NVIDIA and MediaTek Deepen Alliance Across AI, PC, and Automotive

NVIDIA and MediaTek have announced an expansion of their strategic partnership to advance next-generation artificial intelligence computing platforms. The collaboration spans cloud AI infrastructure, client PC computing, and automotive systems, backed by a $3.5 billion investment by NVIDIA into convertible bonds issued by MediaTek.

Context of the Cooperation
Hyperscalers, cloud service providers, and frontier artificial intelligence model developers increasingly seek customized processing units (XPUs) tailored to specific workloads. However, transitioning custom silicon from design into scalable, production-ready rack systems introduces major engineering bottlenecks around high-speed SerDes, memory interfaces, advanced multi-die packaging, and interconnect fabrics.

To address these engineering complexities and establish standardized deployment paths across AI factories, client devices, and vehicles, cooperation was expanded between NVIDIA and MediaTek. The partnership combines NVIDIA’s accelerated computing architecture, graphics technologies, and software stack with MediaTek’s custom silicon design, power-efficient System-on-Chip (SoC) architectures, advanced packaging, and wireless connectivity.

Technical Solution and Responsibilities
The core technical integration centers on MediaTek adopting the NVIDIA NVLink Fusion™ platform as a modular design blueprint for semi-custom multi-die XPU development. Responsibilities are allocated across the companies to align custom chip design with modular rack-scale systems:

NVIDIA supplies the accelerated computing architecture, software platforms, and the underlying NVLink Fusion ecosystem components, including NVIDIA NVLink Fusion Chiplets, NVLink-C2C interconnects, NVHBM memory architectures, and MGX™ rack-scale blueprints.

MediaTek acts as the design and implementation partner, utilizing NVLink Fusion to build custom accelerators tailored to specific customer workloads while managing packaging, physical implementation, and supply-chain logistics.

At a system level, the NVLink Fusion Chiplet connects third-party XPUs to the NVLink scale-up fabric using optical photonics or electrical interconnects. NVLink-C2C provides high-bandwidth, energy-efficient die-to-die communications between customer XPUs, NVIDIA Rosa CPUs, and compatible processors. NVHBM integrates customized memory interfaces to maximize bandwidth and energy efficiency while preserving silicon area for active compute engines.

Deployment or Implementation
The implementation strategy focuses on deployment across three primary computing domains:

Edited by Romila DSilva, Induportals Editor, with AI assistance.