‘DeepSeek Is a Big Threat,’ Says Top Analyst About Nvidia Stock

Nvidia (NASDAQ:NVDA) shares had a nice rally today, surging nearly 5%. Yet, the stock hasn’t fully recovered from the recent bombshell dropped by China’s DeepSeek – an AI model that matches its American rivals in performance but was developed at a fraction of the cost.

While there has been plenty of pushback on DeepSeek’s claims of lowered costs yet similar performance, D.A. Davidson analyst Gil Luria, who ranks amongst the top 4% of Street stock experts, thinks DeepSeek’s entrance suggests a huge shift is about to take place.

“We continue to believe the introduction of DeepSeek was a game changer and a big leap forward for AI,” Luria said. “Our key observation is that AI has become radically more efficient over the last couple of weeks, and we have found significant support that the reduction in compute is in the 90-97% range.”

If true, this could spell trouble for Nvidia.

“We believe we are approaching peak demand, possibly as early as this year as NVDA’s large customers evaluate these recent developments,” warned the 5-star analyst.

Addressing the skepticism around DeepSeek, Luria tackled claims that the model “isn’t really that much better.” He points out that multiple coders have reported significantly improved inference efficiency – an observation echoed by AI leaders, including Anthropic’s founder. A top AI company even reported a jaw-dropping 93% reduction in compute using DeepSeek.

And what about the theory that DeepSeek’s low $2/million token price is just China undercutting the market at a loss? This seems unlikely, says Luria, given its efficiency. Other open-source models now charge $6/million or less, and OpenAI recently dropped its o3 pricing to $4/million.

Additionally, Luria says the argument that the “$6mn training cost claim is bogus” doesn’t alter the previous points. His research suggests costs were over 10x lower, and DeepSeek’s total spend is estimated at $500 million – still far below OpenAI’s $10 billion.

“We are also not sure how this logic plays out for NVDA – either you believe DS used as much advanced NVDA chips as OpenAI, in which case sanctions will address that, or you believe they didn’t, in which case you don’t really need the most advanced chips and training did cost a small fraction,” argues the analyst.

Lastly, the Jevons Paradox has been brought up a lot in the current argument, its use drawing a caustic remark from Luria that this previously obscure concept is now used in any DeepSeek-related discussion. Luria says that while lower prices typically drive higher demand, this effect has historically taken time to materialize. The analyst believes lower inference prices will help enterprises deploy tools faster, but he doesn’t expect a 20-30x demand increase in the short term to offset the reduced need for compute. The “real gating factor” is the availability of scalable applications, which may take 2-3 years to develop. “More importantly as it relates to NVDA,” Luria added, “if these new applications get deployed on device, the increased demand may not increase the TAM.”

All told, Luria rates NVDA shares a Neutral, and his $135 price target suggests shares will climb 8% higher over the next 12 months, far shy of the past year’s 80% surge. (To watch Luria’s track record, click here)

Luria, however, is amongst a minority on Wall Street; two other analysts join him on the sidelines, yet with an additional 37 Buys, the analyst consensus rates the stock a Strong Buy. Going by the $178.32 average price target, a year from now, shares will be changing hands for a ~43% premium. (See Nvidia stock forecast)

To find good ideas for AI stocks trading at attractive valuations, visit TipRanks’ Best Stocks to Buy, a tool that unites all of TipRanks’ equity insights.

Disclaimer: The opinions expressed in this article are solely those of the featured analyst. The content is intended to be used for informational purposes only. It is very important to do your own analysis before making any investment.