Panic selling on DeepSeek news crunches Nasdaq, but insiders say frightenly 'excellent advancement' is being misread -- and it's hardly new news

DeepSeek made a disruptive AI leap eliciting deep fears on Wall Street and the Nasdaq, sending the latter down 2.9%, and hitting shares of AI chipmaker Nvidia particularly hard, dropping 17% at one point.

But analysts say don't panic – and maybe even smile; AI possibly just got much cheaper and better, though it may take time to restore investor faith in chipmakers. – particularly Nvidia (NVDA).

The stock closed at $118.42, down $24.20, or 16.97%, but edged back after hours, gaining $1.79, or 1.51%, at $121.21. 

Nvidia has a near-monopoly on GPU chips that do parallel processing – in effect doing multiple massive computer-cognitive tasks concurrently, the soul of AI models.

DeepSeek apparently combines algorithms and cheaper infrastructure without reliance on pricey GPUs,(Graphics Processing Units), dramatically lowering the price of its apps. 

The market took a long time to awaken to the threat, but "this news is not brand new,” says John Belton, managing director for growth portfolios for Gabelli Funds in prepared remarks.

Strength or weakness?

“DeepSeek published a big paper over a month ago detailing these breakthroughs. They actually released a model with a lot of the methodology a year ago,” he said.  

“Industry execs and insiders have known about DeepSeek for a while. And last week we got some of the most bullish capital expenditure outlooks yet - strange? Therefore, are these investments from US big tech coming from a position of strength or weakness?”

Still, Belton says Nvidia itself almost certainly is bruised by the news.

“Nearer-term creates a new bear case for NVDA. Outstanding questions are: Will DeepSeek’s methods fuel even greater innovation and therefore computing demand from Western customers who have access to even better chips than DeepSeek likely had?”

New ballgame

Dan Ives, senior analyst at Wedbush Securities, brings it down more to the level of how (for decades) nobody ever got fired for choosing IBM – but substituting Nvidia.

“No US Global 2000 [company] is going to use a Chinese start-up, DeepSeek, to launch their AI infrastructure and use cases,” he writes. 

“At the end of the day, there is only one chip company in the world launching autonomous, robotics, and broader AI use cases and that is Nvidia. 

"Launching a competitive LLM model for consumer use cases is one thing. Launching broader AI infrastructure is a whole other ballgame and nothing with DeepSeek makes us believe anything different.”

Mark Klein, CEO of SuRo Capital, sees a case that even Nvidia will rise on the DeepSeek tide.

“Potential Impact on Demand: While there is a contention that DeepSeek's efficient training methods could reduce the demand for high-end Nvidia GPUs, potentially affecting sales, it is also plausible that the more cost-effective approach will result in more demand for hardware for those looking to train proprietary models,” he says in prepared remarks. 

“Additionally, it is not known how this new approach will scale with larger GPU clusters.”

Definite positive

The news of the Chinese startup's advancement in artificial intelligence is likely a net positive for most companies – not least Google, Amazon, and Meta -- because AI possibly just got much cheaper, according to Belton.

“Should be positive big tech broadly with nuances. Clouds could see demand expand as inference costs decline," he said.

It's a “definite positive for AI users like Meta, Google, and Amazon  - lower costs of inferencing. Other AI users across software, healthcare, the internet, and financials also should see compression in compute (inferencing) costs," he added. 

Even Nvidia evinced enthusiasm over the Chinese startup DeepSeek's news

“DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling,” the chipmaker said in a statement. 

"DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely available models and compute that is fully export control compliant. 

"Inference requires significant numbers of NVIDIA GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.”

‘Sputnik moment’

For now, Wall Street is unconvinced. Meta Platforms, Inc. (Facebook), Alphabet Inc. (Google), and Amazon shares all tanked on the day, bringing the Nasdaq down with them. 

For now, it's a shock factor, amplified by Marc Andreessen, co-founder of Silicon Valley venture capital firm Andreessen Horowitz, calling it a “Sputnik moment” on Twitter. 

In 1957, the Soviet Union was the first nation in history to place a rocket-propelled artificial satellite into orbit around the Earth.

Certainly, the investors are not reacting to a nonevent, says Giuseppe Sette, president of Reflexivity.

“DeepSeek has taken the market by storm by doing more with less,” says in a prepared remark. "In layman's terms, they activate only the most relevant portions of their model for each query, and that saves money and computation power. 

"This shows that with AI the surprises will keep on coming in the next few years. And even though that might be a bit of a shocker today, it's extremely bullish in the long-term - because it opens the way for deeper and broader adoption of AI at all scales.

“But where does that leave the US? 

Barriers lowered

DeepSeek's open source approach will be carefully scrutinized by OpenAI and its many contenders for the ultimate prize: AGI [Artificial General Intelligence]."

 AGI is a type of AI that matches or surpasses human cognitive capabilities across a wide range.

"With lower barriers to entry, this ultimate goal is tantalizingly close," he says. 

'Yet, just because China's DeepSeek achieved this advance does not mean the next ones won't happen on U.S. soil.

Thanks to its rich talent and capital base, the US remains the most promising home turf from which we expect to see the emergence of the first self-improving AI."

War rooms

Nancy Tengler, CEO and chief investment officer at Laffer Tengler Investments, agrees that Silicon Valley will react – and, in fact, has been for a while.

“These companies have deep pockets and talent,” she said in prepared remarks. “You heard no doubt that Meta has set up war rooms to analyze DeepSeek. 

"We will soon find out if all the claims are true, and it is my expectation that we will also learn how quickly U.S. firms can adapt.”

Those “war rooms” aren't for nothing, says Belton.

“What DeepSeek has achieved is ‘real’ to an extent. There are a few clear breakthroughs which show the industry a path to cheaper means of both training and inference.”

As for Sputnik, analysts pointed out the United States won the space race.