Urgent.News

What's breaking now, across thousands of outlets.

AI

DeepSeek, Huawei's 'China AI Rise'... '脱Nvidia' dilemma is a challenge

Translated from Korean Read in Korean

As the heightened US export controls on advanced Chinese semiconductors tighten following the "Jensen Huang scandal," China is increasingly focused on finding alternative methods to power AI without relying on Nvidia's chips. Instead of relying on a single high-performance chip, China aims to interconnect thousands of domestic chips and reduce the computational load of AI models. Huawei leads this 'workaround innovation,' with DeepSense leading the software side.

Huawei is developing an 'Atala 950 Super Node' by connecting its AI semiconductor 'Ascend' in large numbers. The concept involves grouping up to 8192 chips into a single computing network. The strategy starts with increasing chip-to-chip communication speed to compensate for individual semiconductor weaknesses on a system level. The dependence on Nvidia's software development environment, CUDA, is rapidly decreasing as well.

Recent attempts by Huawei and Ascend to optimize the operating and communication functions of a system with 128 Ascend chips connected have been published. IDC market research data shows that in the first half of this year, Nvidia's share in China's AI acceleration chip shipments reached 34%, while Huawei's share was 28%. Alibaba's Tenshaid (7%) and ByteDance's Kunrenxin (4%) have also entered the market.

However, the market share does not necessarily indicate operational performance or revenue competitiveness. A critical issue is that despite the growth of domestic chips, replacing Nvidia is not easy, as connecting more chips increases power consumption and cooling and communication costs. Additionally, the software ecosystem shortage is a significant challenge to overcome for sustainable performance in large-scale services over the long term.

According to Reuters, as of last month, China's AI operations based on Nvidia chips alone have exceeded 23 billion won (approximately 46 trillion won). Recently, with Jensen Huang, the former CEO of Nvidia, facing personal issues and stepping down, the US is likely to further tighten its export controls on advanced products. This is driving Chinese companies to seek alternatives to ensure access to Nvidia's computational capabilities in foreign data centers.

In the meantime, the US is pursuing a strategy to block foreign servers from remotely using China's advanced chips. As China speeds up its domestic technology development, the 'break from Nvidia' challenge becomes increasingly apparent.

Written by urgent.news from Hankyoreh's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

This story

This is one outlet's version. Read the fullest account.

Every outlet that reported this, and when →

Read the original at hani.co.kr →

More in AI

More from Monday 16 November →