News reports that Chinese artificial intelligence lab DeepSeek has created an AI model that rivals U.S. models but at a much lower cost sent shockwaves through AI chip stocks on Monday. Nvidia stock and others plunged.
DeepSeek developed its AI system without cutting-edge foreign processors, which China is restricted from obtaining due to U.S. export curbs. The Chinese firm says it built its system for less than $6 million, vs. the billions being spent by rivals. It also is much cheaper to operate than those of U.S. rivals like OpenAI and Meta Platforms.
The news comes after U.S. venture Stargate announced plans to spend as much as $500 billion on AI infrastructure over the next four years.
Silicon Valley venture capitalist Marc Andreessen described DeepSeek-R1 as "AI's Sputnik moment," a reference to when the Soviet Union launched the first artificial Earth satellite in 1957, kicking off the Space Race.
A Wired article said DeepSeek focused on "maximizing software-driven resource optimization," since it didn't have access to the latest AI processors from Nvidia and others.
DeepSeek came up with more efficient methods to train its models, such as using custom communication schemes between chips, reducing the size of fields to save memory, and an innovative use of the mix-of-models approach, according to the Wired article.
On the stock market today, Nvidia stock dropped 17% to close at 118.42. Among AI chip peers, Broadcom fell 17.4% to 202.13 and Marvell Technology sank 19.1% to 100.33. Advanced Micro Devices retreated 6.4% to 115.01.
Other AI-related chip stocks also fell on Monday. Chip foundry Taiwan Semiconductor Manufacturing slid 13.3% to 192.31. Hot IPO stock Astera Labs plunged 28% to 83.16.
Possible Pause In AI Infrastructure Spending
The DeepSeek news could lead to a pause in AI infrastructure spending by hyperscale cloud service providers as they rethink their capital expenditure plans, Evercore ISI analyst Amit Daryanani said in a client note Monday.
"We think some of the risk near term will be around 'Do we need to spend as much on capex given a cheaper alternative?,'" Daryanani said. "It's likely that over time this will end up being a scenario where cloud/enterprises can do even more AI centric work and do it beyond text to images to video."
Jordan Klein, a trading desk analyst with Mizuho Securities, said the DeepSeek news created a "Sell first and ask questions later" type reaction among investors. On the other hand, he would avoid buying AI stocks on the pullback.
"When AI trade and narrative is this crowded and up this much, the damage could take a while to fully play out," Klein said in a client note Monday.
Also, there was misinformation about the cost to develop DeepSeek's AI model and what it actually accomplished, he said.
DeepSeek Panic 'Seems Overblown'
Bernstein analyst Stacy Rasgon said the DeepSeek news does not spell doomsday for AI buildouts.
"The resulting Twitterverse panic over the weekend seems overblown," Rasgon said in a client note Monday. "According to the many (occasionally hysterical) hot takes we saw, the implications range anywhere from 'That's really interesting' to 'This is the death-knell of the AI infrastructure complex as we know it.'"
Meanwhile, demand for AI chips and data centers is still increasing, Rasgon said.
"Investments are still accelerating," he said. "Right on top of all the DeepSeek news flow last week we got Meta substantially increasing their capex for the year. We got the Stargate announcement. And China announced a trillion yuan (about $140 billion) AI spending plan. We are still going to need, and get, a lot of chips."
Rasgon reiterated his buy ratings on Nvidia stock and Broadcom stock.
Could DeepSeek News Be 'China Psyop'?
DeepSeek's impact on the AI technology supply chain "all depends on the veracity of DeepSeek's claims," Truist Securities analyst William Stein said in a client note Monday.
DeepSeek alleges that it built its AI model with "inferior chips and at a fraction of the cost" of U.S. models, Stein said. "We cannot determine the veracity of DeepSeek's claims. If they're true, it magnifies the already-known risk of an AI spending slowdown."
DeepSeek claims it used 2,048 Nvidia H800 chips, a downgraded version of Nvidia's H100 chips designed to comply with U.S. export restrictions. However, some industry officials are skeptical of that claim, saying DeepSeek has more Nvidia processors and more advanced versions than it is letting on.
Some on Wall Street are wondering if the surprising developments from DeepSeek are a "China psyop."
Nvidia Stock, Other AI Chipmakers On Tech Leaders List
Wedbush Securities analyst Daniel Ives maintained his buy rating for Nvidia stock on Monday.
"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," Ives said in a client note. "Launching a competitive LLM (large language 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."
Meanwhile, the DeepSeek news is positive for semiconductor companies that make chips for network edge AI applications, which require less computing power than data center processors, Rosenblatt Securities analyst Kevin Cassidy said in a note Monday.
"This helps proliferate AI applications at the network edge and is positive for our names under coverage including Ambarella, Ceva, Qualcomm, and Synaptics for their network edge AI processors," Cassidy said.
Nvidia stock is on the IBD Tech Leaders list, along with Astera Labs, Broadcom, Marvell and Taiwan Semiconductor.
Follow Patrick Seitz on X, formerly Twitter, at @IBD_PSeitz for more stories on consumer technology, software and semiconductor stocks.
