India AI Summit | Rare, Expensive, and Powerful: NDTV’s First Look At the NVIDIA Chip Everyone Wants

The NVIDIA GPU Explained

• 💡 NVIDIA GPUs are central to global AI, with India currently possessing only 38,000 of these crucial units.
• 🔬 The featured system uses ARM-based CPUs designed by NVIDIA, specifically the Blackwell B200 GPUs, in a compact 1U size.
• 💧 It requires liquid cooling with a dielectric liquid that does not conduct electricity, preventing damage even if leaks occur.

Cost and Scarcity

• 💰 Each NVIDIA rack system costs approximately $5 to $5.5 million, making them exceptionally expensive.
• ⚠️ There is a significant shortage of these GPUs globally due to high demand for AI development and low production capacity.

AI Infrastructure and Power Consumption

• 🚀 These GPUs are integrated into rack systems, with the GB300 system accommodating 18 servers per rack.
• 🌐 Multiple racks can be interconnected with top-of-the-rack switches to form a supercomputer cluster for generative AI applications.
• ⚡ Each rack consumes an immense amount of electricity, equivalent to 50 to 60 households, to process data and fulfill user demands for creative solutions.

India's AI Ambition

• 🇮🇳 India's ability to scale artificial intelligence nationwide is heavily dependent on acquiring many more of these NVIDIA GPUs.
• 📈 While India currently has 38,000, the government is ordering another 10,000, highlighting the massive need for AI infrastructure.
• 🌱 There is a recognized need for these systems to become more energy-efficient and better tailored for the country's specific requirements.