Source: Photo.com
On April 24, local time, Huang personally delivered the world's first DGX H200 to OpenAI. Huang said this will "advance AI, computing and human civilization." ”
OpenAI President and co-founder Greg Brockman made the announcement on the X platform with a photo with Jensen Huang and OpenAI CEO Sam Altman.
The DGX H200 is NVIDIA's latest and most advanced mass-produced AI chip, which is currently the world's most powerful AI-specific chip hardware. This delivery is a pivotal moment for OpenAI, a leader in AI research.
Compared to the previous generation H100, the H200 has a significant increase in performance. Major upgrades include a 1.4x increase in memory bandwidth and a 1.8x increase in memory capacity for a total memory bandwidth of 4.8TB/s and a memory capacity of 141GB.
For OpenAI, acquiring the DGX H200 is an important strategic move that will enhance their AI research capabilities, especially to drive the development of the highly anticipated GPT-5 large model. The H200's increased capabilities will allow OpenAI to push the limits of what its AI models can currently achieve, especially in terms of the speed and complexity of data processing.
AI chips are the mainstream mode of intelligent computing
Accelerated computing based on AI chips is the mainstream mode of intelligent computing. AI chips are co-designed with AI algorithms to meet the ultra-high demand for computing power in AI computing. At present, mainstream AI accelerated computing mainly uses CPU systems equipped with heterogeneous acceleration chips such as GPUs, FPGAs, and ASICs.
The AI chip industry is still in the development stage
The underlying hardware AI chip, the core of artificial intelligence, has experienced many ups and downs and twists and turns, and has been continuously upgraded and improved since its birth. In 2006, Hinton published a paper on "Multi-layer Neural Networks", which proved the possibility of large-scale deep learning neural network learning, and AI chips have developed rapidly since then.
In 2014, Dr. Chen's research team released a series of DianNao papers, opening the field of ASIC chip research. With the rapid development of deep learning algorithms, various application fields have higher and higher requirements for computing power, and the traditional chip architecture cannot meet the demand for computing power for deep learning, and artificial intelligence chips that can accelerate computing processing have entered a stage of rapid development.
AI chips are the new direction
In 2019, the global AI chip market size was $11 billion. With the increasing maturity of artificial intelligence technology and the continuous improvement of digital infrastructure, the commercial application of artificial intelligence will accelerate the landing, promote the rapid growth of the AI chip market, and the global artificial intelligence chip market is expected to reach 72.6 billion US dollars in 2025.
Jacob Bourne, an analyst at Insider Intelligence, said Nvidia's dominance in AI has been consolidated, but competitors such as AMD, Intel, and some startups and big tech companies could erode Nvidia's market share, especially among cost-conscious enterprise customers.
Prospective Economist APP Information Group
For more research and analysis of this industry, please refer to the "Global Artificial Intelligence Chip (AI Chip) Industry Market Research and Development Prospect Research Report" by Qianzhan Industry Research Institute
At the same time, the Prospective Industry Research Institute also provides solutions such as industrial big data, industrial research reports, industrial planning, park planning, industrial investment, industrial mapping, smart investment promotion system, industry status certificate, IPO consulting/fundraising and investment feasibility study, and specialized and special new small giant declaration. To quote the content of this article in any public information disclosure such as prospectus and annual report, formal authorization from Qianzhan Industry Research Institute is required.
More in-depth industry analysis is available in the [Prospective Economist APP], and you can also communicate and interact with 500+ economists/senior industry researchers. More enterprise data, enterprise information, and enterprise development are all in the [Qichamao APP], the most cost-effective and most comprehensive enterprise query platform.