Nvidia Stock Plummets. Why the DeepSeek Chip Selloff Could Be Overblown. -- Barrons.com

Dow Jones
27 Jan

By Adam Clark and George Glover

Nvidia stock was plummeting early Monday. News that a Chinese company has built an artificial-intelligence model competitive with Western rivals for a fraction of the cost was provoking a panic in the chips sector.

Shares in Nvidia plunged 11.5% to $126.05 in early trading. The benchmark S&P 500 index was down 1.7%.

Rival stocks were also tumbling. Marvell Technology fell 14%, Broadcom dropped 11% and Advanced Micro Devices sank 5.5% in early trading. American depositary receipts of Dutch semiconductor equipment maker ASML were down 9.2% and ADRs of Taiwan Semiconductor Manufacturing fell 8.6%.

The big selloff came as Silicon Valley raved about Chinese company DeepSeek's new AI model R1, which appears to perform at a similar level to those developed by rivals such as OpenAI even though it was seemingly developed at a fraction of the cost.

DeepSeek said it spent just $5.6 million training the base model behind its AI. Even though that headline figure excludes research-and-development expenses, it is a striking comparison with the hundreds of millions U.S. companies have spent developing AI models of their own, a figure which was expected to increased to billions with the next generation of the technology.

Marc Andreessen, the Silicon Valley venture capitalist who has been advising President Trump, praised DeepSeek's model and called it a "Sputnik moment" in a post on social-media site X. That refers to the Soviet Union's 1957 launch of the first artificial satellite, named Sputnik, which shook the U.S.'s view of its technological dominance over its geopolitical rival.

That raises questions about whether companies such as Microsoft and Google-parent Alphabet will need or want to continue spending increasingly larger amounts on Nvidia's future chips. It is likely to increase the focus on the potential shift away from training larger models and toward inference -- the process of generating answers or results from the models, where Nvidia's hardware advantage is less marked.

"The DeepSeek news is clearly likely to put pressure on semiconductor stocks today and perhaps in coming days, till there is more clarity on its potential impact on AI chip demand," wrote Jefferies equity sales specialist William Beavington on Monday.

However, there are reasons to think the market could be overreacting.

DeepSeek trained its model on stockpiled Nvidia graphics-processing units (GPUs) rather than Chinese hardware and that's not a feasible strategy forever. Its founder told Beijing's government that U.S. restrictions on chip exports are already an issue, according to The Wall Street Journal.

Some commentators have raised questions over how DeepSeek's model was developed. Alexandr Wang, CEO of Scale AI which provides data for training models, told CNBC in an interview last week that he believed DeepSeek had access to 50,000 H100 Nvidia chips, in breach of U.S. export rules.

"While DeepSeek's achievement could be groundbreaking, we question the notion that its feats were done without the use of advanced GPUs to fine tune it and/or build the underlying LLMs [large language models] the final model is based on," wrote Citi analyst Atif Malik in a research note.

Malik maintained a Buy rating and $175 target price on Nvidia stock.

It's also unclear to what extent Nvidia's customers will be willing to use a Chinese-made model in the face of the Trump administration's backing for the domestic U.S. AI industry.

Last week the White House announced an AI infrastructure plan with Oracle, OpenAI, and SoftBank. The joint venture, called The Stargate Project, will build infrastructure to power the next generation of AI, starting with a data center in Texas. Investments could reach $500 billion over four years. Nvidia was mentioned as a valued partner.

"A...logical implication is that DeepSeek will drive even more urgency among U.S. hyperscalers to leverage their key advantage (access to GPUs) to distance themselves from cheaper alternatives," wrote Raymond James analyst Srini Pajjuri in a research note.

Pajjuri reiterated a Strong Buy rating on Nvidia stock.

Write to Adam Clark at adam.clark@barrons.com and George Glover at george.glover@dowjones.com

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.

 

(END) Dow Jones Newswires

January 27, 2025 09:41 ET (14:41 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10