Chinese startup DeepSeek has made headlines in 2025 by significantly impacting the artificial intelligence landscape. In January, the company launched its R1 model, which not only took the App Store by storm but also led to an 18% drop in Nvidia's stock prices. With a training cost of $6 million compared to OpenAI's $100 million, DeepSeek showcased its potential to revolutionize the industry.
The startup has demonstrated that U.S. chip sanctions are not as effective as intended. Despite training its models on limited Nvidia H800 chips, the performance has been comparable to that of AI giants like GPT-4 and Claude. In December, DeepSeek released its latest model, V3.2, which benchmarks on par with GPT-5, further solidifying its position in the competitive landscape.
This year has seen a shift in the download patterns for open-source AI models, with Chinese versions now leading at 17% compared to 15.8% for American ones. The success of DeepSeek has sparked a race in the open-source arena, prompting major players like Alibaba, Moonshot AI, and even OpenAI to release their own open-source models.
In just one year, China has transformed from a follower into a formidable competitor in the AI space. This development signals a significant shift in the market dynamics, challenging established players and potentially reshaping the future of artificial intelligence technology.
Informational material. 18+.
