After having time to consider the bombshell news about China's low-cost DeepSeek AI advancements, many Wall Street analysts now believe the impact will be a good thing for AI chipmakers. Meanwhile, Nvidia stock looked for support at a key level on Wednesday.
In a client note Wednesday, BofA Securities analyst Vivek Arya said the DeepSeek achievements are "more evolutionary than revolutionary."
The DeepSeek developments are "noteworthy, but not a big course correction," he said.
While DeepSeek's AI models employed some novel software approaches, they were based on the earlier work of others, including Meta Platforms' Llama models. And since DeepSeek shared its work, other AI system developers can copy those ideas and improve upon them, Arya said.
And the reports that DeepSeek used inferior Nvidia processors due to export controls means that companies with access to Nvidia's latest AI processors will be capable of even greater AI advances, Arya said.
DeepSeek is focused on catching up with Western AI models in generative AI. Meanwhile, cloud computing giants in the U.S. are focused on achieving the next level of AI: "artificial general intelligence" or AGI, he said.
Concerns that hyperscale cloud service providers might pause their spending on AI infrastructure because of DeepSeek are misguided, Arya said. Instead, they're likely to keep their foot on the gas pedal.
Arya reiterated his forecast that the AI accelerator market will double from 2025 levels to $500 billion by 2028 or 2029.
Nvidia Stock Looking For Support
He maintained his buy ratings on AI chip stocks Nvidia, Broadcom and Marvell Technology.
On the stock market today, Nvidia stock hovered near its 200-day moving average line, a major support level. It ended the regular session down 4.1% to 123.70.
Elsewhere on Wall Street, analysts pooh-poohed some of DeepSeek's claims, such as what processors it used and the alleged $6 million price tag for its latest AI model. That's 1% to 3% of the cost typically required to train a leading-edge AI model.
"There is a better chance of me lining up next to Saquon (Barkley) and the Eagles in the Super Bowl than DeepSeek creating this LLM (large language model) with $6 million and no next-generation Nvidia GPUs (graphics processing units)," Wedbush Securities analyst Daniel Ives said in a client note Wednesday.
DeepSeek Is 'Not All It's Cracked Up To Be'
Piper Sandler analyst Harsh Kumar said tech executives are telling him that DeepSeek's capabilities are not as good as advertised.
"In talking to some of our CEOs, they implied that DeepSeek's efficiency and model is not all that it's cracked up to be. They mentioned that it's quite a bit behind OpenAI's model," Kumar said in a client note Wednesday.
Kumar reiterated his overweight, or buy, ratings on Nvidia stock and AMD.
"We believe that the stock response to DeepSeek is overdone at this juncture and continue to believe in AI proliferation leading to improving fundamentals for the compute-related chip companies," he said.
Follow Patrick Seitz on X, formerly Twitter, at @IBD_PSeitz for more stories on consumer technology, software and semiconductor stocks.
