Huawei MDC Intelligent Driving Computing Platform and Ascend AI Chips

Huawei MDC (Mobile Data Center) is an automotive-grade intelligent driving computing platform developed by Huawei based on its in-house Ascend AI chips, engineered to deliver core computing power for Advanced Driver Assistance Systems (ADAS) as well as Level 2+ to Level 4 autonomous driving. Regarded as the "brain" behind Huawei's full-stack smart vehicle solutions, it is one of the few domestic intelligent driving platforms offering an entirely self-developed stack spanning chips, hardware platforms, operating systems, and toolchains.

Ascend AI Chips: The Computing Foundation of MDC

Ascend is Huawei's proprietary AI processor series powered by the DaVinci architecture, serving as the core engine behind the MDC platform's high-performance computing capabilities. Ascend processors deliver high energy efficiency alongside robust AI inference performance; for instance, the early Ascend 310 chip achieved 16 TOPS of computing power while simultaneously recognising up to 200 distinct objects under low power consumption.

The current mainstream model, the Ascend 610, is built on a 7nm process and delivers dense AI computing power of 160–200 TOPS (INT8), setting an industry benchmark for real-world energy efficiency. On the cloud training front, Huawei Cloud has deployed over 100,000 Ascend compute cards to accelerate vehicle manufacturers' intelligent driving algorithm iterations, while the next-generation Ascend 950DT chip was officially rolled out in August 2026 to further bolster cloud-based foundation model training.

MDC Platform Product Matrix

The Huawei MDC platform adopts a standardised and modular design philosophy, offering a comprehensive product portfolio tailored to diverse computing demands and application scenarios:

The MDC 610's hardware architecture typically incorporates a core controller setup featuring the "Ascend 610 + Infineon TC397", striking a balance between heavy AI compute workloads and ASIL-D functional safety control.

Technical Architecture and Ecosystem

The MDC platform is far more than just a cluster of silicon; it functions as a comprehensive heterogeneous computing system. Running Huawei's in-house intelligent driving operating systems (AOS/VOS), it is fully compatible with AUTOSAR and ROS, supports mainstream AI frameworks such as Caffe, TensorFlow, and PyTorch, and provides an extensive library of over 400 operators.

Across the wider industry ecosystem, Huawei collaborates with carmakers via a tiered partnership model: firstly, by supplying full-stack solutions through the HI (Huawei Inside) model (e.g., AVATR); and secondly, as a Tier 1 supplier providing the standalone MDC platform or chipsets, enabling car manufacturers to develop their proprietary algorithms (e.g., BYD, GAC, among others).

Development Trends: The "vehicle-to-cloud" synergy of the MDC platform is increasingly coming to the fore. In tandem with Huawei Cloud's AI computing clusters, it enables a rapid development loop of "data collection — cloud training — vehicle deployment". As of 2026, more than 30 OEMs have forged deep partnerships with Huawei Cloud in the intelligent driving space, with total autonomous mileage under the Qiankun ADS system surpassing 736 million kilometres.