Nvidia Announces New AI Server With 10x Performance

Nvidia announces new AI server with 10x performance, 72 integrated chips and MoE architecture, leading the global artificial intelligence field.

Nvidia on December 3 released new data showing that its new artificial intelligence (AI) server is capable of improving the performance of advanced AI models, including popular models from China, by up to 10 times compared to the previous generation of servers.

The data comes as the AI ​​industry shifts its focus from model training, an area where Nvidia currently leads, to deploying models to millions of users, an area that is seeing increased competition from rivals like Advanced Micro Devices (AMD) and Cerebras.

Nvidia says these improvements come largely from its ability to pack a large number of chips into a single server and the high-speed connections between them, an area where it still leads and maintains a clear advantage over its rivals. Nvidia’s latest AI server is equipped with 72 of its top-of-the-line chips.

The data Nvidia released focuses primarily on AI models that use Mixture-of-Experts (MoE) architecture, a method that optimizes the performance of AI models by dividing tasks into separate parts and assigning them to different "experts" within the model for processing.

The MoE architecture gained popularity in 2025, especially after DeepSeek, a Chinese AI company, introduced a high-performance open-source model that required less training time on Nvidia chips than other competitors.

Since then, major companies like OpenAI (the maker of ChatGPT), France’s Mistral, and China’s Moonshot AI have begun applying the MoE method to their models. Moonshot AI released a highly regarded open-source model this past July that uses the technique.

While Nvidia continues to maintain an advantage in AI model deployment, rivals like AMD are also working hard to develop competing products.

AMD is expected to launch a similar AI server next year, integrating many powerful chips with the aim of directly competing with Nvidia's servers in the field of inference (model processing and deployment)./.

(TTXVN/Vietnam+)

Source: https://www.vietnamplus.vn/nvidia-cong-bo-may-chu-ai-moi-co-hieu-suat-cao-gap-10-lan-post1080980.vnp