Huawei Accelerates Ascend 960DT Launch to Match DeepSeek Demand

Huawei has accelerated the release timeline for its next-generation Ascend 960DT AI chip, shifting the launch from the third quarter of 2027 to the first quarter. This two-quarter advance aligns with reports that Chinese AI lab DeepSeek is preparing to deploy at least 160,000 of these accelerators at a new data center in Inner Mongolia.

The timing suggests a coordinated effort to secure domestic alternatives to Nvidia hardware. While Huawei aims to close the performance gap with its rival, DeepSeek’s large-scale order indicates confidence that the current generation of chips can support their training and inference workloads, despite ongoing U.S. export restrictions.

Accelerated roadmap targets early 2027

At its Huawei Connect conference, executives detailed a revised product schedule that pulls the Ascend 960DT forward by roughly six months. The company also outlined performance targets for the subsequent Ascend 960PR, projecting a doubling of 4-bit floating-point throughput. This metric is critical for efficient AI inference, as low-precision math drives the majority of large-scale model operations and directly impacts power costs.

However, the accelerated timeline does not guarantee immediate parity with Nvidia. Independent benchmarks comparing the new Ascend chips to Nvidia’s current data-center lineup are not yet available. The launch remains approximately eighteen months away, leaving a window where the performance gap may persist despite the increased urgency behind Huawei’s development efforts.

DeepSeek bets on domestic silicon

DeepSeek’s planned deployment of 160,000 Huawei accelerators would make it one of the largest known users of the Ascend platform. This move is driven by years of tightening U.S. export controls that have limited access to Nvidia’s most advanced GPUs. By committing to a massive cluster of domestic chips, DeepSeek is signaling that its software optimizations can extract sufficient performance from Huawei hardware to meet its operational needs.

This strategy mirrors a broader industry trend where major players like Microsoft, Meta, and Amazon are developing custom silicon to reduce dependence on any single supplier. For DeepSeek, the bet on Huawei is a practical response to supply chain constraints, relying on its track record of achieving strong results with constrained hardware. The success of this deployment will hinge on the stability of the software stack supporting the Ascend chips.

Global shift away from Nvidia

The convergence of Huawei’s aggressive scheduling and DeepSeek’s infrastructure investment highlights a structural shift in the AI chip market. According to tech-insider.org, these moves occur alongside reports that Western tech giants are pushing their own custom chips into production. This simultaneous effort by both Chinese and Western companies to diversify away from Nvidia suggests a permanent change in the competitive landscape, where specialized silicon plays a larger role in reducing costs and supply risks.