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

Hey, so here’s the deal with Huawei and their latest move with this CloudMatrix 384 thing. It’s like, suddenly, everyone’s talking about it. I didn’t even see it coming. Chinese tech giants are apparently all over this, giving NVIDIA a serious run for its money—especially in China. Bold move, right?

Title? Not Sure: Huawei’s CloudMatrix 384 Shake-Up: Is NVIDIA Sweating in China?

Anyway, I heard Huawei’s really pushing their AI stuff. Like they’re really going for it. Imagine a poker player going all-in, but in the tech world. People are buzzing about this CloudMatrix 384 AI cluster—it’s all made with Huawei’s own gizmos. And guess what? Ten big shots, hush-hush on the names though, are already using it. It’s like a secret fan club, but with data centers.

Sure, I rambled on about CloudMatrix 384 before, but let’s do a quick rerun. This thing is a beefy competitor to NVIDIA’s top dog, the GB200 NVL72. China’s finally making moves to be less dependent on others for tech. How about that?

Random fact: the cluster sports 384 Ascend 910C chips—and get this—an "all-to-all topology." I can’t even pretend to know exactly what that means, but it sounds intense. They crammed in a ton of chips, five times more than NVIDIA’s rig, which is wild. It dishes out 300 PetaFLOPS, which, if I’m being honest, sounds made-up but is apparently double what NVIDIA does. Downside? It guzzles power like you wouldn’t believe, 3.9 times more than GB200. Yikes, right? The energy bill must be something else.

Random Pic Description: Imagine a table of tech specs. It’s like comparing apples to a three-piece suit—random but you get the point.

The cost? Oh man, brace yourself. A CloudMatrix 384 sets you back $8 million! That’s like triple NVIDIA’s sticker price. So, if you’re looking for something wallet-friendly, this ain’t it. But hey, who needs cheap when you want to stick it to the West with your homegrown tech?

So, there’s that. Hope this all makes sense. Anyway—wait, what was I saying? Ah, never mind. Catch you later.