Key Information
- Major Chinese technology companies like ByteDance and Alibaba are planning to adopt Huawei’s new artificial intelligence chip, the 950PR.
- Enhanced compatibility with Nvidia software and improved performance make the 950PR a competitive option for Chinese companies.
- Huawei’s focus on inference workloads reflects the increasing demand for practical AI applications in the Chinese tech sector.
Huawei’s latest artificial intelligence chip, the 950PR, has garnered significant interest from major Chinese technology companies like ByteDance and Alibaba, who are expected to place orders after successful customer tests, according to Reuters. This development represents a crucial advancement for Huawei, which had struggled to convince industry giants to widely adopt its Ascend 910C chip, despite government efforts to promote the use of domestic semiconductors.
Enhanced Compatibility and Improved Performance
Enhanced compatibility with Nvidia’s widely used CUDA software system, as well as improved response speeds, has alleviated previous concerns and made the 950PR a more attractive option for Chinese tech companies. Huawei plans to ship around 750,000 units of the 950PR this year, with mass production set to begin next month.
The launch of Huawei 950PR coincides with challenges for Nvidia in the Chinese market. The US has imposed restrictions on the sale of many Nvidia AI chips to China due to national security concerns. However, a recent development is the H200 chip, more powerful than previously restricted models, which received authorization for sale last year under certain conditions. Nevertheless, its availability in China remains uncertain as regulatory approvals are still pending.
Focus on Practical Applications
Huawei’s 950PR, priced at around 50,000 yuan (6,300 euros) for the standard version and 70,000 yuan for the high-end variant with HBM memory, marks a turning point. It transitions from Huawei’s proprietary CANN software system to a more open system, allowing for easier migration of existing models developed with Nvidia software. While the raw computing power gain is modest compared to the 910C, the 950PR is optimized for inference workloads, which are crucial for the practical deployment of AI models.
This focus on inference aligns with a clear trend, meeting the growing demand in the Chinese tech sector for concrete AI applications. This evolution is driven by the rapid adoption of open-source AI agents like OpenClaw.
(jw)(fc)
Follow Business AM on Google News
If you want to access all articles, subscribe here!
