DeepSeek threat? Softbank, Advantest, other Japanese tech stocks sink up to 10%

Shares of Japanese chip-related firms were feeling the heat on Monday.

Today's slump comes as Chinese AI startup DeepSeek gained traction with its updated AI model, raising fears about potential challenges to US technological dominance.

The sell-off also reflected profit-taking after last week’s rally in tech stocks driven by SoftBank Group Corp.'s AI investment plans.

Investor confidence was weighed down by concerns over DeepSeek's competitive AI model. Nasdaq futures also fell, amplifying the negative sentiment.

Softbank, Advantest slump up to 10%

Advantest Corp., a supplier to Nvidia, dropped 8.2%. Disco Corp. fell 3.3%. SoftBank Group declined 6.28% after soaring 16% last week.

Data-center cable makers Furukawa Electric Co. and Fujikura Ltd. each over 8%.

Last week, Japanese tech shares rallied as SoftBank Group announced plans to collaborate with companies like OpenAI and Oracle Corp. to invest in AI infrastructure in the US.

However, the emergence of DeepSeek and its potentially competitive AI model has introduced fresh uncertainty into the market, sparking a wave of sell-offs in tech and chip-related stocks.

Andrew Jackson, head of Japan equity strategy at Ortus Advisors Pte, told Bloomberg:

The DeepSeek threat

China's DeepSeek has shocked the global tech industry by releasing AI models that rival the best in the US.

These models are said to be built at a fraction of the cost and use less powerful hardware, raising concerns about the diminishing lead of American AI.

In December, DeepSeek unveiled a free, open-source large language model that was developed in just two months for under $6 million.

The lab used Nvidia's H800 chips, less powerful than the restricted H100 chips, which are subject to US export controls.

Despite this, the model has delivered results that outperformed Meta’s Llama 3.1, OpenAI’s GPT-4o, and Anthropic’s Claude Sonnet 3.5 in third-party benchmarks.

Last week, DeepSeek launched another breakthrough model called r1, a reasoning model that surpassed OpenAI’s latest o1 in areas like problem-solving, math, and coding.

This development has intensified concerns over Silicon Valley's massive spending on AI models and data centers, as DeepSeek has achieved competitive results with significantly fewer resources.

The advancements from DeepSeek have sparked global attention, including remarks from Microsoft CEO Satya Nadella at the World Economic Forum. Nadella praised the model's efficiency.

He added:

The emergence of DeepSeek has shaken up the AI landscape majorly dominated by the US.

The lab's ability to build cost-effective, high-performing models signals growing competition to US's leadership in field of AI.