DeepSeek Challenges AI Giants with Cost-Effective Models

Chinese AI firm DeepSeek unveiled open-source models last week, claiming performance comparable to OpenAI and Meta at a fraction of their cost. The flagship model, DeepSeek R1, has sparked debates on the sustainability of massive AI infrastructure investments, which have reached hundreds of billions of dollars in the U.S.

Bernstein analysts lauded the models’ ability to compete with industry leaders while using significantly fewer resources. However, they dismissed claims that DeepSeek replicated OpenAI’s capabilities for just $5 million, labeling market fears as exaggerated.

“If DeepSeek reduced model costs by 10x, current cost trajectories remain unsustainable due to scaling laws,” Bernstein stated, emphasizing the need for innovations like DeepSeek to optimize existing hardware and sustain AI advancements.

Despite DeepSeek’s achievements, Bernstein reaffirmed that increasing AI infrastructure is crucial to meet growing demand. The brokerage advised against reacting to “doomsday scenarios” on social media and maintained its bullish outlook on Nvidia (NASDAQ:NVDA) and Broadcom (NASDAQ:AVGO), citing strong AI growth prospects.

Nvidia shares, however, fell sharply on Friday and seemed set to extend losses Monday as concerns over reduced AI investment weighed on tech stocks. Wall Street futures also declined, reflecting uncertainty about the future of AI infrastructure spending.

DeepSeek’s breakthrough highlights the industry’s evolving dynamics, but analysts remain confident that demand for advanced computing will sustain investment in AI technologies.