DeepSeek AI entered the market with massive efficiency, Nvidia shares dipped marking the lowest of 2025. Today marked a technological leap that shocked Silicon Valley. China unveiled a newly built open-source AI model that beats some of the most powerful precise AI model’s in the market.
Emerge of DeepSeek AI
AI model build in DeepSeek AI lab appears to be one of the best open challenges yet. One of the major shocking things behind the development is its cost. It took only $5.576 M in training costs to be in action as compared to Open AI which took 5 billion dollars per year and Google made 51 billion dollars. Microsoft who invested 13 billion dollars in Open AI is curious as well. What is even more stunning is how the DeepSeek AI model outperforms all of these. It is shocking to see how the supermodel does inference time and how it computes more efficiently.
DeepSeek AI
DeepSeek AI model outstands all other GPTs in solving Maths computation as well as coding fixing bugs and many other parameters. It uses the Reasoning model which leads all of the other models. Founder and CEO of Scale AI Alexander Wang told that they have done this under Humanities last exam which involves training the model with all the subjects in DeepSeek AI research lab. Paper was released yesterday.
Hardships behind DeepSeek AI model
This is done by the Chinese lab despite of the restrictions led by the US government to curb the export supply of semiconductor chips which cuts off China from getting the most effective chips to do computing. Now due to this advancement, China proved that Nvidia’s chips are no longer required to make competitive AI models as was believed by all Western world. That is the reason Nvidia shares dipped marking the lowest of 2025. Deepseek -V3 requires only 2.788M H800 GPU hours for its full training. It proved that Chip control was not the chokehold Washington D.C. intended.
Related news
Very little is known about the lab and founder Liang Wenfeng. According to the Chinese news channel, Deepseek AI was made by young geniuses of a low-key team. It is a threat to the world amid this Face-off as China accurately entered the market. Along with that, some other Chinese startup carved their way up amid low resources.
One of them is the startup of Kai-Fu Lee who builds a $1 billion startup in 8 months. Alibaba is also in a race the cut costs by 85 percent. Google former Chairman Eric who stated that China is 3-4 years behind the USA Has now changed his statement that China caught up in 6 months aggressively and proved a better approach to the world.
Now that open source has built upscaled AIs the companies need not to make their’s from scratch but make over the existing ones more significantly with smaller budget and smaller teams. It tells us that any company that is in the race today may not be in race tomorrow. Now the companies can do distillation by making small models needed from the large models.