Huida delivers the world’s first DGX H200 to OpenAI, Hon Hai is expected to benefit | Industry Hotspots | Industry

Foreign reports pointed out that OpenAI president and co-founder Greg Brockman revealed that Nvidia handed over the world’s first DGX H200 artificial intelligence (AI) server to the company. With Nvidia’s new generation AI servers are being delivered one after another, and the main supplier Hon Hai (2317) will benefit the most, and will drive a substantial growth in the performance of its cloud network department.

In fact, Hon Hai has been cooperating with Huida to develop servers as early as 2017, when it launched the world’s first AI server HGX1. When OpenAI was founded, there was a photo showing Huang Renxun, founder and CEO of Huida. He gave Musk an AI server as a gift and signed it. The server at that time was the first-generation AI server jointly developed by Hon Hai and Huida. It can be seen that Hon Hai Group, Huida and Open AI have cooperated for a long time and are very close.

Brockman posted a tweet, showing a photo of Sam Altman, Jen-Hsun Huang, and Brockman, and in front of them was the first delivered DGX H200.

DGX H200 is designed to bypass the limitations of traditional standard connections such as InfiniBand (the interconnection standard between servers and storage) and Ethernet by using Huida’s customized NVLink Switch (exclusively manufactured by Hon Hai) , providing maximum throughput for maximum workloads.

OpenAI’s chatbot ChatGPT sparked an AI craze, and a global rush to pursue AI products and investments has led to a growing demand for generative AI computing in large enterprises and cloud data centers, which in turn has led to a new wave of supercomputers and AI servers. device needs.

Hon Hai Chairman Liu Yangwei also said before that the AI ​​factory (Factory) jointly developed by Hon Hai and Huida will be seen soon, and there will also be smart manufacturing, smart electric vehicles, etc. These three major platforms should see progress in the first half of the year. . The legal person believes that as the application of generative AI continues to expand, it is expected to drive Hon Hai’s shipments of Huida’s new generation servers to gradually increase.

Huida released a new generation of DGX GH200 AI supercomputer in May last year. It is the first supercomputer to use Grace Hopper super chip with NVLink Switch.

It connects 256 Grace Hopper superchips into an ultra-large, 1-Exaflop, 144TB GPU giant model used to drive generative AI, recommendation systems and data analysis. At this year’s Huida GTC conference, the brand-new GB200 AI server was further released.

Huang Renxun said: “Generative AI, large-scale language models and recommendation systems are the engines that drive the development of the modern economy. The DGX GH200 supercomputer integrates Huida’s most advanced accelerated computing and network technologies to expand the development of AI.” .