New Delhi: When the whole world was busy preparing for the Lunar New Year, there was a stir in the AI world. China’s Hangzhou-based company DeepSeek has announced its new DeepSeek-R1 model, whose claims have not only shocked Silicon Valley, but have also caused the shares of companies like NVIDIA and TSMC to collapse.
What is the game of DeepSeek-R1?
According to the company, this model has beaten OpenAI’s GPT-4o in mathematics and coding, while it claims to be at par in programming. Surprisingly, its training cost is just 6 million dollars (10% of Meta’s Llama 3.1). Also, it has been made completely open-source and available for free. “This move directly challenges Silicon Valley’s ‘pay-to-use’ model,” explains tech expert Rajiv Mehra.
Will US sanctions work?
The success of DeepSeek-R1 has raised questions about the US AI chip ban. Some analysts believe that China has proved that sanctions cannot stop the pace of its AI dreams. “This is just the beginning,” says Beijing AI researcher Li Jiayao, “Our teams are working day and night.”
Stock market panic!
After this news, NVIDIA shares fell by 3%, and TSMC’s by 2.5%. But will this decline be permanent? Mumbai market expert Amit Shelar says, “There is nothing to panic now. The demand for AI chips will remain for a long time. The real competition will be in the era of Edge AI.”
Controversies are also not less!
Some experts are questioning DeepSeek’s claims. They say that it is wrong to say that the cost of this model using 50,000 GPUs (10,000 H100, 30,000 H20) is 5-10%. “This is just the calculation of training cost, infrastructure and electricity costs are hidden,” tweeted Toronto analyst Michael Chen.
What does it mean for India?
This development could also impact the Indian tech sector. Priya Reddy, CEO of Bengaluru-based AI startup NeuroBot, says, “Open-source models are a boon for companies like us. Now we will test DeepSeek-R1 in our healthcare projects.”
Is Silicon Valley’s silence broken?
Meanwhile, Meta has decided to keep the code of its Llama model open-source for now, but it is being said that they are trying to ‘reverse engineer’ the technology of DeepSeek. At the same time, OpenAI has hinted at a major update next month.
Finally: Is this a new chapter of ‘Dragon vs Eagle’ in the AI race? Only time will tell the answer, but one thing is certain – now ‘Made in China’ cannot be ignored in the world of technology.
