Chinese AI firms are giving their all to come up with in-house solutions, as a new report claims that the tech giant Alibaba has developed a new AI chip to replace NVIDIA's hardware.
Alibaba's AI Chip is Claimed To Be Targeted Towards Inferencing Workloads, Possibly Utilizing SMIC's 7nm Process
Beijing has made it a priority to pivot away from America's AI tech stack, which is why the administration has been somewhat 'convincing' local tech giants to shift towards domestic solutions. Firms like Huawei, Cambricon, MetaX, and many others are increasing their investments in developing in-house chips, and now, according to a report by WSJ, Alibaba has developed its own AI chip, which is claimed to fill the 'NVIDIA void', and more importantly, relies completely on domestic technologies.
There is an essential element around China's push for local AI chips, and it is that the nation is far away from competing with the likes of NVIDIA right now. Not only are Chinese manufacturers barred from high-end semiconductors, but when it comes to the software stack as well, Team Green is the go-to option. However, one way that Beijing is currently focused on is by increasing 'inferencing capabilities' on domestic AI chips, which is why the Ascend chips from Huawei are claimed to be ideal for inferencing workloads, but for model training, they are no match.
Alibaba's AI chip is claimed to have been developed by a Chinese company, likely SMIC, and it is currently being tested on inferencing workloads. Given that SMIC currently has 7nm production lines up and running, it is likely that Alibaba has utilized the process node with its AI chip, which puts it on par with solutions from Huawei and Cambricon. Apart from all the optimism, China still faces a massive capacity issue, since its fabs cannot support high-volume production, mainly due to old equipment, such as DUV lithography, and inefficient techniques.
China still needs NVIDIA's AI chips for model training, and no options can replace the American tech stack when it comes to creating capable LLMs. More importantly, NVIDIA's CEO is pushing towards a 'Blackwell-based' product for China as well, showing that there's a dire need for NVIDIA to come up with a more capable solution for the region to stay relevant.
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.
