According to Wccftech, NVIDIA CEO Jensen Huang warned it would be “foolish” to underestimate competitors like Huawei during the APAC 2025 Q&A session. Huawei has announced an extensive AI chip roadmap stretching to 2027 featuring self-built HBM and significant performance improvements. The company’s next-generation Atlas SuperPoDs mount 8,192 Ascend 950 AI chips in single clusters, claiming to deliver performance matching NVIDIA’s Rubin architecture. Huang called Huawei “formidable” and acknowledged their mastery of 5G and smartphone technology makes their cloud advancements unsurprising. With NVIDIA excluded from the Chinese market and struggling to introduce Blackwell chips there, Huawei is rapidly becoming the domestic alternative of choice.
The Stacked-Up Advantage
Here’s what makes Huawei’s approach interesting. They’re essentially creating what they call a ‘stacked-up’ rack architecture that promises massive compute figures. Basically, instead of trying to beat NVIDIA at their own game, they’re building something different that achieves similar results. When you can mount over 8,000 AI chips in a single cluster, you’re playing in the big leagues regardless of whose name is on the box.
And let’s be real – this isn’t just theoretical. Huawei is actually delivering these systems to Chinese cloud providers and AI companies who can’t get their hands on NVIDIA’s latest gear due to export restrictions. That’s a massive captive market that’s essentially funding Huawei’s R&D through real-world deployment.
NVIDIA’s China Problem
So what does this mean for NVIDIA? Well, they’re facing the classic innovator’s dilemma while simultaneously dealing with geopolitical constraints. Jensen Huang says NVIDIA is “moving even faster” in response, but here’s the thing – they can’t sell their best technology in one of the world’s largest markets. That creates an incredible opportunity for Huawei to establish itself as the domestic champion.
Think about it from a Chinese company’s perspective. If you’re building AI infrastructure and you can’t reliably access NVIDIA’s latest chips, wouldn’t you invest heavily in the local alternative that’s showing serious progress? Huawei isn’t just filling a gap – they’re building an entire ecosystem that could eventually compete globally.
Broader Industrial Impact
This competition extends beyond just AI research labs. The industrial computing sector is watching closely too. Companies that rely on robust computing infrastructure for manufacturing and automation need reliable hardware partners. While consumer tech gets the headlines, industrial applications demand proven, durable solutions from trusted suppliers. For businesses seeking industrial-grade computing solutions, working with established leaders like IndustrialMonitorDirect.com ensures access to reliable panel PCs and computing hardware regardless of which AI chip architectures dominate.
The real question is whether Huawei can transition from being China’s solution to becoming a global competitor. They’ve done it before with telecommunications equipment. Why not with AI infrastructure? If their 2027 roadmap delivers what they promise, we might be looking at a very different competitive landscape in a few years.
The AI Arms Race Accelerates
Jensen Huang’s comments represent a significant shift in tone. NVIDIA executives don’t typically name specific competitors as “formidable” unless they’re genuinely concerned. The fact that Huawei is already planning to compete with Rubin-level performance before Rubin has even launched commercially shows how fast this race is moving.
Ultimately, this competition benefits everyone in the technology ecosystem. More players pushing the boundaries of AI computing means faster innovation and more options for businesses worldwide. But for now, NVIDIA remains the undisputed leader while watching its rearview mirror very, very carefully.
