Understand any stock in simple English
Enter a ticker to learn what the company does and how it makes money.
DeepSeek builds large artificial intelligence models and gives them away cheaply, making money by selling API access and chat services.
DeepSeek is a Chinese AI company that builds large language models — the kind of software that powers chatbots and coding assistants. People use its free chatbot app and website to ask questions, write code, and get help with work, much like ChatGPT or Claude.
The business makes money mainly by selling API access, which lets other companies and developers plug DeepSeek's models into their own products and pay per use. It also offers a paid chat service. Its models are unusually cheap compared with rivals, and DeepSeek says its software is efficient, so it needs less computer power to run. That matters because the biggest cost in AI is renting time on powerful chips. DeepSeek is one of the most closely watched AI labs in the world. The company is backed by the Chinese quant trading firm High-Flyer, which was co-founded by DeepSeek's CEO, Liang Wenfeng. In January 2025, DeepSeek released an AI model it claimed was built for a fraction of the cost of top US models, and it briefly became the most downloaded app in the US App Store. That shook markets and raised questions about whether US restrictions on advanced chip exports to China were really slowing the country's AI progress.
DeepSeek started in the summer of 2023 as a side project of High-Flyer, a quantitative hedge fund run by Liang Wenfeng, working out of a shared office in Hangzhou. The fund had stockpiled Nvidia chips and paid the bills itself, no outside investors. From day one the team aimed at something idealistic: open source, cheaper AI, and AI for everyone.
DeepSeek shipped its first big model in early 2024, then kept releasing new versions through the year, each built on a design that only wakes up the parts of the model it needs for a given task, which saves a lot of computing. The December model matched top closed rivals like GPT-4o, and the striking part was the price: it was trained for a few million dollars, a tiny fraction of what the American leaders spent.
In January 2025 DeepSeek released R1, a model built to reason through problems step by step, and it landed like a bomb. It performed near the best US models at a fraction of the cost, and the investor Marc Andreessen called it a Sputnik moment. US tech stocks sold off and DeepSeek became a global name overnight.
Days later the DeepSeek app passed ChatGPT as the most downloaded free app on the US iPhone App Store, a stunning turn for a company almost nobody outside China had heard of. Markets read it as proof that AI could be done far more cheaply than assumed, and Nvidia's share price fell 18% in a single day.
Through 2026 the company finally took outside money. In July it closed its first external funding round, pulling in more than 50 billion yuan from investors including a national AI industry fund. By September it had hired CITIC Securities to prepare a listing on Shanghai's STAR Market.
In April 2026 DeepSeek previewed its V4 series, which could handle a million tokens of context, roughly the length of several books, and released it under a permissive open license. Chinese chipmakers Huawei and Cambricon adopted it, tying DeepSeek's software more tightly to domestic hardware.
DeepSeek raised around $15 billion in one of China's biggest financing rounds ever, twice raising its target as investors piled in. The money was meant to keep pace with Western frontier labs and fund expansion abroad, and it set up the company for a public listing.
DeepSeek is now preparing to list on Shanghai's STAR Market within the year, with CITIC Securities again doing the work. It had already raised 7.4 billion in June 2026 from Liang, Tencent and CATL, at a valuation that later reached $75 billion. The push reflects China's AI IPO fever and the race to fund data centers and talent against US rivals.
That moment tells you both the promise and the risk. DeepSeek is a serious technical player with a loyal following, but it competes against far richer and better-funded companies like OpenAI, Google, and Anthropic. It cannot easily buy the most advanced chips because of US export rules, so it leans on clever engineering instead. It is also a Chinese company, which makes some businesses and governments nervous about using its services. It does not publish detailed financial results, so its revenue, profitability, and funding remain largely a private matter.
Saved answer · last researched 20 Oct 2026