China's research institute has urged local AI firms to stick with NVIDIA's AI solutions, given that using domestic alternatives would not be a viable replacement.
Low Supply Coupled Up With High Transfer Costs Makes Solutions From Huawei & Others Costly For AI Firms, Says CAICT, Asks Firms To Utilize AI Chips From NVIDIA
There is an ongoing debate on how China will move forward when it comes to acquiring the necessary AI computing power for domestic industries, and the growing influence of US sanctions has made it difficult for Chinese AI firms to rely solely on hardware from the likes of NVIDIA. While domestic solutions from Huawei have played their part in mitigating the problem, a new report by the China Academy of Information and Communications Technology (CAICT) claims that Chinese data centers shouldn't consider shifting completely to domestic products, given their limited capabilities.
If the conditions allow, [data centres] can choose [Nvidia’s] A100 and H100 high-performance computing units. If the need for computing power is limited, they can also choose H20 or alternative domestic solutions.
- CAICT via SCMP
Well, China has been a significant marketplace for NVIDIA, constituting more than 10% of the firm's YoY revenue despite being a region struck by US restrictions. The importance of Team Green's hardware is immense for Chinese AI firms like ByteDance and Tencent, given that with the large consumer base they have on hand, they require significant AI computing power onboard. CAICT has disclosed that GPU-based computing power for AI training and inference saw a YoY rise of 70% in the region, highlighting China's massive demand for AI hardware.
As a solution to the growing demands, companies like Huawei and BirenTech have presented their own AI chips to combat NVIDIA's influence in the markets, with Huawei, in particular, gaining massive sales traction, especially for its Ascend 910B AI GPU. The firm is also rumored to present the more advanced Ascend 910C as well but with domestic solutions, code porting and transferring existing LLMs built upon NVIDIA's compute stack is a difficult task to achieve, which the CAICT believes isn't worth the hustle.
Well, China can't certainly rely on NVIDIA. Given that the US administration has a knack for revising policies quite frequently and domestic AI solutions aren't working out too well, the country needs to resort to a concrete alternative. Sure, workarounds such as GPU renting services and getting AI hardware through "grey channels" are also present, yet they aren't sustainable in the long run.
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.
