Huawei is already trying to push Windows out of China’s PCs. Now it wants to take on Nvidia, with new AI chips and a system designed to connect up to one million processors
Huawei has already decided that China doesn’t need Windows. Now it appears to have Nvidia in its sights too. Last month, we looked at how Huawei’s HarmonyOS PCs could begin pushing Microsoft’s Windows out of China’s computer market. Now the Chinese technology giant is aiming at another American technology giant, announcing two new AI chips for 2027 as it tries to build an alternative to Nvidia’s hardware and software empire.
The first, called the 960DT, is scheduled to arrive in the first quarter of 2027, followed by the Ascend 960PR in the third quarter. But the chips themselves may not be the most interesting part. Huawei is also developing a technology called UnifiedBus that can connect enormous numbers of AI processors and make them work together as a single computing system. Huawei says its biggest systems could eventually link up to one million AI processors.
Huawei Doesn’t Just Want Another Chip
The interesting thing about Huawei’s strategy is that it isn’t simply trying to produce a faster piece of silicon and hope developers show up. Nvidia’s dominance comes from an entire ecosystem: powerful GPUs, networking technology, software tools and a huge developer community built around its hardware.
Huawei appears to be taking a similar approach. Its UnifiedBus technology is designed to let large numbers of AI chips communicate and operate together efficiently. The company says it has already developed 11 semiconductors based on the technology for use in its large AI systems. Its biggest linked machines, which Huawei calls superclusters, can support up to one million AI processors.
That is an eye-catching number, although it is Huawei’s claimed system capacity rather than a demonstration that a million chips are already operating together in a production system.
The company has already started selling smaller versions of this idea. Huawei says it has shipped more than 1,000 supernodes to more than 370 customers. A supernode combines multiple AI chips into a single system capable of tackling the same task. Huawei has not disclosed how many chips are inside each supernode, who all those customers are or how the shipments are divided between different models.
Still, the numbers suggest Huawei isn’t waiting for its next generation of chips before trying to build a domestic AI computing market. The company is effectively betting that if it can’t freely buy the world’s most advanced AI processors, it can compensate by connecting large numbers of locally available processors into increasingly powerful machines.
The Nvidia Problem Is Bigger Than Silicon
That strategy matters because China’s access to advanced computing hardware has been restricted by U.S. export controls. The restrictions cover certain advanced AI chips as well as semiconductor manufacturing equipment, making it harder for Chinese companies to simply buy whatever hardware is available to American AI labs. Nvidia remains the global leader, and its advantage isn’t just raw chip performance. Developers have spent years building software around Nvidia’s CUDA ecosystem, while cloud companies and AI labs have designed huge computing infrastructures around its processors.
Huawei therefore has a much bigger problem to solve than designing a competitive chip. It needs to make an entire alternative computing stack attractive enough for Chinese developers and companies to use.
Huawei says that the ecosystem is growing. David Wang, the company’s rotating chairman, said Huawei’s AI chip ecosystem now has 5,270 monthly active developers. That number is tiny compared with the enormous global community surrounding Nvidia, so Huawei still has plenty of ground to cover. But China’s unusual position gives it a powerful incentive to keep building. If domestic AI companies cannot reliably access the most advanced foreign processors, an alternative that is good enough, available locally and supported by Chinese developers becomes considerably more valuable.
From Windows to Nvidia
That makes Huawei’s AI push surprisingly similar to its operating-system strategy. When the company started pushing HarmonyOS for PCs, the goal wasn’t merely to produce another operating system. Huawei was trying to build an ecosystem that could reduce dependence on American software. The same logic is now appearing in AI computing. Hardware, networking, operating systems, developer tools and AI applications all become pieces of the same puzzle. Huawei is trying to make the pieces work together.
And there is a particularly interesting twist here. Huawei isn’t necessarily trying to beat Nvidia by producing a single chip that is dramatically faster than Nvidia’s best processor. Its UnifiedBus approach suggests another route: connect more chips and make the entire system behave like one giant machine.
That matters because modern AI models increasingly require more computing power than a single processor can provide. Fast communication between thousands of processors can therefore become almost as important as the processors themselves. Huawei says it has already built the technology to connect large numbers of chips, and its 2027 chips will give the company another generation of hardware to plug into that infrastructure.
Huawei’s Bigger Nvidia Bet
Huawei is still a long way from replacing Nvidia globally. Nvidia has an enormous head start in chips, software, networking and developers, while Huawei remains constrained by access to advanced manufacturing technology. But China doesn’t necessarily need Huawei to win the global AI chip war for this strategy to matter.
If Huawei can build a sufficiently capable domestic alternative, China could gradually become less dependent on Nvidia while creating its own AI hardware ecosystem from the chip upward. A few weeks ago, Huawei’s ambition looked like replacing Windows on Chinese PCs. Now it is building the machinery underneath China’s AI ambitions too.
The bigger question is how far that ecosystem can grow before the rest of the world starts paying attention.
In case you missed:
- China Wants a Future Without Windows. It Starts This September.
- Goodbye Blackwell, Hello Rubin: Nvidia’s new AI platform is here!
- NVIDIA’s Taiwan Problem: The AI Boom Is Now Dependent on One Island
- QpiAI Unveils Kaveri, India’s First Aatmanirbhar 64-Qubit Quantum Chip
- Tata Electronics to Manufacture Intel Chips & AI Laptops in India!
- Not Just Oil Part 2: How the Iran War is Disrupting Helium, Chips, and the AI Boom
- You Can Now Host an NVIDIA Data Centre in Your Home. Some Neighbours Already Want Them Gone.
- One Chip to Rule Them All: The 6G Chip at 100 Gbps!
- Aatmanirbhar Push: India Becomes 7th Nation to Crack Gallium Nitride Chip Technology
- Apple, Tesla, Tata: The Cyberattack That Hit India’s Manufacturing Boom
