Nithin Kamath on DeepSeek: China’s AI surge highlights India’s research gaps

As China's AI model DeepSeek creates international buzz, Zerodha founder & CEO Nithin Kamath has said DeepSeek is the latest example of China's scientific and technological progress. Giving a historical background about both countries, Kamath said in the 1960s-1970s, India and China had roughly the same per capita GDP. "They (China) started their reforms in the 1980s, and by 1990, they had overtaken our per capita GDP. Say what you will about the differences in our worldviews and economic models, but their scientific and technological progress is undeniable across disciplines—DeepSeek is just the latest example."

Chinese hedge fund High-Flyer-led DeepSeek launched two of its latest models in the past month, topping the number of downloads on Apple's App Store. As the details about DeepSeek's extremely low cost and performance made global headlines, shares of the US chipmakers saw massive drops, wiping out more than $1 trillion from their market cap.

"I saw this chart in an Economist (link in next tweet) article. It gives you a quick and dirty idea of the progress China has made despite the fact that papers and citations can be gamed. India wouldn’t even show up on the chart," says Kamath.

He says India has always been plagued by "short-termism". "Problems are typically addressed through a patchwork or the Jugaad mentality. This is true when it comes to business, politics, regulatory approaches, etc. Many problems that require focused long-term thinking instead get band-aid fixes. This is not to say that we haven't made progress. In my lifetime, we have come a long way, but not nearly enough."

Kamath says one can’t just buy GPUs and expect Indians to create groundbreaking AI applications. "Without the right talent and an enabling ecosystem that facilitates innovation, all the GPUs in the world will be pointless."

Advocating for building research capabilities, Kamath says India does produce great researchers, but it does not seem to offer a conducive environment for them, which is why the majority of them go to the US. "Again, this is not something that will show results instantly. In the case of China, it was 2 decades at the bare minimum focusing just on research. If we start focusing on building our research and scientific capabilities, we will hopefully see results in 5 to 10 years."

He says that's all the more important in the world that AI will shape where being mediocre won't be an option.

Market sentiment on AI and big tech has turned negative following the initial success of China-based DeepSeek’s V3 and R1 large language models, which are reportedly trained at significantly lower training costs with deep discounts on inference. DeepSeek (DS) is 100% owned by a highly successful AI-driven quant fund in China, High-Flyer. High-Flyer created DS in April 2023 to focus on artificial general intelligence (AGI/LLM). V2 was launched in May 2024, with a reported cost of only Rmb2/output token. It achieved a No. 7 ranking on the University of Waterloo's LLM leaderboard. Last month, it launched V4, trained on a data set of 14.8tn tokens (13tn for GPT4-o), at a training cost of only $5.6 million.