Huawei Technologies has launched a new artificial intelligence (AI) architecture that is reportedly set to rival Nvidia’s offerings in addressing computing power bottlenecks. The CloudMatrix 384 Supernode was described as a “nuclear-level product” that matched Nvidia’s NVL72 system in alleviating computing bottlenecks for AI data centers. Nvidia’s NVL72, launched in March last year, features a 72-graphics processing unit (GPU) NVLink domain that functions as a single, powerful GPU, enabling real-time inference for trillion-parameter large language models (LLMs) at speeds 30 times faster than previous generations. NVLink is a high-speed interconnect technology developed by Nvidia that allows multiple GPUs to communicate with each other and share data more efficiently. Huawei’s new supernode, currently deployed in the company’s data centers in Wuhu in Anhui province, achieved 300 petaflops of computing power, compared with the 180 petaflops offered by Nvidia’s NVL72, the report said, citing data from Huawei.
Supernodes are AI infrastructure architectures equipped with more resources – such as central and neural processing units, network bandwidth, storage and memory – than standard counterparts. This allows them to function as relay servers, enhancing the overall computing performance of clusters and significantly speeding up the training of foundational models. If confirmed, Huawei’s advancement highlights its progress in achieving self-sufficiency in computing power amid heightened U.S.-China tech tensions.
Chinese short video firm Kuaishou Technology unveiled an upgraded version of its Kling AI 2.0 video generation artificial AI model, touting it as “the world’s most powerful” as competition intensifies in the race to develop video-generating models. Kling now has more than 22 million global users who have generated over 168 million video clips and 344 million images, according to Kuaishou Senior Vice President Gai Ku. Kuaishou is among a group of Chinese Big Tech firms rushing to claim new advances in AI tools capable of producing video, following developments from TikTok owner ByteDance and Alibaba Group Holding. Kuaishou's previous-generation models are currently ranked as the world’s top image-to-video model and the second-best text-to-video model, trailing only Google’s Veo 2, according to Artificial Analysis, a third-party service that tests and ranks global AI models. Kuaishou rolled out the first version of Kling AI in mid-2024, making it one of the first Chinese tech firms to introduce a competitor to OpenAI’s Sora video model, which was previewed in February 2024, the South China Morning Post reports.
Huawei Technologies also unveiled an upgraded version of its artificial intelligence (AI)-powered eyeglasses, ratcheting up competition in the already crowded market for wearable hardware. Priced at CNY2,299, the new eyewear comes with a sleek titanium frame and enhanced AI-powered capabilities to handle a range of tasks from gesture control and voice commands to simultaneous translation. Huawei’s latest AI eyewear offers up to 11 hours of music playback or nine hours of talk time. It also provides continuous “intelligent announcements”, with updates such as weather forecasts and calendar reminders.
The AI glasses sector is poised to grow exponentially, as players from Big Tech firms to start-ups take advantage of breakthroughs in generative AI to enhance offerings in the segment. Market consultancy IDC estimated that China’s total AI glasses shipments will reach 2.75 million this year. Separately, Wellsenn XR forecasts that global AI eyewear shipments would surpass 80 million by 2030.
