Is Huawei a Real Threat to Nvidia’s AI Empire?

In a rare public warning, Nvidia CEO Jensen Huang acknowledged that while his company’s AI technology is a “generation ahead,” Chinese rival Huawei “has got China covered” if U.S. export controls continue to restrict American participation in the world’s second-largest economy. The comments underscore a critical tension at the heart of the global technology race: Washington’s efforts to slow Beijing’s technological ascent may be accelerating the very outcome it fears—the rise of a potent, self-sufficient Chinese competitor.

Huang’s remarks frame the central conflict for the AI industry. Nvidia sits atop a $3 trillion valuation built on its dominance in the chips that power AI. But as US restrictions tighten, a state-backed Huawei is rapidly emerging as China’s national champion, raising questions about the durability of Nvidia’s moat and the future of the global AI landscape.

What makes Nvidia so dominant?

Nvidia’s power extends far beyond just making the fastest chips. Its dominance is built on a comprehensive ecosystem developed over two decades, often referred to as a “moat.” At its core is CUDA, a proprietary software platform that allows developers to harness the parallel processing power of Nvidia’s GPUs. Millions of developers are trained on CUDA, and the vast majority of AI research and applications, including dominant frameworks like PyTorch and TensorFlow, are optimized to run on it. This creates enormous switching costs for anyone trying to move to a competing hardware platform.

Beyond CUDA, Nvidia offers a “full-stack” solution that includes not just its industry-leading GPUs like the H100 and the new Blackwell series, but also proprietary high-speed interconnects (NVLink) that link thousands of GPUs together into a single, cohesive supercomputer. This systems-level integration is critical for large-scale AI training and has proven exceptionally difficult for competitors to replicate.

How have US export controls changed the game?

Washington’s strategy has centered on creating chokepoints in the semiconductor supply chain to deny China access to the most advanced technology. This began with restricting sales of Nvidia’s top-tier AI chips, like the H100, and has expanded to include the advanced lithography equipment needed to manufacture them, primarily the extreme ultraviolet (EUV) machines sold exclusively by Dutch firm ASML.

In response, Nvidia developed less-powerful, export-compliant chips for China, such as the H20. However, the Trump administration tightened the rules again in April 2025, requiring licenses even for these downgraded chips. The move forced Nvidia to take a $5.5 billion charge on its China-specific inventory and effectively froze it out of what Huang estimates will be a $50 billion AI market within a few years. These restrictions have had the dual effect of fueling China’s drive for self-sufficiency while creating an opportunity for domestic players to fill the vacuum left by Nvidia.

How formidable is Huawei’s challenge?

Once crippled by US sanctions that nearly destroyed its smartphone business, Huawei has re-emerged as Beijing’s primary weapon in the chip war. With substantial state support, the company is developing a full stack of AI hardware to rival Nvidia’s. Its Ascend series of AI processors, particularly the 910B and 910C, are gaining traction with major Chinese tech firms like ByteDance and Tencent.

In April 2025, Huawei unveiled the CloudMatrix 384, a computing system connecting 384 of its Ascend 910C chips. By some metrics, the system delivers more raw computing power and memory than Nvidia’s flagship GB200 NVL72 rack, though it is significantly less power-efficient. Huawei reportedly has the capacity to produce over a million Ascend chips, enough for thousands of clusters. While this is still a fraction of Nvidia's output, it demonstrates that China now has a viable, domestically produced alternative for large-scale AI training, a situation that did not exist a few years ago. Huang himself has acknowledged that Huawei’s latest technology is now comparable to Nvidia's H200, one of its most powerful previous-generation chips.

Can China build a self-sufficient ecosystem?

Creating a viable alternative to Nvidia requires more than just capable hardware; it demands a competitive software ecosystem to rival CUDA. This remains China's biggest hurdle. Huawei is promoting its own platform, CANN (Compute Architecture for Neural Networks), but attracting a critical mass of developers away from the well-established and highly optimized CUDA ecosystem is a monumental task that could take many years.

Furthermore, China remains dependent on foreign technology at key chokepoints, most notably in manufacturing. China’s top foundry, SMIC, can produce 7-nanometer chips—a significant achievement demonstrated in Huawei’s Mate 60 smartphone—but it relies on older deep-ultraviolet (DUV) lithography. To reach the most advanced nodes (3nm and below), EUV lithography is essential, and ASML remains the world's sole supplier. With the US pressuring the Netherlands to restrict ASML’s sales to China, Beijing is years away from developing its own comparable EUV machines.

However, not all AI progress requires cutting-edge chips. Chinese startup DeepSeek recently shocked the industry by training a world-class AI model using only a fraction of compute that its Western counterparts used. This breakthrough suggests that algorithmic and software efficiencies could partially offset hardware limitations, allowing China to remain competitive even while trailing on the most advanced manufacturing processes.

What does this mean for the global AI landscape?

The US strategy has effectively triggered the bifurcation of the global AI supply chain into two competing spheres: a US-led stack built around Nvidia, and a burgeoning Chinese stack centered on Huawei. Huang has warned that forcing China’s 50% of the world’s AI researchers to develop on a domestic platform could have unintended long-term consequences for American technological leadership.

At the same time, the geopolitical turmoil is fueling a new global trend: “sovereign AI.” Nations from the Middle East to Europe are now investing billions to build their own domestic AI infrastructure to ensure they aren't reliant on either the US or China. This creates new, multibillion-dollar markets for Nvidia, potentially offsetting losses from China and reducing its dependence on a handful of US hyperscalers. For now, Nvidia remains the undisputed king of AI hardware, but the accelerating efforts of a determined and well-funded rival in a protected market ensure that its reign will not go unchallenged.

Reference Shelf:

Huawei 'has got China covered' if the U.S. doesn't participate, Nvidia CEO tells CNBC (CNBC)

Nvidia's CEO says China is not far behind the U.S. in AI capabilities (Tom's Hardware)

Huawei Develops New AI Chip, Seeking to Match Nvidia (WSJ)

Why ASML and TSMC Are the Chokepoints in Global Chipmaking (ARPU)