DeepSeek is a Chinese artificial intelligence company that develops open-weight large language models (LLMs). Based in Hangzhou, Zhejiang, it is owned and funded by High-Flyer, a Chinese hedge fund. The company was founded in July 2023 by Liang Wenfeng and focuses on research-oriented AI development, with its models made publicly available under open-weight licenses.
DeepSeek's approach emphasizes computational efficiency and broad recruitment beyond traditional computer science fields. Its models have been hosted by major cloud providers and have gained international attention for their performance and accessibility, particularly in regions seeking alternatives to Western AI platforms.
Founding and Early History
DeepSeek traces its origins to High-Flyer, co-founded by Liang Wenfeng in June 2015. High-Flyer began using GPU-dependent deep learning models for stock trading on 21 October 2016, transitioning from earlier CPU-based linear models. By the end of 2017, most of its trading operations were driven by AI.
In 2019, High-Flyer constructed its first computing cluster, Fire-Flyer, at a cost of 200 million yuan. The cluster contained 1,100 GPUs interconnected at 200 Gbit/s and operated for 1.5 years before being retired. By 2021, Liang had begun purchasing large quantities of Nvidia GPUs, reportedly obtaining 10,000 Nvidia A100 GPUs before United States restrictions on chip sales to China took effect.
Construction of Fire-Flyer 2 began in 2021 with a budget of 1 billion yuan. In 2022, the cluster's capacity was used at over 96%, totaling 56.74 million GPU hours, with 27% of capacity supporting scientific computing outside the company. Fire-Flyer 2 initially used 5,000 PCIe A100 GPUs in 625 nodes, each containing 8 GPUs, later incorporating NVLinks and Nvidia Collective Communications Library for larger models.
On 14 April 2023, High-Flyer announced the launch of an artificial general intelligence (AGI) research lab. Two months later, on 17 July 2023, that lab was spun off into an independent company named DeepSeek, with High-Flyer as its principal investor. Initially, venture capital investors were reluctant to provide funding due to concerns about quick exits.
2024 Model Releases
In 2024, DeepSeek released several significant models under its open-weight framework. DeepSeek-V2, introduced in early 2024, demonstrated advances in model architecture and training efficiency. The company continued refining its algorithms to maximize computational efficiency, leveraging older hardware and reducing energy consumption due to U.S. chip restrictions.
DeepSeek-R1, launched in January 2025 alongside an eponymous chatbot, marked a major milestone. The model gained international recognition for its reasoning capabilities and cost-effective training approach. Since January 2025, DeepSeek has released its models under free and open-source software licenses.
Company Operations
DeepSeek is headquartered in Hangzhou, Zhejiang, with Liang Wenfeng serving as CEO. As of May 2024, Liang personally held an 84% stake in DeepSeek through two shell corporations. The company focuses on research and has stated it does not have immediate plans for commercialization, allowing it to skirt certain provisions of China's AI regulations aimed at consumer-facing technologies.
Due to its open-weight framework, DeepSeek models have been hosted natively by cloud providers including Microsoft Azure and Perplexity AI. In February 2026, Anthropic accused DeepSeek of using thousands of fraudulent accounts to generate millions of conversations with Claude to train its own LLMs. In April 2026, investors began discussions for a $300 million funding round, which would bring the company to a $10 billion valuation. DeepSeek completed a May 2026 Series A round receiving US$7 billion, reaching a post-money valuation of US$52 billion. In July 2026, Bloomberg and the Financial Times reported preparations for an IPO as soon as 2027, with another discussion targeting a US$70 billion pre-money valuation.
Training Framework
DeepSeek operates the Fire-Flyer and Fire-Flyer 2 computing clusters. Fire-Flyer 2, still in operation as of 2025, consists of co-designed software and hardware architecture. The hardware uses Nvidia GPUs with 200 Gbps interconnects, divided into two zones with cross-zone task support. The network topology uses two fat trees for high bisection bandwidth.
Key software components include 3FS (Fire-Flyer File System), a distributed parallel file system designed for asynchronous random reads using Direct I/O and RDMA Read. Since each data read is random and not reused, caching is unnecessary. Another component is hfreduce, an asynchronous communication library originally designed to replace Nvidia Collective Communications Library functions.
Strategy and Hiring
DeepSeek's hiring approach emphasizes skills over lengthy work experience, resulting in many hires fresh out of university. The company recruits individuals without computer science backgrounds to expand expertise in areas such as poetry and advanced mathematics. According to The New York Times, dozens of DeepSeek researchers have or have previously had affiliations with People's Liberation Army laboratories and the Seven Sons of National Defence.
Since 2025, DeepSeek's chatbot was adopted in non-combat roles by the People's Liberation Army, including hospitals, the People's Armed Police paramilitary, and national mobilization organizations. The company has also expanded into Africa, offering more affordable and less power-hungry AI solutions, bolstering African language models and generating startups, for example in Nairobi. Along with Huawei's storage and cloud computing services, DeepSeek's impact on sub-Saharan Africa's tech scene is considerable, offering local data sovereignty and flexibility compared to Western AI platforms.
Reception and Impact
DeepSeek's models have been recognized for their efficiency and open accessibility within the Large language model ecosystem. The company's focus on research over commercialization distinguishes it from competitors like OpenAI and Anthropic. Its training innovations, including the Fire-Flyer clusters and custom software, contribute to broader developments in Machine learning and Deep learning.
DeepSeek's expansion into Africa and its cost-effective solutions have positioned it as a significant player in global AI adoption, particularly in regions with limited infrastructure. The company's open-weight approach allows for local customization and deployment, fostering a growing community of developers and startups.