Huawei Begins Delivering CloudMatrix 384 AI Clusters to Chinese Customers; Costs Triple That of NVIDIA’s GB200 NVL72 System

Whoa, okay, so Huawei has cooked up something pretty wild with their CloudMatrix 384 AI cluster, right? Chinese tech companies are biting, apparently. These guys are like, “Hey NVIDIA, watch out!” I mean, the CloudMatrix sounds pretty beastly, especially over in China where, yeah, let’s face it, NVIDIA’s not having the easiest time.

Huawei’s super busy… they’re on this mission to bulk up their AI hardware lineup. They’re like, “We see you, NVIDIA, and we’re coming.” So, this CloudMatrix 384 — kind of a mouthful, huh? — was born, all grown-up and ready to rumble. Heard about it from the Financial Times, no less, where they said ten big clients have hugged this Huawei baby. That’s huge, right?

The word is, big players have already thrown CloudMatrix into their data centers. Mysterious lot, those customers. Not a peep about who they are, but they’re hooked on Huawei gear. We talked about CloudMatrix before, but to sum up — it can tango with NVIDIA’s GB200 NVL72, big time. China’s in it for self-reliance, using its own tech muscle.

Now, let’s dive into the specs. CloudMatrix 384 rocks these 384 Ascend 910C chips — apparently mingling in this “all-to-all topology” thing. Just a fancy way of saying, they’ve crammed in five times more chips than NVIDIA’s offering. More chips, more power, right? Nearly 300 PetaFLOPS in computing power, which is like double what NVIDIA’s brings to the party. But, and it’s a big ‘but,’ it guzzles power like nobody’s business — 3.9 times more. Imagine the electric bill on that!

And wow, the price tag. A cool $8 million per cluster. Not exactly trying to beat NVIDIA on price, are they? Nope, it’s more about showing they’ve got homegrown chops to take on the Western tech giants.

Yeah, it’s all a bit crazy. Something tells me this saga is just getting started. Maybe it’s just me, but I’m kinda intrigued to see where these tech battles end up.