Huawei Ascend 960 AI Chips: New Chip Roadmap Targets Nvidia Amid US Export Controls

Huawei Ascend 960 AI chips are emerging as an important part of the Chinese technology company’s long-term strategy to expand its artificial intelligence computing business and compete with Nvidia. However, the latest information shows that the Ascend 960 is planned for 2027, rather than being a chip already launched in 2026.

Huawei has announced plans for multiple next-generation AI processors, including the 960DT and Ascend 960PR, while continuing to develop infrastructure that can connect very large numbers of AI chips into unified computing systems. The developments come as US export controls continue to restrict China’s access to some advanced AI processors and semiconductor manufacturing technologies.

The development highlights how export restrictions are contributing to a broader shift in the global AI semiconductor industry, with Chinese companies investing heavily in domestic alternatives.

Huawei’s New AI Chip Roadmap

Huawei’s latest roadmap includes two AI processors planned for 2027.

The company plans to launch the 960DT in the first quarter of 2027, followed by the Ascend 960PR in the third quarter of 2027, according to Huawei rotating chairman David Wang.

Huawei had previously described the Ascend 960 as part of a longer-term chip roadmap. In a 2025 presentation, the company said the Ascend 960 would offer approximately twice the computing power, memory capacity, memory bandwidth and number of interconnect ports compared with the Ascend 950 generation. Huawei’s earlier plan placed the Ascend 960 launch in the fourth quarter of 2027.

This means the company’s AI semiconductor strategy is not centered on a single chip. Instead, Huawei is developing a sequence of processors alongside increasingly large AI computing systems.

Why Huawei Is Targeting Nvidia

Nvidia remains a dominant force in the global AI accelerator market, with its GPUs and software ecosystem widely used for training and running advanced AI models.

Huawei is attempting to build an alternative ecosystem around its own Ascend processors.

The competition is particularly important in China, where US export restrictions have limited the availability of some advanced Nvidia products. Nvidia’s own regulatory filings acknowledge that export controls have affected its ability to compete in China’s data-center computing market and can create opportunities for competitors offering products outside the scope of those restrictions.

Huawei has consequently become one of the most prominent domestic alternatives for Chinese organizations seeking AI computing hardware.

Nvidia CEO Jensen Huang has previously acknowledged Huawei’s growing position in China’s AI chip market. Reuters reported in August that Huang said Nvidia had effectively conceded the Chinese AI chip market to Huawei, although Nvidia continues to pursue opportunities in China where permitted.

Ascend 960 Is Designed for Larger AI Workloads

The significance of the Huawei Ascend 960 AI chips extends beyond their individual processing performance.

Modern AI systems increasingly depend on large clusters containing thousands of processors. Training and operating advanced models can require multiple accelerators working together, making communication between chips a critical part of overall system performance.

Huawei is therefore developing its UnifiedBus technology to connect large numbers of AI processors.

According to Reuters, Huawei has developed 11 semiconductors using the technology and is working toward large computing clusters capable of connecting up to one million AI processors. The company has also delivered more than 1,000 smaller systems known as supernodes to more than 370 customers.

This approach allows Huawei to compete at the system level rather than relying solely on the specifications of an individual processor.

Huawei’s Ascend 950 and 960 Roadmap

Huawei’s broader roadmap shows how quickly the company is attempting to advance its AI hardware.

In its 2025 presentation, Huawei said the Ascend 950 would be followed by the Ascend 950DT, with the latter planned for the fourth quarter of 2026. The company said the 950DT would be optimized for AI inference and training and support several data formats, including FP8, MXFP8, MXFP4 and HiF8.

The Ascend 960 is positioned as the next major step.

Huawei said the 960 would double several important specifications compared with the Ascend 950 and support its proprietary HiF4 data format, which is designed for high-efficiency four-bit AI processing.

The company has also outlined plans for an Ascend 970 after the 960, indicating that Huawei intends to maintain a relatively rapid product-development cycle.

SuperPoD and SuperCluster Strategy

Huawei is also focusing on complete AI infrastructure rather than individual chips.

The company has developed its Atlas SuperPoD systems, which combine large numbers of Ascend processors into a single computing platform.

Huawei previously said its planned Atlas 960 SuperPoD would contain up to 15,488 Ascend 960 chips. The company projected that the system could deliver 30 exaflops of FP8 computing power and 60 exaflops of FP4 computing power.

Huawei also outlined an even larger SuperCluster concept involving more than one million NPUs.

These ambitions demonstrate the importance of interconnect technology in the company’s strategy. Instead of attempting to match Nvidia only at the single-chip level, Huawei is attempting to create large-scale computing systems capable of handling increasingly demanding AI workloads.

US Export Controls Are Reshaping China’s AI Chip Industry

The development of Huawei Ascend 960 AI chips is taking place against the backdrop of extensive US semiconductor export controls.

Washington began imposing restrictions on advanced semiconductor and AI-related technologies for China in 2022. The controls cover certain AI chips as well as semiconductor manufacturing equipment and related technologies.

Nvidia has said in regulatory filings that the restrictions include parameters such as processing performance, performance density, interconnect bandwidth and memory bandwidth. The company has also warned that additional restrictions could further affect its ability to compete in China.

These restrictions have increased pressure on Chinese companies to develop domestic alternatives.

For Huawei, that has meant investing in chip design, AI processors, interconnect technologies and complete computing systems.

Domestic AI Hardware Gains Importance in China

China’s AI industry has been expanding rapidly, creating substantial demand for computing infrastructure.

The restrictions on foreign AI processors have simultaneously created challenges and incentives for Chinese semiconductor companies.

Huawei is currently one of the leading domestic players, while other Chinese companies are also developing AI accelerators. Reuters has reported that firms such as Biren, Enflame, Moore Threads and MetaX are developing competing AI hardware products.

Biren, for example, reported significant revenue growth in the first half of 2026 as demand for domestically produced AI accelerators increased. However, its scale remains much smaller than Nvidia’s global business.

This suggests that China’s AI chip market is developing into a broader domestic ecosystem rather than relying on a single Huawei alternative.

Can Huawei Directly Challenge Nvidia?

The Huawei Ascend 960 AI chips represent a serious effort to narrow the technology gap, but comparisons with Nvidia require caution.

Nvidia has advantages that extend beyond processor specifications. Its CUDA software ecosystem, large developer community, networking technologies and global customer base are important parts of its competitive position.

Huawei is attempting to build its own ecosystem around Ascend processors and UnifiedBus infrastructure.

The two companies are therefore competing across several layers:

• AI accelerator performance
• Memory bandwidth
• Chip-to-chip communication
• AI software ecosystems
• Data-center infrastructure
• Developer adoption
• Supply-chain availability
• Cost and energy efficiency

Independent assessments have also highlighted continuing manufacturing and performance challenges for China’s AI chip industry. A Council on Foreign Relations analysis, for example, has argued that China’s semiconductor manufacturing constraints could make it difficult for Huawei to match Nvidia’s highest-end performance in the near term.

Software Could Be Just as Important as Hardware

AI chip competition is not determined by hardware specifications alone.

Nvidia’s CUDA platform has become deeply integrated into AI development, allowing researchers and companies to build applications using a mature collection of libraries, frameworks and development tools.

Huawei therefore needs to encourage developers and enterprises to adopt its own software ecosystem.

The company has been investing in its Ascend ecosystem and AI infrastructure to make its processors useful across a wide range of workloads.

If Chinese developers increasingly optimize AI models for Huawei hardware, the company could strengthen its position in China’s domestic market even if it remains behind Nvidia in some global benchmarks.

Global AI Semiconductor Competition

The Huawei Ascend 960 AI chips are part of a much larger transformation in the semiconductor industry.

The AI boom has made advanced accelerators strategically important for technology companies and governments. The United States wants to restrict China’s access to certain cutting-edge technologies, while China is investing heavily in domestic semiconductor capabilities.

This competition is also affecting companies outside China.

Nvidia has warned that changing export regulations can complicate product development, distribution and customer relationships. The company has also said that export controls could encourage customers in other markets to reduce their dependence on US semiconductor products because of compliance concerns.

As a result, the AI chip industry is becoming increasingly connected to broader supply-chain and technology policy decisions.

What Comes Next for Huawei’s AI Chips?

The next major milestones will be the launch of Huawei’s 960-series processors and the expansion of its large-scale AI computing infrastructure.

The 960DT is planned for the first quarter of 2027, while the Ascend 960PR is expected in the third quarter of 2027. Huawei is also continuing work on UnifiedBus and large AI clusters.

The success of these products will depend on several factors, including manufacturing capacity, real-world performance, software compatibility, energy efficiency and customer adoption.

Huawei will also need to maintain a steady supply of processors while improving the broader ecosystem around them.

Looking Ahead

The Huawei Ascend 960 AI chips represent a significant step in China’s effort to develop advanced domestic AI computing infrastructure.

Rather than focusing exclusively on individual processors, Huawei is building an integrated strategy involving AI accelerators, high-speed interconnects, supernodes, SuperPoDs and massive computing clusters.

The company’s plans do not mean that Nvidia’s global position has been displaced. Nvidia continues to benefit from its established hardware, software and developer ecosystem. However, Huawei’s growing presence demonstrates how export controls are encouraging the development of alternative AI semiconductor supply chains.

With the 960-series processors scheduled for 2027, the next stage of the Huawei-Nvidia competition will increasingly be determined by real-world performance, production scale, software ecosystems and the ability to deploy large AI computing systems efficiently.

Frequently Asked Questions

1. What are Huawei Ascend 960 AI chips?
Huawei Ascend 960 AI chips are next-generation artificial intelligence processors that Huawei is developing as part of its broader Ascend semiconductor roadmap.

2. When will the Ascend 960 launch?
Huawei’s earlier roadmap placed the Ascend 960 in the fourth quarter of 2027. More recent plans also identify 960-series variants, including the 960DT and 960PR, for 2027.

3. Is Huawei already selling the Ascend 960?
The Ascend 960 is a future-generation product rather than a chip already broadly available in 2026. Huawei has announced planned 2027 launches.

4. Why is Huawei developing new AI chips?
Huawei is expanding its domestic AI computing capabilities amid restrictions on China’s access to certain advanced foreign AI processors and semiconductor technologies.

5. Is Huawei trying to compete with Nvidia?
Yes. Huawei’s AI chip and computing-system development is explicitly aimed at strengthening its position against major AI hardware companies including Nvidia.

6. What is UnifiedBus?
UnifiedBus is Huawei’s interconnect technology designed to allow large numbers of AI processors to work together as a larger computing system.

7. How many chips could Huawei’s future systems connect?
Huawei has discussed systems capable of connecting up to one million AI processors and has developed smaller supernode systems for customers.

8. What role do US export controls play in Huawei’s AI strategy?
US export controls restrict access to certain advanced AI chips and semiconductor technologies for China. The restrictions have increased the importance of domestic alternatives such as Huawei’s Ascend family.

9. Is Huawei’s AI hardware as powerful as Nvidia’s?
Performance comparisons depend on the specific chip, workload and software environment. Nvidia continues to have major advantages in its global hardware and software ecosystem, while Huawei is developing increasingly capable domestic alternatives.

10. What is the significance of the Ascend 960 for China’s AI industry?
The Ascend 960 is part of China’s broader effort to develop domestic AI computing infrastructure and reduce dependence on restricted foreign semiconductor technologies.