A Chinese startup is making waves in the world of artificial intelligence, with DeepSeek launching advanced AI models which the company claims are as good, or even better, than the top models from U.S. tech giants. What’s more, DeepSeek’s models are much cheaper to develop and use-a fact that has raised a lot of interest and questions over the future of the global AI industry.

DeepSeek’s latest AI model, called DeepSeek-V3, was trained using less than $6 million worth of computing power from Nvidia H800 chips. This is a fraction of the cost that other companies, like OpenAI and Meta, spend on their AI models. DeepSeek’s AI Assistant, powered by DeepSeek-V3, has already become the top-rated free app on Apple’s App Store in the U.S., surpassing OpenAI’s ChatGPT. The success has made many people wonder why U.S. tech companies are spending billions of dollars to develop AI when a small company like DeepSeek can achieve similar results at a fraction of the cost. The rise of DeepSeek has meanwhile seen the big tech companies, including Nvidia, which makes the chips needed to train the models, see their stock prices take hits. The justification of high AI development costs in the U.S., investors are asking themselves increasingly, when companies like DeepSeek can so easily turn out high-quality models for so much less.

Why is DeepSeek Making Headlines?

When OpenAI, a San Francisco-based company, released ChatGPT in late 2022, it kicked off a mad race among tech companies around the world racing to develop their own AI chatbots. In China, companies like Baidu quickly launched their versions of ChatGPT, but many were disappointed by the gap in quality between Chinese and U.S. AI models.

DeepSeek has turned this narrative around. Silicon Valley executives and engineers have hailed the performance and cost efficiency of the company’s models, DeepSeek-V3 and DeepSeek-R1. DeepSeek says its models can do what the most advanced ones from OpenAI and Meta do but are a fraction of the cost to use. For example, the DeepSeek-R1 model, launched recently, is a good 20 to 50 times cheaper to use than models from OpenAI, depending on the task.

Yet, not everyone is convinced. Several experts have come forward to doubt DeepSeek’s claims. Recently, Scale AI CEO Alexandr Wang suggested that DeepSeek may be using 50,000 Nvidia H100 chips, which would violate U.S. export controls banning the sale of advanced chips to Chinese companies. DeepSeek has not responded to these allegations yet.

Analysts at Bernstein also said the total training costs for DeepSeek’s V3 model are way higher than the $5.58 million the company reported. They also said the training costs for the R1 model have not been disclosed.

Who is Behind DeepSeek?

DeepSeek is headquartered in Hangzhou, China, and is backed by Liang Wenfeng, co-founder of a quantitative hedge fund called High-Flyer. In March 2023, High-Flyer said it was sidelining its trading operations to focus on AI research. The fund set up a research team to work on a branch of AI known as Artificial General Intelligence, referring to AI capable of performing better than humans on most tasks. DeepSeek was established later that year.

How much High-Flyer has invested in DeepSeek is not known, but the two share an office building. High-Flyer also owns patents related to chip clusters for training AI models. In July 2022, High-Flyer’s AI unit said on its official WeChat account that it operates a cluster of 10,000 A100 chips, which are used for AI training.

How Does China View DeepSeek?

The success of DeepSeek has not gone unnoticed in China’s political circles. On January 20, the same day DeepSeek-R1 was released, its founder Liang Wenfeng attended a closed-door meeting hosted by Chinese Premier Li Qiang along with business leaders and experts; Liang’s presence gives a sign that DeepSeek’s achievements mean something for China to reach self-sufficiency in strategic industries like AI.

This is not the first time a Chinese technology chief has been invited to such a high-ranking meeting; last year, the CEO of Baidu, Robin Li, was also invited to a symposium. Such meetings show how seriously the Chinese government takes the rearing of homegrown tech companies to reduce reliance on foreign technology.

What Does DeepSeek’s Success Mean for the AI Industry?

DeepSeek’s rise is a testament to how competitive the global AI industry is becoming. While American companies such as OpenAI and Meta have been dominating the field for the last couple of years, the success of DeepSeek proves that smaller companies can also make worthy contributions. DeepSeek challenges the belief that massive investments are required to develop high-quality AI models by developing high-quality AI models at lower costs.

But the success of DeepSeek raises other questions, too, about the future of AI development: Will big tech companies have to rethink their strategy if smaller firms can achieve similar results for a fraction of the cost? And how will governments respond to an increasingly competitive AI sector, especially when it comes to export controls and national security?

For now, DeepSeek focuses on further research and development. The company is working toward the development of AI systems that can outperform humans in a wide range of tasks, a concept known as Artificial General Intelligence. If DeepSeek can achieve this, it could change the way we think about AI and its potential to transform industries and improve lives.

DeepSeek is among the companies that are a force to be reckoned with in the AI space. The unique approach it has taken toward the development of AI, combined with cost efficiency, has already positioned it as one of the key players in the global tech arena. As this company continues to grow, perhaps it will also encourage other startups to follow in its footsteps and continue further with more advancements in the field of artificial intelligence.