Chinese Startup DeepSeek Disrupts Global AI Landscape

Chinese startup DeepSeek is reshaping the AI industry with its advanced and cost-efficient models, challenging U.S. dominance. The Hangzhou-based firm’s AI model, DeepSeek-V3, has garnered global attention for requiring less than $6 million in Nvidia (NASDAQ:NVDA) H800 chip power for training, a fraction of typical costs.

DeepSeek’s AI Assistant, powered by DeepSeek-V3, has surpassed OpenAI’s ChatGPT to become the top-rated free app on Apple’s (NASDAQ:AAPL) U.S. App Store. Its innovative models, including the recently launched DeepSeek-R1, are on par with top offerings from OpenAI and Meta (NASDAQ:META), while being 20 to 50 times cheaper depending on the task.

Skepticism remains, with Scale AI CEO Alexandr Wang alleging that DeepSeek possesses 50,000 Nvidia H100 chips, potentially violating U.S. export restrictions. Additionally, Bernstein analysts question the accuracy of DeepSeek's claimed training costs, suggesting they may exceed the reported $5.58 million.

DeepSeek’s success is backed by Liang Wenfeng, co-founder of the hedge fund High-Flyer. The fund, which pivoted to focus on Artificial General Intelligence (AGI), has patents and operates clusters of A100 chips, further strengthening DeepSeek’s AI capabilities.

Beijing has acknowledged DeepSeek's significance in advancing China’s AI ambitions. Founder Liang attended a closed-door symposium with Premier Li Qiang on the day of DeepSeek-R1’s launch, underscoring the strategic importance of self-sufficiency in AI amidst U.S. export controls.

DeepSeek’s rapid ascent challenges global tech giants and showcases China’s potential to rival U.S. AI innovation, sparking shifts in the technology landscape.