Huawei continues to introduce new solutions on the AI battleground, and in the current landscape, the company has launched the ATLAS 350 AI card. The new creation is capable of phasing out powerful Nvidia chips in terms of performance and efficiency.
AI accelerator card refers to a specific hardware component designed to speed up AI operations, machine learning, and deep learning neural network workloads.
In the latest edition, Huawei introduced the Atlas 350 AI card that claims to deliver 1.56 petaflops of FP4 computing power. FP4 is low-precision in computing, which enables the AI card to transfer data faster than all of the previous versions.
That’s nearly 2.8-times improvement over the Nvidia H20 chipset.
The company unveiled the new AI solution at the China Partner Event on March 20. It runs on an Ascend 950PR processor and promises enhanced computing power + storage capabilities for AI inferencing.
Huawei launched the Ascend 950PR last year in September as a key step in AI model inference that ensures an efficient processing of all input tokens. Perhaps it has now made perfect use of this chipset.
Details further suggest that the new Atlas AI card targets growth in AI inference operations regarding search recommendation, multimodal generation, and LLMs. With the new solution, the OEM aims to match or even exceed its industry peers.
Huawei continues to enhance its computing hardware and storage tech solutions. It plans to unveil a FusionCybe A1000 cabinet for deploying AI quickly in small and medium-sized enterprises.
The President of Huawei Data Storage Product Line, Yuan, said:
“While the first half of the AI era focused on computing power, the second half will be defined by data. In 2026, Huawei will continue to upgrade its storage product lines and will closely work with major Chinese data infrastructure projects.”
(Image Credits: Huawei)
The post Huawei Atlas 350 AI card debuts, outshining Nvidia H20 chip appeared first on Huawei Central.
