI infrastructure, those deploying it and those simply riding the theme is critical. Leading hyperscalers and dominant chipmakers may have the balance sheets and competitive positions to weather a valuation reset. Smaller companies with weak business models, heavy funding needs or limited pricing power face a tougher test.

“The biggest falls could be in companies that should not have existed and oversold themselves in this AI bubble narrative,” Jain says. “Only a few have business models that will survive this fall.” That does not rule out opportunities among the survivors. But investors may have to wait for the earnings to catch up with valuations. “Microsoft stocks did not go anywhere from 2004 to 2016, but earnings doubled,” Jain says.

“Some stocks are already 50-60% down. It is giving that 2000s kind of feeling.” The question confronting investors now is now whether to buy the current dip but what to buy in the current dip — and how long investors one is prepared to wait for the returns.