•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
By combining DeepSeek, Huawei’s AI chips and the cloud platforms of Alibaba and Tencent, China has for the first time built a closed AI ecosystem that does not rely on NVIDIA chips or U.S. technology. After DeepSeek unveiled its V4 model last week, demand for Huawei’s Ascend 950 AI chips surged, according to three people familiar with the matter. Chinese tech giants, including ByteDance, Tencent and Alibaba, have been contacting Huawei about new chip orders, while cloud computing and GPU rental providers are also rushing to place orders.
The development marks the first time a major Chinese AI model runs entirely on domestic chips without U.S. technology. DeepSeek V4 is optimized for Huawei’s chips, signaling a strategic shift away from dependence on American semiconductors toward domestic AI hardware, a Beijing priority in the race for technological dominance.
Last week, Huawei said its Ascend SuperNode hyperscale infrastructure, built on Ascend 950 chips, would fully support DeepSeek V4 models. Huawei’s Ascend SuperNode lineup has been tuned for V4 inference—using trained AI models to answer queries and perform tasks.
Among Chinese chip makers, Ascend 950—specifically the 950PR variant—is described as the only domestic chip that supports AI compute in a compressed numeric format. That design is intended to allow the chip to process more calculations per second at lower cost. While 950PR still lags NVIDIA’s more advanced H200, it is described as significantly superior to the H20, which is the strongest NVIDIA product allowed to be sold in China until Beijing blocked imports last year.
The 950PR chip represents Huawei’s breakthrough after years of struggling to win large orders from China’s technology sector. Reuters reported in March that customer testing of the chip went smoothly earlier this year. Companies including ByteDance and Alibaba planned to place orders after samples were distributed in January.
The Huawei chip race highlights how DeepSeek’s V4 release accelerated demand for domestic AI hardware as U.S. export controls continue to restrict access to NVIDIA’s most advanced processors. It also reflects validation of Huawei’s chip performance to date.
To gain early access, Alibaba Cloud’s Bailian platform offered DeepSeek V4 on launch day, providing both V4-Pro and V4-Flash at prices aligned with DeepSeek’s official pricing. Tencent Cloud launched a Preview V4 service on its TokenHub platform the same day, deploying the model on domestic nodes and Singapore’s international gateway to serve global users.
Because major cloud platforms can deploy the model quickly, millions of users and developers can access V4, which in turn is expected to boost AI query volume and underlying chip demand.
DeepSeek is offering developers a 75% discount on its new model through May 5, 2026. The company also said V4-Pro prices could fall materially in the second half of 2026 after large-scale delivery of Huawei’s Ascend 950, though restrictions are expected to persist until production ramps up, reflecting tight supply of high-end domestic AI chips.
DeepSeek V4 comes in two versions: V4-Pro with a total of 1.6 trillion parameters and V4-Flash with 284 billion parameters. Both versions support a one-million-token context window. The models are released under the MIT open-source license, allowing companies to use, modify and commercialize the technology.
Despite the demand surge, 950 production is not expected to meet requirements due to U.S. export restrictions on advanced chip manufacturing tools that prevent China from acquiring cutting-edge equipment.
Huawei plans to ship around 750,000 units of 950PR this year, with mass production starting in April and full deliveries beginning in the second half of 2026, according to people familiar with the plan.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
