India’s AI Vision

Union IT Minister Ashwini Vaishnaw said that India will develop multiple sovereign AI foundation models within 8 to 10 months. This comes as China’s DeepSeek AI model gains global attention.

Governments AI Strategy GPU Procurement

• India AI Mission
  • Approved March, 2024
  • Initially aimed to procure 10,000 GPUs now expanded to 18,693 GPUs.
  • 10 companies such as Reliance Jio Tata Communications and Yotta Data Services will provide the startups researchers and academia with GPUs.
• Cost Efficiency
  • The rate of a GPU is set at Rs.115.85 per hour for low end and Rs.150 per hour for high end. This is significantly cheaper than the global benchmark of $2.5–$3 per GPU hour.
• Financial Support
  • Direct central funding equity based funding to encourage private investment.
  • 40% discount for startups, researchers and educational institutions.
  • Long term users 6 months or more may get additional discounts.
• Competing with DeepSeek Global AI Leaders
  • Chinas DeepSeek recently built an AI model at a fraction of global costs raising concerns.
  • India will host DeepSeek on its servers and thereby mitigate risks associated with privacy and data security.
• Indias AI Models vs Global Benchmarks
  • DeepSeek was trained on 2,000 GPUs.
  • ChatGPT4 was trained on 25,000 GPUs.
  • The procurement of 18,693 GPUs by India is a strong foundation for competing in AI development.
• Evaluation Criteria for AI Model Developers
  • Companies will be evaluated based on
    • Technical ability and methodology.
    • Applicability of the use cases to India agriculture education climate change etc.
• Ethical implications and scalability
  • Financial feasibility and sustainability of the models.
• Indigenously Developing GPU AI related Applications
  • India is actively developing indigenous GPUs to reduce reliance on foreign technology.
• 18 AI applications have been selected for funding focusing on:
  • Agriculture
  • Learning disability solutions
  • Climate change mitigation

DeepSeek AI