DeepSeek, an AI company based in Hangzhou, China, has stunned the tech and financial markets with its low-cost, high-performance AI model, R1.
What was a simple university startup just a few years ago, has rapidly risen demonstrating an ambition to develop artificial general intelligence (AGI)—a level of human-like intelligence that no company has yet achieved. Although AGI remains elusive, DeepSeek’s innovative approach to AI development sets it apart from its US-based competitors.
DeepSeek has focused on creating a model that is significantly cheaper to operate. This efficiency may partly stem from the need for Chinese computer scientists to navigate US restrictions on chip availability, forcing them to innovate with fewer resources. Despite its significant technical achievements, the company has maintained a relatively low profile compared to high-profile Silicon Valley firms like OpenAI and Meta. However, its blue whale logo has begun to gain visibility as its app becomes increasingly popular worldwide.
The R1 model, which rivals OpenAI’s GPT-4, is causing a stir for its ability to perform complex reasoning tasks, particularly in mathematics and coding.
What sets it apart is its unique internal architecture that reduces memory requirements and operational costs. Researchers have praised its efficiency, with DeepSeek claiming it developed the model in just two months for under $6 million—a stark contrast to the billions spent by US tech giants. Although some question these cost estimates, the model’s ability to deliver comparable results with minimal resources is remarkable.
Liang Wenfeng, DeepSeek’s founder and a former quantitative hedge fund manager, is a key figure in the company’s rise. He emphasises that DeepSeek aims to lead technological innovation rather than merely monetise applications, marking a shift in the role of Chinese firms in global tech development. This vision is reflected in the company’s focus on advancing AI ecosystems rather than looking at short-term profits.
Why is this important?
The success of DeepSeek’s R1 model has already sent shockwaves through the AI industry. News of its performance has already caused US tech stocks to dip, with Nvidia—a major supplier of AI hardware—losing $600 billion in market value. DeepSeek’s efficient model raises questions about the necessity of vast investments in computing infrastructure and energy, challenging the prevailing narrative among US tech firms.
Although DeepSeek has not achieved AGI, it has demonstrated that cutting-edge AI can be developed and operated at a fraction of the cost. This breakthrough suggests a potential shift in the industry’s future, reducing reliance on expensive hardware and energy. While the long-term implications remain uncertain, DeepSeek’s rise signals a transformative moment for AI development worldwide.