Huawei’s Bold AI Bet Aims to Fill Nvidia’s Void in China
00:00/00:00
Listen to this article
1x
Subscribe now
Listen to the full version
Listen to the full version
Huawei first launched its Ascend series in 2018
Facing tightening U.S. sanctions that bar access to advanced semiconductors, Chinese tech behemoth Huawei Technologies Co. Ltd. is spearheading a pragmatic, large-scale strategy to deliver the artificial intelligence computing capacity China urgently needs.
Once a major customer of global chipmakers, Huawei has now recast itself as a linchpin in China's push for self-reliance in AI. Last month, rotating Chairman Eric Xu Zhijun announced a multiyear plan to launch four successive generations of its Ascend AI chips, with each iteration aimed at nearly doubling computing performance year over year.
You've accessed an article available only to subscribers
VIEW OPTIONS
Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.
Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.
- RELATED
Share this article
Open WeChat and scan the QR code
- DIGEST HUB
Digest Hub
Back
Explore the story in 30 seconds
- Huawei is leading China’s push for AI self-reliance amid U.S. sanctions, focusing on developing four generations of Ascend AI chips and large-scale cluster-based computing systems.
- Its Atlas 900 cluster uses 384 Ascend 910C chips and, though each chip lags behind Nvidia’s, the system can deliver nearly double Nvidia’s AI processing throughput at higher energy cost.
- Huawei faces challenges in AI software, open-sourcing its CANN platform as an alternative to Nvidia’s CUDA, while rivals pursue CUDA compatibility for easier adoption.
AI generated, for reference only
Explore the story in 3 minutes
[para. 1]
AI generated, for reference only
Who’s Who
- Huawei Technologies Co. Ltd.
- Huawei Technologies Co. Ltd., a Chinese tech giant, is leading China's AI self-reliance efforts amidst US sanctions. They are developing Ascend AI chips and building cluster-based supercomputers to challenge Nvidia. Huawei's strategy focuses on system-scale engineering to overcome chip restrictions, planning to deliver progressively larger systems to support AI development.
- Nvidia Corp.
- Nvidia Corp. once dominated China's AI chip market, controlling approximately 95%. However, U.S. export restrictions severely impacted its supply to China, creating a critical opening for domestic players like Huawei. While Huawei's chips may consume more power, they are designing cluster-based systems to deliver competitive AI processing throughput.
- Moore Threads Technology Co. Ltd.
- Moore Threads Technology Co. Ltd. is a Chinese startup, founded by veterans from Nvidia and AMD. It is actively developing its own alternatives to Nvidia's AI chips, aiming to fill the vacuum created by US export restrictions. The company is also seeking an initial public offering on Shanghai's STAR market.
- MetaX Integrated Circuits Co. Ltd.
- MetaX Integrated Circuits Co. Ltd. is a Chinese startup, founded by veterans from Nvidia and AMD. Following U.S. export restrictions that impacted Nvidia's supply in China, MetaX is actively developing its own AI chip alternatives. The company is part of a broader trend of indigenous hardware development in China, aiming to address the void left by foreign chipmakers and potentially pursue an initial public offering on Shanghai's STAR market.
- Baidu Inc.
- Baidu Inc. is a major Chinese firm that is intensifying its in-house chip development efforts. This initiative aligns with China's broader tech industry shift towards indigenous hardware and self-reliance in artificial intelligence, especially as U.S. export restrictions impact the supply of foreign AI chips. Baidu-backed Kunlun Tech Co. Ltd. is also introducing CUDA-compatible chips.
- Alibaba Group Holding Ltd.
- Alibaba Group Holding Ltd. is mentioned as a major firm rapidly increasing its efforts in developing in-house chips. This move is part of a broader trend within China's tech industry to pivot towards indigenous hardware, driven by U.S. export restrictions that have impacted the supply of foreign AI chips.
- Taiwan Semiconductor Manufacturing Co. Ltd.
- Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) was barred from making chips for Huawei in 2020 due to U.S. sanctions. This restriction impacted Huawei's access to advanced chipmaking technology, as TSMC had a lead in manufacturing technology that Chinese mainland contract chipmakers lacked.
- ASML Holding NV
- ASML Holding NV is a company that produces state-of-the-art extreme ultraviolet lithography equipment. Due to sanctions, Chinese mainland contract chipmakers cannot buy this equipment, which has hindered their manufacturing technology. This restriction affects Huawei's ability to access advanced chipmaking nodes.
- Cambricon Technologies Corp. Ltd.
- Cambricon Technologies Corp. Ltd. is a Chinese company that develops chips. Along with Alibaba's T-Head unit and Baidu-backed Kunlun Tech Co. Ltd., it's a domestic rival to Huawei in the AI chip market. These companies are introducing CUDA-compatible chips, which makes it easier for developers to adopt them.
- Alibaba’s T-Head unit
- Alibaba's T-Head unit is one of the domestic rivals introducing CUDA-compatible chips. This move aims to simplify developer adoption within China's AI ecosystem. Alongside other companies like Cambricon Technologies Corp. Ltd. and Kunlun Tech Co. Ltd., T-Head is contributing to the competitive landscape as China seeks self-reliance in AI hardware.
- Kunlun Tech Co. Ltd.
- Kunlun Tech Co. Ltd. is a Baidu-backed company mentioned in the article. It is developing CUDA-compatible chips, which makes it easier for developers to adopt their technology. This positions Kunlun Tech as a competitor to Huawei and other domestic rivals in China's AI chip market.
AI generated, for reference only
What Happened When
- 2018：
- Huawei launched its Ascend series and the Ascend 910 chip outperformed then-mainstream Nvidia GPUs.
- 2020：
- U.S. sanctions barred TSMC from making chips for Huawei.
- 2025：
- Huawei introduced the Atlas 900 super-node, comprising 384 Ascend 910C chips.
- 2025：
- Atlas 900 system already deployed by Huawei, delivering 300 petaflops.
- August 2025：
- Huawei open-sourced the CANN ecosystem for outside developers.
- As of September 2025：
- Huawei's rotating Chairman Eric Xu announced a multiyear plan for four successive generations of Ascend AI chips.
AI generated, for reference only
- PODCAST
- MOST POPULAR
Sign in with Google
Sign in with Facebook
Sign in with 财新