Nvidia, Nasdaq and power utilities tumble, DeepSeek reports attack
Make us a preferred source
Make us a preferred source on Google
More news: DeepSeek has restricted new user registrations to Chinese mainland phone numbers only after what the Chinese AI startup said were "large-scale malicious attacks" on its services.
The app surpassed OpenAI’s ChatGPT as the most downloaded free app on Apple’s US App Store. The chatbot was encountering issues on Monday, according to its status page, but had resolved login and API issues.
Meanwhile, the buzz around DeepSeek’s low-cost AI models triggered a global tech stock sell-off, with Nvidia shares leading losses - down more than 17% to US$117.26, putting it on track for a potential one-day market value loss record for any company - and the Nasdaq falling 3.25% on Monday local time.
The Philadelphia Semiconductor Index dropped 7.9%, and power utilities like Vistra and Constellation Energy saw declines of over 18%.
Analysts said DeepSeek’s efficiency could disrupt AI development costs and demand for chips and data centres, while investors turned to safe-haven assets, driving US Treasury yields lower.
What they said: "DeepSeek's online services have recently faced large-scale malicious attacks," the company said, adding that existing users could log in as usual.
China’s DeepSeek AI rattles global tech markets
The news: Chinese AI startup DeepSeek triggered a selloff in global tech stocks with the release of cost-effective AI models, which rival OpenAI and Meta’s offerings while using less advanced hardware.
The numbers: The Nasdaq 100 futures fell over 5% on Monday (Tuesday AEDT), while Nvidia futures and ASML shares dropped 10% and 12%, respectively. The Nasdaq 100 and Europe’s Stoxx 600 technology sub-index were together set for a market capitalization wipeout of USD1.2 trillion ($1.91 trillion), if the losses hold, according to Bloomberg.
Bonds rallied and bitcoin sank as investors sought safety.
DeepSeek’s AI assistant was released last Monday and has already overtaken ChatGPT on the Apple App Store as the top free download.
The context: Founded by Liang Wenfeng, a former quant fund chief, DeepSeek developed the open-source DeepSeek-V3 AI model, training it using Nvidia’s H800 chips at an alleged cost of less than USD6 million, significantly undercutting the typical costs of over USD100 million.
It's latest model, DeepSeek-R1, is a reasoning-focused AI model designed to challenge OpenAI’s o1 model with advanced capabilities, leveraging reinforcement learning for self-improvement.
The developments challenge the need for massive AI investments by US firms and highlights vulnerabilities in the effectiveness of Washington’s export controls targeting China’s advanced chip and AI capabilities.
While US and European tech stocks tumbled, some Chinese firms, including Baidu and Alibaba, saw gains.
What they said: Analysts warned the development has the potential to derail the western investment case, based on massive AI hardware spending from a handful of hyperscalers, and disrupt US dominance in AI.
"We still don’t know the details and nothing has been 100% confirmed in regards to the claims, but if there truly has been a breakthrough in the cost to train models from USD100 million+ to this alleged USD6 million number, this is actually very positive for productivity and AI end users as cost is obviously much lower meaning lower cost of access," said Jon Withaar, a senior portfolio manager at Pictet Asset Management.
Silicon Valley venture capitalist Marc Andreessen described DeepSeek’s release as AI’s “Sputnik moment,” referencing the former Soviet Union satellite during the 1950s space race.
