India’s GPU Acquisition from NVIDIA: A Deep Dive
A few years ago, a significant event marked the AI industry’s trajectory: NVIDIA’s CEO, Jensen Huang, personally delivered the first NVIDIA DGX AI supercomputer to OpenAI. This move set the stage for a broader question: How long will it take for Indian giants like Reliance or Tata to achieve a similar level of AI success as OpenAI?
OpenAI’s GPU Arsenal
OpenAI’s GPU collection is impressive. According to Huang Jensen, CEO of NVIDIA, the company has over 10,000 GPUs, which is nearly 40 AI supercomputers. Each of these supercomputers houses 256 GPUs. With this in mind, Huang estimates that India is on track to acquire tens of thousands of GPUs, approximating to less than 400 AI supercomputers.
India’s AI Future: A Vision by NVIDIA
Huang is bullish about India’s AI prospects. He believes that while companies like OpenAI take months to train GPT models, India could achieve comparable outcomes in mere weeks. He further revealed that India would soon receive the world’s fastest computers, which have yet to hit production. These machines, expected by next year, will be up to 100 times faster, drastically reducing the cost of training foundational AI models.
Accessible AI for Indian Enterprises
Reliance and Tata are set to benefit from DGX Cloud, a service by NVIDIA. This platform allows businesses to harness the power of an AI supercomputer via a standard web browser, eliminating the complexities of on-premises infrastructure. Huang emphasized the affordability of this service, noting that creating a large language model akin to ChatGPT could cost between $10-$20 million.
NVIDIA’s Offerings Accelerate AI Development
NVIDIA’s advanced GH200 Grace Hopper Super chip and the DGX Cloud are expected to fast-track AI development. However, it’s worth noting that OpenAI has chosen Microsoft’s Azure for training its models, including GPT-3 and GPT-3.5. This collaboration has integrated NVIDIA’s AI stack, adding a significant number of NVIDIA A100 and H100 GPUs.
India’s Competitive Advantage in AI
India’s diverse linguistic landscape offers a unique advantage in AI. Huang believes that India should focus on developing in-house LLMs and foundational models without exporting data. With the country’s rich talent pool of computer scientists and the upcoming NVIDIA supercomputers, India is well-equipped to lead in AI.
The Dawn of India’s AI Revolution
Partnerships between NVIDIA, Reliance, and Tata signify the beginning of India’s AI revolution. Reliance aims to develop AI applications for its vast Jio customer base, while Tata focuses on generative AI applications for enterprises. With NVIDIA’s infrastructure, these companies are poised to redefine the AI landscape in India.