Global technology shares experienced sharp declines for a second consecutive day, driven by the emergence of a cost-efficient artificial intelligence (AI) model from Chinese startup DeepSeek. The selloff, which began this week, has raised concerns about the sustainability of high valuations for major AI firms.
Nvidia, one of the leading beneficiaries of the AI boom, saw its shares fall nearly 17% on Monday, erasing $593bn from its market value. According to data from LSEG, this marks the largest single-day loss in market capitalisation for any Wall Street company. Other major tech firms were also impacted, with Broadcom’s shares falling 17.4%, Microsoft losing 2.1%, and Alphabet dropping 4.2%.
The market reaction stemmed from the release of DeepSeek’s AI assistant, which the company claims operates with significantly reduced costs and data compared to US-built models. The Hangzhou-based startup reported that its DeepSeek-V3 model, launched on 10 January, was trained using Nvidia’s H800 chips at a cost of under $6m. This figure contrasts with the billions spent by Silicon Valley companies on similar AI advancements.
DeepSeek’s R1 model, which was made open source, has reportedly outperformed existing models from US companies, including OpenAI. Nvidia described the model as “an excellent AI advancement” despite its role in sparking a selloff in the company shares. OpenAI CEO Sam Altman also commented on the development, calling it an “impressive model,” on social media platform X.
The launch of DeepSeek’s AI assistant has highlighted the increasingly competitive nature of the global AI sector. Over the weekend, DeepSeek became the most downloaded free app on Apple’s US App Store, overtaking ChatGPT.
The broader selloff has highlighted the vast sums of money funnelled into AI research and development in recent years. Since November 2022, when OpenAI’s ChatGPT sparked the AI boom, the combined market value of the “Magnificent Seven” tech firms has increased by approximately $10 trillion. However, the emergence of cost-efficient models like DeepSeek’s has raised concerns about whether these inflated valuations are sustainable. DeepSeek’s success contrasts sharply with the muted reception of Baidu’s ChatGPT-like model, underscoring the disparities in AI capabilities between China and the US.
Cyberattack hits DeepSeek amid rising popularity
DeepSeek has faced additional challenges amid its surge in popularity. The company temporarily disabled new registrations for its DeepSeek-V3 platform, citing a large-scale cyberattack targeting its services. “Due to large-scale malicious attacks on DeepSeek’s services, we are temporarily limiting registrations to ensure continued service,” the company stated on its status page. “Existing users can log in as usual. Thanks for your understanding and support.”
The attack, believed to be a distributed denial-of-service (DDoS) incident, disrupted the platform’s API and web chat services. In a DDoS attack, a significant volume of traffic is directed at specific servers, overwhelming resources and causing disruptions. While DeepSeek has not provided details about the perpetrators, the incident has raised questions about potential involvement from competitors or other threat actors.
It also remains unclear as to how DeepSeek is using user input data to train and re-train its models, Dataiku’s director of AI governance, Jacob Beswick, told Tech Monitor. “In the context of China’s different data regime,” asks Beswick, “what data has gone into training their model? And are we individually, collectively, and organisationally, aligned to whatever approach they’ve taken?
“Ultimately, the UK government needs to be asking itself whether DeepSeek aligns with its collective expectations when it comes to AI safety, security, and risk. These are questions that should be asked of any provider, Chinese or otherwise.”
Read more: OpenAI poised to launch o3 mini language model
