DeepSeek, a Chinese artificial intelligence lab, has unveiled a new AI model that challenges prominent U.S. competitors, sending shockwaves through AI chip stocks. The model was developed at a fraction of the cost, estimated at under $6 million, while rivals invest billions in comparable technologies. This announcement has significantly impacted stocks of major companies like Nvidia, which saw a 15.5% decline, as well as Broadcom and Marvell Technology, both of which recorded over 16% drops.
DeepSeek’s innovative approach involved building its AI without relying on advanced foreign processors, a necessity due to U.S. restrictions on exports to China. The firm has optimized its software-driven processes, utilizing custom communication methods between chips, reducing memory size requirements, and implementing a unique mix-of-models strategy. This has allowed DeepSeek to achieve high efficiency and operational cost savings compared to models from companies like OpenAI and Meta.
The surprising news led to a wave of caution among investors, prompting analysts, including Mizuho’s Jordan Klein, to advise against buying AI stocks during this correction period. Evercore ISI analyst Amit Daryanani noted that the revelation might cause hyperscale cloud service providers to reconsider their capital expenditures in AI infrastructure, given the existence of a cheaper alternative. However, the panic in the market has been deemed overblown by some experts, with Bernstein analyst Stacy Rasgon affirming the ongoing demand for AI-related chips and investments, citing Meta’s capital expenditure increases and China’s announcement of a $140 billion AI funding plan.
Despite skepticism over the claims made by DeepSeek, including its use of Nvidia H800 chips, which some industry insiders believe may be more advanced than disclosed, the news does not necessarily spell disaster for U.S. companies. Analysts are still optimistic about the continued growth in AI investments and semiconductor demand, particularly for network edge AI applications, which benefit semiconductor firms like Qualcomm and Ambarella.
In summary, while DeepSeek’s achievements pose new challenges for U.S. tech companies and their stock prices, ongoing demands for AI technologies and infrastructure imply a robust future for the industry overall. This situation also prompts a re-evaluation of spending habits within the AI sector, but it offers a potential for innovation that could benefit smaller chip companies creating solutions for edge computing.