Huawei unveils chipmaking, computing power plans for the first time

By Brenda Goh and Che Pan

Shanghaı (Reuters) – Huawei said on Thursday that the Ascend Ai chip would publish four new repeats in the next three years and break the privacy to raise the ambitions of chewing and the ambitions of competing against it. Nvidia first.

The Chinese technology giant became one of the leading players in developing a domestic semiconductor production industry, aiming to reduce the confidence in a supply chain dominated by the United States.

In the first quarter of the year, after the launch of the Ascend 910C, Vice President Eric XU said that the company’s succession of the company’s successor Ascend 950 was planned to start two variants in 2027 and follow the 960 version and 970 in 2028.

“The information process power has always been and even more important for the key to artificial intelligence and even China’s artificial intelligence.” He said.

The Ascend 950 chip will be strengthened by the company’s own special high bandwidth memory, he said that China has overcome an important bottleneck in technology in technology, and has been limited to South Korean and US suppliers for years.

Huawei also plans to offer new information processing supernodes, Atlas 950 and Atlas 960, defined by XU as the most powerful, 8,192 and 15,488 rise chips in the world.

Chips are the successors of Atlas 900, also known as Cloudmatrix 384, which uses 384 of Huawei’s latest 910C chips.

In some metrics, the Huawei product performs better than the GB200 NVL72, which uses NVIDIA’s 72 B200 chip.

Huawei says the system uses “Supernode” architecture that allows chips to be connected to each other at super high speeds.

(Reporting by Brenda Goh and Che Pan; Editing by Jacqueline Wong and Clarence Fernandez)