India praises the Laboratory of the Chinese Depseek, plans to organize its models on local servers

The Minister of IT of India praised on Thursday the Progress of Deepseek and said that the country will be the host of the great language models of the Chinese Laboratory on national servants, in a rare openness for Chinese technology in India.

“You have seen what Deepseek has done: $ 5.5 million and a very powerful model,” said Minister Ashwini Vaishnaw on Thursday, responding to the criticisms that New Delhi has received for his own investment in AI, which has been much less than many that many Other countries.

Since 2020, India has banned more than 300 applications and services linked to China, including Tiktok and Wechat, citing national security concerns. The approval to allow Depseek to stay in India seems to depend on the platform storage and processing of all data from Indian users nationwide, in line with strict data location requirements of India.

“Data privacy problems with respect to Depseek can be addressed by hosting open source models on Indian servers,” Vaishnaw said at an industry conference.

Deepseek models will probably be housed in the new computer installation of India. The installation works with 18,693 graphics processing units (GPU), almost twice as its initial objective: almost 13,000 of them are GPU NVIDIA H100, and approximately 1,500 are GPU NVIDIA H200. Around 10,000 GPUs are ready to be deployed, and the installation is scheduled to start operations “in the next few days,” according to the minister.

The installation will also offer computer services with few discounts to India companies. Vaishnaw said that standard IA computer science would be offered with a 42% discount to market rates, and that high precision computer science would be discarded by 47%.

The minister’s comments arrive a day after Deepseek’s homonymous application was taken from Apple and Google application stores in Italy, after the data protection regulator of that country said he was asking how the Chinese company was using and storing the personal data of the Italians.

The launch of the Depseek “Reasoning” model, based on a supposedly modest budget, sent shock waves through the technology industry this week, which caused the market capitalization of the Nvidia chips giant to decrease by $ 600 billion. The model has quickly been under intense scrutiny and has caused heated debates around Copyright problemsUS export controls. And how even more money should be poured into the efforts of AI.

Beyond the organization of foreigners’ models, India is also trying to promote the development of AI models and related technology in their own grass. “The main chips designers are willing to work with India to develop indigenous GPUs,” said Vaishnaw.

Vaishnaw estimated that India would see an investment of $ 30 billion in hyperscalers and data centers in the next two or three years. One of the largest conglomerates in the country, Reliance, is to plan to build what could become the largest data center in the world in the city of Jamnagar, with a capacity of 3 gigawatts, Bloomberg reported last week.

“We believe there are at least six main developers who can develop AI models in six to eight months at the external limit, and four to six months in a more optimistic estimate. A common computing installation is the most important component to create a robust ia ecosystem, ”said Vaishnaw.

The computer installation will also support the broader initiatives of India. Vaishnaw said that 18 applications promoted by AI that focus on agriculture, climate change and learning disabilities have been selected for initial financing.

To supervise the development of these AI initiatives, India will establish a regulatory body using what Vaishnaw described as a “center and radio model”, which allows multiple institutions to collaborate in the security frameworks. “We will keep our models open and focused on the application,” he said.