Buzz
The NVIDIA H100 chip boasting the most powerful graphics processing capability today is one of the top-priced chips globally, yet not easily accessible to everyone.
In a breakthrough move, NVIDIA has introduced the world's most powerful GPU chip named NVIDIA H100. However, the production quantity of H100 is quite limited, leading to it quickly becoming a scarce phenomenon with a price tag of up to $40,000 (almost a billion dong). This price point may prompt many to wonder if peak performance is worth the significant amount they have to invest.
NVIDIA CEO - Jensen Huang, has hailed the NVIDIA H100 as the 'world's first system designed for super AI'NVIDIA CEO - Jensen Huang, has hailed the NVIDIA H100 as the 'world's first system designed for super AI'. With three times the power of its predecessor A100 and only a 1.5-2 times increase in operating costs, the H100 marks a significant step forward in the AI industry. In AI training, GPUs can process batch computations simultaneously, giving them an advantage over CPUs.
This makes high-performance GPUs like the H100 vital tools in building AI chatbot applications, which are at the core of AI. But why has the H100 become a scarce phenomenon in the market? The sudden AI explosion is one of the main reasons. Currently, only NVIDIA can produce powerful GPUs like the H100, and with the increasing demand for GPUs for AI applications, the supply cannot meet current needs.
With three times the power of its predecessor A100 and operating costs only increasing by 1.5-2 timesAI-focused entities are racing to secure H100s. OpenAI - the parent company of ChatGPT needs tens of thousands of chips to operate GPT-5. Inflection AI - a startup project backed by Microsoft and NVIDIA, has purchased 22,000 H100 GPUs to build its own AI center. Even Meta, Google Cloud Platform, Amazon Web Service, and Microsoft Azure are vying to acquire tens of thousands of H100 GPUs for their projects.
Even small tech firms have placed orders for GPU H100s. King Abdullah University of Science and Technology in Saudi Arabia is a prime example, having ordered 3,000 H100s worth $120 million. According to experts' estimates, global sales of H100s this year could exceed 430,000 units with a total value of around $15 billion. However, the Financial Times predicts NVIDIA could produce 550,000 H100s this year and they are already sold out.
King Abdullah University of Science and Technology in Saudi Arabia is a prime example, having ordered 3,000 H100s worth $120 millionElon Musk has even described sourcing GPUs as 'harder than finding addictive substances'. In a tech-disruptive era, continuing to provide powerful and efficient GPUs is a challenge for NVIDIA and the GPU industry as a whole.
The importance of graphics chips in artificial intelligence development is undeniable. Graphics chips, like Nvidia H100, provide parallel processing capabilities, significantly enhancing the efficiency of AI model training compared to CPUs. This not only helps create more powerful AI applications but also reduces the time and resources needed.
In AI training, GPUs excel at parallel processing, giving them an advantage over CPUs.With the explosion of artificial intelligence, the demand for computing power is increasing, and graphics chips are indispensable choices to meet these challenges. They play a crucial role in developing advanced AI applications such as intelligent chatbots, complex data analysis, and many other applications, driving the advancement of artificial intelligence for life and industry.
Read more:
- Will Apple and Intel utilize TSMC's brand-new 2nm process? Is it 'too little too late'?
- The U.S. 'choke off' China in the semiconductor manufacturing industry
- Explore section article
