DeepSeek V3 is an AI model developed by Chinese artificial intelligence company DeepSeek Artificial Intelligence Co., Ltd. Founded in 2023 by Liang Wenfeng, co-founder of Chinese hedge fund Hi-Flyer, the company is gaining attention in the global AI scene by developing high-quality AI models at low cost.
The DeepSeek-R1 model they developed is very low cost, even when compared to models like OpenAI’s ChatGPT. While the DeepSeek-R1 model cost just $5.6 million to build, American labs have spent between $100 million and $1 billion on similar models.
This achievement of DeepSeek caused a huge drop in the stock values of American tech companies. Companies like Nvidia, Tesla, Google, Amazon, and Microsoft suffered huge losses in the market.
The secret to DeepSeek’s success lies in the new technical approaches they used. They made this possible using old generation GPUs. Computer GPUs are required to train AI models. DeepSeek found ways to increase the capacity of old GPUs using the cheap H800 GPU. In addition, using the multi-headed delayed attention technical design and the presence of young researchers, the model was ready with a state-of-the-art AI system at a low cost. It works faster than previous models.
The secret behind DeepSeek’s success lies in its innovative application of existing AI frameworks. They can develop more cost-effective models using open source technologies. This opens up new opportunities for small startups and research institutions to compete in the AI field.
DeepSeek’s achievement has had a huge impact on the global AI field. American tech stocks, especially companies like Nvidia, Tesla, Google, Amazon, and Microsoft, have seen their market value fall significantly. As DeepSeek’s models achieve high performance at low cost, the high-cost AI of big tech companies is being questioned.
DeepSeek’s breakthrough also demonstrates China’s AI capabilities. It is remarkable that Chinese companies have been able to develop advanced AI models with limited resources, despite US chip export restrictions.
This breakthrough by DeepSeek opens up new possibilities and discussions in the field of AI. The development of high-performance models at low cost is redefining the future of AI development. This opens up new opportunities for small startups and research institutions, while also questioning the existing models of large tech companies.
