DeepSeek, a Chinese startup, has managed to plummet the world’s largest company by market capitalization on the stock market.
It is the largest daily drop in value of a listed company in history whose chips are, or at least were, the favorite to drive artificial intelligence applications. Nvidia has dropped 16.9%, representing the largest drop for a company in Wall Street history, losing nearly $600 billion in market capitalization.
This plunge has been seconded by sharp falls in other companies in the sector, such as AMD, Micron, Arm Holdings, and technology giants such as Microsoft, Meta and Alphabet.
What is DeepSeek and why is it causing such a stir?
The Chinese company is dedicated to artificial intelligence. It is also the name of its AI chat, a proprietary alternative to others such as Copilot, Gemini. This artificial intelligence, has already managed to match and even surpass ChatGPT in downloads in mobile app stores. Moreover, in performance and reasoning tests, its R1 model has proven to be on par with other leading models.
Who is behind it?
DeepSeek, a startup founded in July 2023 in Hangzhou, China by Liang Wenfeng, a computer engineer with an ambitious vision of surpassing human capabilities with artificial general intelligence.
Under his leadership, DeepSeek has grown rapidly and has attracted the attention of investors and AI experts for its innovative approach and ability to produce high-quality models at a significantly lower cost.
What models has the company developed?
The company has developed AI linguistic models such as DeepSeek-V2 and the recent DeepSeek-V3, which want to compete with the hitherto American titans, but at a much lower cost, which has allowed training these advanced models, without the costs associated with the large infrastructures of other companies, has been developed for much less money, about $ 5 million.
Innovation and key AI features, DeepSeek-R1
Its models, such as DeepSeek-R1, have been innovative, capable of performing complex tasks such as reasoning, coding and mathematics, competing directly with other models such as ChatGPT.
DeepSeek’s latest models, released this month, are extremely fast and inexpensive. The DeepSeek-R1, the latest of the models developed with fewer chips, challenges the dominance of giants such as OpenAI, Google and Meta.
Advanced capabilities: DeepSeek-V3 has proven to be surprisingly efficient in reducing the amount of resources needed for training, using only a fraction of the infrastructure needed for other companies’ models.
IS DEEPSEEK FREE?
Free to users: Unlike other advanced AI services, such as OpenAI or Claude, DeepSeek is available for free. This has attracted a broad user base, without the costs associated with its competitors.
Open source model: DeepSeek stands out as an open source model, which allows any developer to customize and enhance the model. This is an advantage over models from companies such as OpenAI, which do not release their models in such an accessible way.
This is the first advanced artificial intelligence system of its kind available to users at no cost. Other powerful systems such as OpenAI or Claude Sonnet require a paid subscription, and some of these subscriptions even impose additional fees. Although Google Gemini also offers a free version, this is limited to older models. For the time being, DeepSeek has no such restrictions.
Conclusion: Is DeepSeek shaping the future of AI?
DeepSeek is proving that artificial intelligence does not have to be expensive or rely on large hardware infrastructures. Its ability to deliver advanced and affordable solutions is changing the AI landscape, and its growing popularity could alter the balance of power in the global technology industry. In a world where AI has the potential to transform every sector, DeepSeek is positioning itself as an emerging leader challenging the traditional giants with a more cost-effective and efficient offering.