Huawei is reportedly moving towards making its Ascend AI chips an option for companies beyond China, thereby expanding the influence of Chinese technologies.
Huawei Hopes to Sell Its Highest-End Ascend 950 AI Chips To Korea, Expanding Its Influence to Global Markets
When it comes to Chinese manufacturers offering AI infrastructure, Huawei is the largest, not just in the size of the compute portfolio it provides, but also in terms of production capacity. The firm's Ascend AI chips are known to see the most widespread adoption among Chinese hyperscalers, especially at a time when Beijing had imposed restrictions on NVIDIA's AI hardware. According to the Korean media outlet YNA, Huawei plans to offer its Ascend portfolio to Korean customers after China, aiming to provide an alternative to NVIDIA.
Huawei Korea plans to officially launch AI computing cards and AI data center-related solutions next year. We want to provide Korean companies with a second option other than NVIDIA.
Huawei Korea's CEO, Balian Wang, disclosed the announcement at an official company event, stating that, unlike NVIDIA, the firm will offer only cluster configurations to its customers in Korea, which will include networking and storage infrastructure as well. Huawei claims that customers will receive entire "End-to-End" solutions, suggesting that the move targets clients who cannot access NVIDIA's hardware and need an alternative.
Interestingly, Huawei is expected to offer its most capable chip, the Ascend 950, to customers in Korea. For those unaware, these models are available in two primary configurations: the 950PR and the 950DT. Both of these chips are known to have Huawei's self-built HBM integrated, and the Chinese tech giant is most likely to offer cluster-based configurations, which are the Atlas SuperPod models. These clusters are claimed to rival NVIDIA's Rubin lineup, according to information disclosed by Huawei's official announcements, which means that Korean customers are getting cutting-edge technology from the Chinese giant.
| Feature | Ascend 950PR | Ascend 950DT |
|---|---|---|
| Target Workloads | Inference & recommendation | Decode & training |
| HBM Type | HiBL 1.0 | HiZQ 2.0 |
| HBM Capacity | ~128 GB | ~144 GB |
| HBM Bandwidth | ~1.6 TB/s | ~4 TB/s |
| Interconnect Bandwidth | ~2 TB/s | ~2 TB/s |
| Compute Orientation | Prefill/Inference | Training + Decode |
NVIDIA's CEO has repeatedly discussed how firms like Huawei plan to aggressively expand their influence in the global market through a strategy similar to the one the company employed with 5G networking technology. Jensen calls it China's "Belt & Road" initiative, but with a focus on AI, and it appears that this is unfolding. The only concern that remains for the Chinese tech giant is whether it has sufficient production capacity on board to fulfill demand from global markets.
News Source: Jukan
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
