DeepSeek, a project backed by the Chinese hedge fund High-Flyer, is making waves in global AI and technology circles. Modeled similarly to OpenAI’s ChatGPT, DeepSeek was reportedly trained on a budget of just US$6 million — a stark contrast to the roughly $100 million used for training OpenAI’s GPT-4 in 2023 and a fraction of Meta’s budget for LLaMA 3.1.
Launched on January 10th, DeepSeek initially rolled out quietly, as a free AI assistant in the same vein as ChatGPT. Promoted as requiring far less data and developed at a fraction of current AI model costs, DeepSeek gained massive popularity within days. It quickly became the most downloaded AI chatbot app globally.
However, the sudden success of DeepSeek sent shockwaves through financial markets. On January 27th, global investors dumped tech stocks amid fears that DeepSeek could challenge the industry dominance of traditional leaders like NVIDIA. The reaction was swift and dramatic: NVIDIA lost $593 billion in market capitalization, with other major tech companies experiencing similar declines.
What is DeepSeek?
DeepSeek R1 is a next-generation AI chatbot model, designed to operate much like ChatGPT — accepting user prompts and delivering responses in seconds. Early testers have found its capabilities comparable to Google’s Gemini and OpenAI’s GPT-4.
What truly stands out is DeepSeek’s ultra-low development cost. Its creators claim the entire project cost just $6 million to train, a fraction of the resources invested by competitors.
Additionally, reports suggest DeepSeek strategically navigated U.S. export restrictions on advanced AI chips. Despite a ban on exporting Nvidia A100 chips to China from September 2022, DeepSeek’s team secured just enough A100 chips and supplemented them with cheaper, less sophisticated hardware — managing to achieve surprisingly efficient results.
This has alarmed U.S. tech leaders, especially since Washington has tried to limit the supply of advanced chips to China for security reasons. DeepSeek’s backers further revealed, via their official WeChat account, that they have operated a cluster of 10,000 A100 chips since July 2022.
DeepSeek is also engineered to use less memory, contributing to its cost efficiency. Its combination of speed and affordability drove it to the number-one spot on Apple’s App Store in its first week of release in the United States.
Who is Liang Wenfeng, the Founder of DeepSeek?
DeepSeek was developed under the leadership of Liang Wenfeng, who co-founded High-Flyer, a hedge fund based in Hangzhou, China. Founded in February 2016, High-Flyer traditionally focused on trading but shifted gears in March 2023 when Liang announced an ambitious AI research initiative.
Through its official channels, the hedge fund declared it would refocus efforts on building a new research group dedicated to Artificial General Intelligence (AGI) — systems capable of outperforming humans across a broad range of tasks.
In November 2023, DeepSeek released its first product, DeepSeek Code, followed later that month by the DeepSeek LLM (large language model) series. In January 2025, the company introduced its first public-facing chatbot, based on the DeepSeek R1 model, for Android and iOS platforms. By January 27th, 2024, DeepSeek had overtaken ChatGPT as the top free app on the U.S. Apple Store, contributing to an 18% plunge in NVIDIA’s share price and triggering wider market volatility.
DeepSeek’s Impact on the Generative AI Industry
Despite the market panic, many experts view DeepSeek’s arrival as a positive development for generative AI. Its success highlights China’s progress in AI, demonstrating that cutting-edge capabilities can be delivered at far lower costs.
If ChatGPT showed the transformative potential of generative AI, DeepSeek has proven that similar results can be achieved on a budget, broadening global access to the technology.
According to Andrew Yao Chi-chih, a Turing Award winner from Tsinghua University, DeepSeek’s open-source nature could ultimately benefit the wider research community.
Local Chinese companies have also embraced DeepSeek enthusiastically. Firms such as Lenovo, Shenzhen-based UBTech, and automaker Geely have already integrated DeepSeek’s technology to enhance their products and streamline operations.
In summary, DeepSeek is a landmark development that not only challenges industry giants but also signals a shift toward more cost-effective, democratized AI worldwide.