Huawei’s CloudMatrix 384 AI cluster has started to see adoption from Chinese tech giants, who are reportedly impressed by the server’s capabilities relative to NVIDIA’s offerings.
Huawei’s Latest CloudMatrix 384 AI Cluster Puts NVIDIA’s AI Dominance In Jeopardy, Especially In Regions Like China
Well, it seems like Huawei won’t stop when it comes to expanding its AI computing portfolio and capitalizing on NVIDIA’s vulnerable position in China, as the firm has stepped up its development process. We recently disclosed how Huawei had prepared the CloudMatrix 384 AI cluster, which is said to be entirely developed through in-house products, and gave performance that would likely give NVIDIA a tough time competing in the domestic market. Now, the Financial Times reports that ten different clients have adopted the CloudMatrix 384 AI server, signaling a massive breakthrough for Huawei’s AI ambitions.
It is claimed that Chinese tech giants have already integrated the AI cluster into their data center portfolio, and while the names of customers haven’t been disclosed yet, it has been revealed that they are primary customers of Huawei’s product offerings. We did cover CloudMatrix 384 extensively in the past, but for a quick summary, it is said that Huawei’s solution successfully competes with NVIDIA’s most powerful AI server, the GB200 NVL72, showing that China no longer depends on anyone else for its computational needs.
Regarding specifications, the CloudMatrix 384 (CM384) AI cluster features 384 Ascend 910C chips connected in an “all-to-all topology” configuration. Interestingly, Huawei has managed to cover the architectural flaws by housing five times as many Ascend chips as NVIDIA’s GB200. A CloudMatrix cluster is said to deliver 300 PetaFLOPS of BF16 computing, almost two times higher than GB200 NVL72. However, the only drawback here is the power CloudMatrix 384 is expected to consume, which is said to be 3.9x the power of a GB200 NVL72 and somewhat “awful” perf/watt figures across AI workloads.
Interesting, the price of a single CloudMatrix 384 AI cluster is said to be $8 million, which is almost three times higher than NVIDIA’s GB200 NVL72 configuration, so the key motive behind Huawei’s product is not to provide a cost-effective performance, rather a product that is made through in-house resources and is capable enough to compete with Western alternatives.
