BY COMFORT OGBONNA
Demand for Huawei’s Ascend 950 artificial intelligence chips has surged dramatically following the release of DeepSeek V4, a powerful new model designed to run on the Shenzhen-based firm’s hardware. The sudden spike in interest has triggered a rush among China’s largest technology companies to secure chip supplies, according to people familiar with the matter.
Major internet giants including ByteDance, Tencent, and Alibaba have reportedly initiated discussions with Huawei to place new orders. At the same time, firms specializing in cloud computing and GPU rental services are also scrambling to obtain the chips, reflecting growing demand across multiple sectors of China’s rapidly expanding AI ecosystem.
The momentum behind Huawei’s chips highlights a shifting balance in the global semiconductor landscape. The Ascend 950PR chip is said to outperform the Nvidia H20, previously the most advanced processor the U.S. firm was allowed to sell in China before export restrictions tightened. However, it still trails Nvidia’s more advanced H200 chip, which remains largely unavailable in China due to ongoing regulatory disputes between Washington and Beijing.
Despite approvals on paper, shipments of the H200 to China have yet to materialize, as both governments continue to negotiate the terms of its sale. This delay has created a significant opening for Huawei to expand its foothold in the domestic semiconductor market, positioning its Ascend series as a viable alternative to American-made chips.
The Ascend 950PR represents a major breakthrough for Huawei after years of struggling to secure large-scale adoption within China’s tech sector. Earlier testing phases reportedly yielded strong results, with companies like ByteDance and Alibaba planning procurement after receiving sample units at the start of the year. The latest surge in demand suggests that confidence in Huawei’s chip performance is growing rapidly.
At the center of this momentum is the launch of DeepSeek V4, which has sparked intense excitement across the AI industry. The model’s optimization for Huawei’s chips signals a strategic pivot toward domestically produced hardware, reducing reliance on foreign technology. This shift aligns with Beijing’s broader push to achieve technological self-sufficiency, particularly in critical sectors such as artificial intelligence and semiconductors.
Huawei recently confirmed that its Ascend supernode infrastructure, built on the 950 series chips, fully supports DeepSeek V4. The company also noted that its entire Ascend SuperNode lineup has been adapted to handle V4 inference tasks—allowing the model to process queries and execute complex operations efficiently at scale.
One of the standout features of the Ascend 950 series is its ability to support advanced computational techniques that use more compressed numerical formats. This allows the chips to perform more calculations per second while reducing operational costs, giving them a competitive edge among domestic alternatives.
The rapid adoption of DeepSeek V4 has further accelerated demand. Alibaba Cloud quickly integrated the model into its Bailian platform on the same day of its release, offering both V4-Pro and V4-Flash versions at competitive pricing. Similarly, Tencent Cloud launched preview services through its TokenHub platform, deploying the model across both domestic infrastructure and international gateways such as Singapore.
This swift rollout by major cloud providers has made the model instantly accessible to millions of developers and users, significantly increasing the volume of AI workloads. As more applications adopt DeepSeek V4, the need for high-performance chips to support these operations continues to rise.
However, supply remains a major challenge. While DeepSeek has introduced aggressive pricing strategies—including a temporary 75% discount for developers—industry insiders warn that chip availability may not keep pace with demand. The company itself has acknowledged that pricing could drop further in the future, but only once Huawei’s Ascend 950 supernodes reach large-scale production.
The V4 model comes in two versions: V4-Pro, featuring an enormous 1.6 trillion parameters, and V4-Flash, with 284 billion parameters. Both versions support a one-million-token context window, making them among the most advanced open-source AI models available. Released under the permissive MIT license, the models allow businesses to freely use, modify, and commercialize the technology.
Even with strong production plans, Huawei is expected to face ongoing constraints. U.S. export controls on advanced chipmaking equipment continue to limit China’s ability to manufacture cutting-edge semiconductors at scale. As a result, output of the Ascend 950 series is likely to fall short of soaring demand in the near term.
According to sources familiar with Huawei’s plans, the company aims to ship around 750,000 units of the 950PR this year. Mass production began in April, but full-scale shipments are not expected until the second half of 2026. Until then, supply bottlenecks could remain a defining feature of China’s AI hardware market.
Overall, the surge in demand for Huawei’s chips underscores a pivotal moment in the global tech race. With geopolitical tensions reshaping supply chains and access to foreign technology, China’s push toward self-reliance is gaining momentum—and Huawei appears to be emerging as a central player in that transformation.