news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

NVIDIA CEO Jen-Hsun Huang said they will start shipping engineering samples of Blackwell GPUs for AI applications around the world this week, and NVIDIA plans to bring Blackwell architecture products to market in the fourth quarter of this year.

This is what Huang Renxun said at Siggraph. In fact, NVIDIA hardware partners such as Foxconn, Quanta, Wistron, Pegatron and ASUS have already demonstrated Blackwell-based servers at this year's Taipei Computer Show. They may have had Blackwell GPUs for some time. However, not all software developers can get NVIDIA's latest Blackwell GPUs at present.

NVIDIA and its partners will charge a higher fee for each AI server cabinet equipped with Blackwell GPUs. The two reference server cabinets launched by NVIDIA are the NVL36 equipped with 36 B200 GPUs, which costs about US$2 million, and the NVL72 equipped with 72 B200 GPUs, which starts at US$3 million.

These cabinets will be provided by traditional partners such as NVIDIA, Foxconn, Quanta, and Wistron, as well as new entrants such as ASUS. NVIDIA is expected to ship 60,000 to 70,000 B200 server cabinets next year, which will bring in at least $210 billion in revenue. It is generally expected that major companies such as AWS, Dell, Google, Meta and Microsoft will adopt NVIDIA's latest Blackwell GPU.