DeepSeek has rapidly gained widespread attention. The Chinese AI lab, DeepSeek, captured the public interest after its chatbot application ascended to the top of the Apple App Store and Google Play Store charts. DeepSeek’s AI models, which employ compute-efficient techniques, have prompted Wall Street analysts and technologists to question the United States’ ability to maintain its lead in the AI competition and whether the demand for AI chips will persist.
DeepSeek’s origins trace back to High-Flyer Capital Management, a Chinese quantitative hedge fund utilizing AI for trading decisions. Liang Wenfeng, a co-founder and AI enthusiast, initiated High-Flyer in 2015. Wenfeng, who began engaging in trading during his time at Zhejiang University, launched High-Flyer Capital Management as a hedge fund in 2019 with a focus on AI algorithm development and deployment.
In 2023, High-Flyer established DeepSeek as a lab dedicated to researching AI tools, separate from its financial operations. With High-Flyer as an investor, the lab later spun off into an independent company, also called DeepSeek. From its inception, DeepSeek developed its own data center clusters for model training. However, like other Chinese AI companies, DeepSeek has been affected by U.S. export restrictions on hardware, leading them to use Nvidia H800 chips—a less powerful version than what U.S. companies access—for training one of its models.
The technical team at DeepSeek is reportedly young, with aggressive recruitment of Ph.D. AI researchers from top Chinese universities. The company also hires individuals without computer science backgrounds to broaden the tech’s understanding across various subjects, according to The New York Times.
DeepSeek introduced its initial set of models—DeepSeek Coder, DeepSeek LLM, and DeepSeek Chat—in November 2023. The AI industry began to notice the company following the release of its next-generation DeepSeek-V2 model family last spring. DeepSeek-V2, a versatile text and image analysis system, performed well on numerous AI benchmarks and was significantly more cost-effective than its contemporaries, compelling domestic competitors like ByteDance and Alibaba to lower usage prices for some models and offer others for free.
The DeepSeek-V3 model, launched in December 2024, further increased the company’s prominence. According to internal benchmarks, DeepSeek V3 outperforms both downloadable models like Meta’s Llama and API-accessible models like OpenAI’s GPT-4o. DeepSeek’s R1 reasoning model, released in January, is claimed to match OpenAI’s o1 model on key benchmarks. As a reasoning model, R1 effectively self-verifies, making it more reliable in fields such as physics, science, and mathematics. However, due to its Chinese origins, DeepSeek’s models are subject to scrutiny by China’s internet regulator to ensure their responses align with “core socialist values,” limiting R1’s ability to address sensitive topics like Tiananmen Square or Taiwan’s autonomy.
In March, DeepSeek surpassed 16.5 million visits, although its traffic had decreased by 25% from February. Despite trailing behind ChatGPT, which had over 500 million weekly active users, DeepSeek’s rise has been noteworthy.
While DeepSeek’s business model remains unclear, it prices its offerings below market value and some are free. The company has not yet accepted venture capital funding, citing efficiency breakthroughs as enabling cost-effectiveness. Nonetheless, some experts challenge the accuracy of DeepSeek’s claims.
Developers have embraced DeepSeek’s models, which, though not open source, are available under licenses permitting commercial use. According to Clem Delangue, CEO of Hugging Face, the platform hosts over 500 derivative models of R1, totaling 2.5 million downloads.
DeepSeek’s ascent has been described as significant yet potentially over-hyped, influencing industry trends and financial outcomes, including an 18% drop in Nvidia’s stock price in January. The increasing influence of DeepSeek has prompted a public acknowledgment from OpenAI CEO Sam Altman and defensive strategies from other industry leaders. Additionally, the U.S. Commerce department has banned DeepSeek on government devices.
Microsoft has incorporated DeepSeek into its Azure AI Foundry service, while Meta continues to emphasize AI infrastructure spending. OpenAI has labeled DeepSeek as “state-subsidized,” urging the U.S. government to consider banning it. Despite this, Nvidia’s CEO praised DeepSeek for its innovation, which necessitates significant computational resources.
Some countries and companies have banned DeepSeek due to perceived risks, and New York State prohibits its use on government devices. Looking ahead, the U.S. government appears increasingly cautious about the potential impacts of foreign tech influences, with the possibility of banning DeepSeek on government devices being considered.
This narrative was initially published on January 28, 2025, and will continue to be updated.
