What Kind of Company is DeepSeek? (Founding Story)

1. What Kind of Company is DeepSeek? (Founding Story)

In early 2024, Chinese AI startup DeepSeek made waves in the global AI market by unveiling its large language model, DeepSeek R1. Despite its relatively short history, DeepSeek has demonstrated China's AI capabilities on a global scale, positioning itself as a competitor to OpenAI, Mistral AI, and other major players in the field. Generative AI is no longer just a technological innovation; it has become a central element in the global battle for technological supremacy. As DeepSeek rapidly rises alongside OpenAI’s GPT series, Google’s Gemini, Meta’s LLaMA, and Mistral AI’s open-source models, the global AI market is shifting toward a triangular competition among the U.S., China, and Europe.

DeepSeek has gained attention by releasing a model that claims to provide powerful performance at a cost 27 times lower than its competitors, advocating for the democratization of AI. This strategy aligns with China's broader effort to build an independent AI ecosystem, countering the dominance of U.S. tech giants in AI development.

DeepSeek takes a distinct approach compared to existing Chinese AI companies. Its founder, Liang Wenfeng (梁文峰), emphasizes solving fundamental AI research challenges rather than merely achieving commercial success. His philosophy bears similarities to OpenAI CEO Sam Altman’s vision for developing Artificial General Intelligence (AGI).

DeepSeek's key technological differentiators include:

  1. High-performance, low-cost AI models—By optimizing algorithms and utilizing distributed learning techniques, DeepSeek has achieved global competitiveness despite China’s limited AI semiconductor resources.
  2. Open-source accessibility—DeepSeek R1 was released as an open-source model, maximizing accessibility for AI researchers and developers and expanding DeepSeek’s influence within the AI ecosystem.
  3. Adaptation to U.S. semiconductor sanctions—DeepSeek is structured to continue advancing AI training and inference technologies despite U.S. restrictions on semiconductor exports, aligning with China’s broader AI development strategy.

Unlike traditional Chinese AI firms, DeepSeek minimizes direct government control and fosters a more creative AI research environment. This approach is crucial for technological innovation and directly impacts the speed of AI industry development.

As U.S.-China AI competition intensifies, the policies of Donald Trump’s second term are beginning to reshape the AI industry. The Trump administration is expected to reverse AI regulations from the Biden era, prioritizing AI semiconductor and data industry investments to strengthen U.S. companies.

Recently, President Trump announced a $5 trillion 'Stargate AI Infrastructure' investment plan, marking an unprecedented level of support for the AI sector. The Stargate AI initiative aims to reinforce U.S. dominance in AI by establishing a national supercomputing network, AI semiconductor manufacturing hubs, and cloud computing infrastructure.

The three key pillars of Trump's second-term AI policy are:

  1. Expanded government support for U.S. AI firms—Companies like OpenAI, Google, and Microsoft will receive increased funding for AI R&D, semiconductor development, and cloud infrastructure expansion.
  2. Stronger restrictions on Chinese AI companies—U.S. export controls on AI semiconductors will persist, and firms like Huawei and DeepSeek may face tighter restrictions on global expansion.
  3. Integration of AI into critical industries—The U.S. will enhance AI applications in defense, finance, and cybersecurity to bolster economic and national security competitiveness.

These policies could further widen the technological gap between the U.S. and China in AI semiconductors and model training environments. Conversely, China is likely to respond by accelerating domestic AI semiconductor production and optimizing its own AI models.

The Trump administration is also expected to leverage the 'Department of Government Efficiency (DoGE)' to drive AI and technology policies. This move could foster stronger collaboration between the U.S. government and tech leaders like Elon Musk, potentially accelerating AI projects by companies such as Palantir, Tesla, and xAI through regulatory rollbacks.

2. How Does DeepSeek Compare to OpenAI’s O3-High and Deep Research?

DeepSeek is a large-scale AI model developed in China, designed to compete with OpenAI’s GPT series.

  • O3-High (OpenAI): A high-performance model based on GPT-4, optimized for U.S. and global markets with access to a vast AI ecosystem.
  • Deep Research (OpenAI): Specializes in research paper summarization and scientific analysis, excelling in academic data processing and advanced analytical capabilities.
  • DeepSeek (China): Primarily optimized for Chinese-language applications and aligned with China’s AI regulatory environment. It serves as a strategic model to strengthen China’s independent AI capabilities.

While DeepSeek is tailored for the Chinese market, its data accessibility and performance still lag behind OpenAI’s models in the global AI ecosystem.

3. How Do Mistral AI and Naver Clova Compare from a Sovereign AI Perspective?

Sovereign AI refers to a nation's ability to independently develop and operate AI technologies using domestic data and infrastructure.

  • Mistral AI (France): A leading proponent of Europe’s Sovereign AI strategy, focusing on open-source AI models to reduce Europe’s dependence on U.S. and Chinese AI technologies.
  • Naver Clova (South Korea): Aims to build an independent Korean AI and cloud ecosystem, with a focus on Korean-language AI models and regional AI infrastructure.

Key Differences:

  • Mistral AI is committed to European AI sovereignty, developing open-source models to establish an independent European AI ecosystem.
  • Naver Clova focuses on South Korea and Asian markets, working with the Korean government and local businesses to advance AI development.

Despite their differences, both companies prioritize data sovereignty and AI independence, ensuring that their respective regions maintain control over AI infrastructure and technological advancements.

4. How Will DeepSeek and BYD Collaborate?

In February 2025, BYD and DeepSeek announced a partnership in autonomous driving. The collaboration will integrate DeepSeek’s AI software into BYD’s vehicles, enhancing its ‘God’s Eye’ driver assistance system. This system includes automated parking and highway navigation features, with plans to introduce them even in budget models priced below 69,800 RMB (~€9,200).

DeepSeek’s AI technology will improve the accuracy and reliability of BYD’s autonomous driving systems, offering a personalized driving experience. The partnership aims to give BYD a competitive edge over Tesla in the global EV market.

DeepSeek’s R1 model is particularly strong in logical reasoning, mathematical problem-solving, and real-time decision-making, which are critical for self-driving systems.

Key Aspects of the Partnership:

  • Integration into 21+ BYD models, including budget-friendly options like the Seagull hatchback.
  • Focus on making autonomous driving features a standard safety feature, not just a premium add-on.
  • BYD’s stock surged 4.5% on the Hong Kong Stock Exchange following the announcement, reflecting investor confidence.

Potential Challenges:

  • Data privacy concerns—DeepSeek’s data-sharing practices have raised fears of government surveillance due to its connection to China Mobile’s servers.
  • BYD’s global expansion risks—Data privacy concerns could hinder BYD’s entry into international markets.
  • No plans for sales in South Korea—BYD Korea confirmed it will not sell DeepSeek-integrated vehicles due to concerns over AI chatbot data leaks.

Additionally, DeepSeek continues to release open-source AI projects on GitHub. Its DeepEP library, designed for Mixture-of-Experts (MoE) and Expert Parallelism (EP), enables efficient distributed AI training. This FP8-enabled communication framework could be critical for both self-driving technology and broader AI model development.

댓글

댓글 쓰기

이 블로그의 인기 게시물

About the Author: Drytree

About the Author: Drytree (Jaehoon Lee) I am a senior researcher specializing in deep tech startups, technology commercialization, and R&BD strategy at a national research institute in South Korea. With a deep curiosity about the rapid evolution of technology and its impact on society, I have been writing under the pen name Drytree since 2021, aiming to make complex technological trends more accessible and foster informed discussions on innovation and policy. My academic background is diverse, spanning Political Science (B.A.), International Politics (M.A.), and currently, a Ph.D. in Technology Management & Entrepreneurship . I have published 13 academic papers in domestic and international journals, covering topics such as patent analysis, technology security, and industrial policy . Books & Publications I have authored several books exploring technology, startups, and geopolitical implications of innovation , including: Elon Musk’s DOGE: The Dawn of a New Economic Exper...
자세한 내용 보기

Leading the AI Era: NVIDIA’s Journey of Innovation and Transformation

  NVIDIA’s story began in 1993 when Jensen Huang, Chris Malachowsky, and Curtis Priem founded the company in Silicon Valley with a bold vision: to revolutionize graphics processing and redefine the multimedia experience. At a time when computer games and graphic software were rapidly gaining traction, existing hardware struggled to keep pace with demand. NVIDIA set out to overcome these limitations by developing high-performance graphics hardware that would transform not only gaming but the entire computing landscape. The Birth of a Game-Changer: The Rise of the GPU In its early years, NVIDIA’s primary focus was on creating a graphics processing unit (GPU) that could deliver unprecedented performance for gaming and multimedia applications. This vision was brilliantly realized in 1999 with the launch of the GeForce 256—the world’s first GPU. The GeForce 256 was a breakthrough product, capable of real-time 3D rendering with support for 32-bit color, transformation, and lighting effec...
자세한 내용 보기

Elon's DOGE will expand their influence in NASA

Elon Musk’s Department of Government Efficiency (DOGE) is expanding its influence over U.S. federal agencies to the space sector by intervening in NASA’s budget execution. According to reports from Bloomberg and Reuters, DOGE officials are set to conduct a comprehensive review of NASA’s expenditures, signaling a new shift in the government’s budget management approach. Related Articles: 🔗 Reuters - Acting NASA Chief Says DOGE Plans to Examine Space Agency’s Spending (Feb 12, 2025) 🔗 Bloomberg - Musk’s DOGE Workers Are Headed to NASA for Spending Review (Feb 12, 2025) NASA’s acting administrator stated that DOGE intends to analyze the agency’s financial management and assess its spending structure. This process may lead to adjustments in the budget for space exploration and research and development. As the leading institution for U.S. space development, NASA is spearheading major projects such as the Artemis program and the Lunar Gateway . Given that Musk’s SpaceX collaborates with N...
자세한 내용 보기