DeepSeek hasn’t only caused the US stock market to crash and caused US President Donald Trump to say it’s a “wake up call” for the US AI industry, but it’s also got the Indian government to take notice.
India’s IT Minister Ashwini Vaishnaw has praised Chinese AI model DeepSeek. “We have seen what DeepSeek has done..(it’s a) powerful model. Very soon we will have our own LLM,” he said at the Make In Odisha Conclave.
Additionally, Vaishnaw said that the DeepSeek model will be available on Indian servers. “We will soon host DeepSeek on Indian servers,” he said. “It is an open-source model. Like LLama, which is open source, this too can be hosted on Indian servers. Data privacy issues regarding DeepSeek can be addressed by hosting open source models on Indian servers,” he added.
Vaishnaw also announced that India had empaneled 18,693 GPUs for the country’s AI mission against a target of 10,000 GPUs that it had set last year. “Our startups researchers, academia, they all want to access compute facility. That’s why in the India AI mission we have taken the common compute as the most important part of the mission,” he said. Vaishnaw said that India had 12,896 Nvidia H100 GPUs, and 1,480 are Nvidia H200s. These are the most advanced GPUs that NVIDIA makes.
These are important steps, but one wonders if it might be too little, too late. DeepSeek’s parent company is rumoured to have 50,000 GPUs, and DeepSeek has been releasing models since 2023. If an Indian lab were to start making LLMs now, it could be years before they could catch up to where cutting-edge Chinese and US models are today, by which time US and Chinese models will again be a generation ahead.
And that could have repercussions for the future of India on the global stage. Anthropic CEO Dario Amodei has called for export controls on chips to China, saying that if China were to have as many chips as the US, it would be an equal counterpart in AI, which could create a bipolar world with US and China as the two poles. India, sadly, doesn’t feature anywhere in the scheme of things — its private AI companies like Sarvam and Krutrim had started off with some fanfare, but have nothing meaningful to show for their efforts yet. India is lagging in the AI arms race, and it’ll take a war-time effort — with the participation of both the public and private sector — for it to have any hopes of catching up.