Nvidia Is ‘Tip Of The Spear’ For AI Trade, Says Dan Niles — Warns ‘At A Certain Point Either The Bond Market’s Wrong Or Stock Market Is Wrong’

Dan Niles, founder and portfolio manager at Niles Investment Management, said on Monday that he’s growing more cautious on tech stocks as bond yields climb, but remains bullish on Nvidia (NVDA), which he sees as a standout beneficiary of the artificial intelligence boom.

In an interview with CNBC, Niles said that investors are facing a growing disconnect between the bond and stock markets, particularly as higher yields put pressure on equity valuations.

“At a certain point either the bond market’s wrong or the stock market is wrong,” he stated, adding that he doesn’t hold a broad bullish stance on technology stocks. Instead, the technology sector specialist said he is being selective and holding more cash while looking for companies where growth and valuation can justify the risk.

“Cash is one of my favorite positions right now because of bond yields spiking,” he said. “For me, I’m trying to be very selective even though I do like some names underneath this.” Nvidia remains one of those names.

Niles Calls Nvidia ‘The Tip Of The Spear’

Niles said he likes Nvidia “quite a lot” heading into its next earnings report, pointing to both its valuation and growth outlook. He also sees the chipmaker positioned to gain additional share among hyperscale customers as power availability becomes a more important constraint on data-center expansion.

“If we’ve got all these data center moratoria and pushback on power, we have the best token per watt out there,” Niles said, referring to Nvidia’s efficiency advantage in AI workloads.

He also sees Nvidia’s valuation as reasonable relative to the growth he expects from the company. “You look at it and say, ‘Wow, it’s trading at a 16x multiple on 70% revenue growth for calendar 27, which is a market multiple,’” Niles said. “Nvidia is a great valuation for terrific growth.”

He called the stock “the tip of the spear for the AI trade and for the market in general.”

NVDA stock gained around 1% in midday trade on Monday, its fourth consecutive session in the green. On Stocktwits, retail sentiment around the Jensen Huang-led AI bellwether shifted to ‘bearish’ from ‘neutral’ territory.

In a note to investors cited by Thefly, BNP Paribas on Monday raised its price target on Nvidia to $345 from $285 and kept an ‘Outperform’ rating on the shares. The shares hit a record high of over $237 on Friday. BNP Paribas’ price target implies an upside of over 45% from that peak.

Niles Flags Falling AI Prices As A Bigger Risk

According to Niles, the broader risk emerging within the AI trade is the falling prices for AI inference and token usage, which could eventually undermine the revenue growth investors are expecting.

“You may generate a lot more tokens, but if the pricing drops a lot because you have an alternative that’s 90% cheaper, ultimately that’s a problem,” he said.

Niles pointed to rising token volumes alongside declining token spending among major AI providers as a trend worth watching. If that pressure persists, he said investors may need to temper their enthusiasm because ultimately “profits matter.”

Niles remains constructive on Meta Platforms (META), pointing to the company’s efforts to monetize its heavy AI spending. According to him, public cloud compute will be the next phase for the company.

He is also bullish on Alphabet (GOOGL), stating that Google’s latest AI model has reached the frontier and pushing back on concerns that the company has fallen behind its rivals.

GOOGL stock edged 0.6% higher in midday trade on Monday, while META stock jumped as much as 2%. On Stocktwits, retail sentiment around Alphabet trended in ‘bullish’ territory over the past day, while sentiment around Meta remained in the ‘neutral’ zone.

Longer term, Niles said he expects Alphabet and Apple (AAPL) to benefit from controlling the platforms consumers use to access AI agents. He forecast that users will increasingly interact with assistants through their phones without necessarily caring which underlying AI model powers them, potentially making the device and operating-system layer more important than any individual AI application.