AI trade entering cooling phase amid two uncomfortable realities; India emerging as key ‘anti-AI’ allocation

The artificial intelligence trade is moving into a cooling phase as a series of uncomfortable realities unsettle conviction in the sector, said veteran strategist Ed Yardeni, President of Yardeni Research. In a recent conversation with CNBC TV18, he said that the market is now confronting two major sources of uncertainty that are removing “excitement and froth” from AI stocks, and that may keep sentiment subdued for a while.

Opaque GPU accounting, intensifying competition among LLMs

Yardeni flagged what he called a major “known unknown”: the opaque accounting around GPU depreciation at companies buying Nvidia’s hardware. The investment community still does not know whether these chips are being depreciated over 1-3 years -- as their rapid obsolescence cycles might imply -- or overextended schedules of 5-6 years. This ambiguity distorts earnings quality and clouds valuations, especially as the competitive race accelerates.

Further, that competitive pressure has intensified sharply after the launch of Google’s Gemini 3, which has topped performance charts previously dominated by OpenAI’s ChatGPT. The assumption that the OpenAI ecosystem would be the runaway winner has been jolted, forcing peers into longer product cycles and heavier capital expenditure, with returns becoming harder to model.

Yardeni said that while he uses multiple AI tools, including Gemini 3, he remains convinced that “the companies that sell the large language models are not going to be the ones who make the money. It’ll be the cloud providers,” which are expanding capacity fastest to meet surging demand.

• Also read | Nvidia shares whipsaw from post-earnings surge to sharp reversal as AI-bubble, fine-print worries weigh

High profile investors’ flag AI bubble

The shift in tone comes against a broader backdrop in which several high-profile investors have exited Nvidia -- among them Peter Thiel’s hedge fund and SoftBank’s Masayoshi Son -- adding to anxiety around frothy valuations.

Investor Michael Burry, who has publicly bet against Nvidia, has accused cloud providers of artificially boosting earnings by extending depreciation schedules for AI hardware at a time when chip models refresh annually and risk rapid obsolescence. Economist David Rosenberg has argued that AI-related market pricing implies an eightfold expansion in industry size within five years, a trajectory he describes as a “bubble of epic proportions”. UBS has also cautioned that investors “do need to worry about bubble risks,” even as some asset managers remain underweight megacap tech despite recent earnings beats.

Capital rotation into ‘non-AI’ India stocks

In this context, the possible rotation of capital into non-AI segments is gaining relevance, as pointed out by Mahesh Nandurkar, MD and Head of Research at Jefferies. He said that global value-focused funds -- previously underweight India -- are now reversing course as the stretched valuations of global AI names become harder to justify.

“We are seeing clearly as a trend that a lot of these value-focused funds are making a comeback and looking at adding to India, and funding that by selling some of these expensive names,” he told CNBC TV18, describing India as a “massive anti-AI trade”. Nandurkar added that if the global AI theme were to peak or even plateau, India stands to be a significant relative beneficiary of capital rotation.

Investors seeking growth at reasonable prices continue to find domestic pockets offering more durable visibility than many AI-linked global stocks. The Telecom sector is the cheapest consumer name in India, while the long-term real-estate cycle offers more durable visibility.

---