Home » Technology » Artificial Intelligence » India to Host Chinese AI Lab DeepSeek’s Models Amid Push for AI Development

India to Host Chinese AI Lab DeepSeek’s Models Amid Push for AI Development

India’s IT minister, Ashwini Vaishnaw, on Thursday praised the progress of DeepSeek, a Chinese AI lab, and announced plans to host its large language models on domestic servers. This marks a rare instance of Chinese technology gaining approval in India, following years of restrictions.

DeepSeek’s Entry into India

DeepSeek’s advancements, including its $5.5 million R1 “reasoning” model, have gained global attention. Vaishnaw called the model “very, very powerful” during an industry conference, addressing concerns about India’s relatively modest investments in AI compared to other nations.

India has banned over 300 Chinese apps and services, such as TikTok and WeChat, since 2020, citing national security concerns. However, the approval to host DeepSeek’s models in India comes with strict conditions. All data from Indian users must be stored and processed domestically, adhering to India’s robust data localization laws.

“Data privacy issues regarding DeepSeek can be addressed by hosting open-source models on Indian servers,” Vaishnaw said.

AI Compute Facility to Power Hosting

DeepSeek’s models will be hosted on India’s newly launched AI Compute Facility, which boasts 18,693 GPUs, nearly twice its original target. This includes almost 13,000 Nvidia H100 GPUs and approximately 1,500 Nvidia H200 GPUs. Of these, 10,000 GPUs are ready for immediate deployment, and the facility is expected to begin operations soon.

The facility is also set to offer steep discounts on computing services. Standard AI computing will be available at 42% below market rates, while high-precision computing will be discounted by 47%. These measures aim to make AI resources more accessible to firms across the country.

Challenges and Global Scrutiny

The announcement comes at a time when DeepSeek is under scrutiny internationally. Its app was recently removed from Apple’s and Google’s app stores in Italy after the country’s data protection regulator raised concerns about how the company handles personal data. Additionally, DeepSeek’s R1 model has sparked debates around copyright issues, U.S. export controls, and the need for increased investment in AI.

Despite its modest budget, the release of the R1 model caused a sharp reaction in the tech world, leading to a $600 billion dip in Nvidia’s market cap.

India’s Broader AI Push

While hosting DeepSeek’s models is a significant step, India is also focusing on developing AI technologies domestically. Vaishnaw revealed that major chip designers are discussing partnerships with India to create indigenous GPUs. He projected investments of $30 billion in hyperscalers and data centers over the next two to three years. Reliance, one of India’s largest conglomerates, aims to build the world’s largest data center in Jamnagar, with a 3-gigawatt capacity, as reported by Bloomberg.

According to Vaishnaw, India has at least six developers who can create AI models within four to eight months. The AI Compute Facility will play a central role in supporting this ecosystem, alongside 18 AI-driven applications focusing on agriculture, climate change, and learning disabilities, which have been selected for initial funding.

Regulatory Measures and Collaboration

To oversee AI initiatives, India plans to establish a regulatory body using a “hub-and-spoke model” to encourage collaboration between institutions. “We will be keeping our models open and application-focused,” Vaishnaw said, emphasizing the importance of safety frameworks and ethical AI development.

 

As India ramps up its AI capabilities, hosting DeepSeek’s models and building its own AI ecosystem signal the country’s ambition to become a global player in AI innovation.

Leave a Reply