Blog
NVIDIA invests $3.5 billion in MediaTek and expands NVLink Fusion for custom XPUs
On 31 August 2026 NVIDIA and MediaTek said they are deepening a partnership covering cloud AI factories, local AI PCs, and automotive. MediaTek will adopt NVIDIA NVLink Fusion so customers can bring custom XPUs into NVLink rack-scale systems. NVIDIA also invested $3.5 billion in MediaTek convertible bonds.
NVIDIA and MediaTek on 31 August 2026 announced an expanded collaboration across cloud AI factories, local AI computing, and automotive platforms. NVIDIA also invested $3.5 billion in convertible bonds issued by MediaTek.
The centerpiece for frontier-model builders is MediaTek’s adoption of the NVIDIA NVLink Fusion platform. NVIDIA says that gives hyperscalers, cloud providers, and frontier model developers a prevalidated path to design custom XPUs and plug them into NVIDIA NVLink-connected, rack-scale AI factories.
Custom accelerators on NVIDIA’s scale-up fabric
MediaTek will offer NVLink Fusion as a design foundation for customers building custom AI accelerators. NVIDIA describes the stack as prebuilt and system-prevalidated for multi-die XPU development, covering silicon, advanced packaging, and rack-scale integration.
The platform, as listed in the joint release, includes:
- The NVLink Fusion chiplet, connecting XPUs to the NVLink scale-up fabric over NVIDIA photonics or electrical interconnects.
- NVLink-C2C, for high-bandwidth links between XPUs, NVIDIA Rosa CPUs, and other compatible processors.
- NVHBM, NVIDIA’s custom high-bandwidth memory approach, which the release says increases bandwidth and energy efficiency while dedicating more silicon area to compute.
NVIDIA’s pitch is that customers can focus on differentiated compute while NVIDIA and MediaTek supply connectivity, memory architecture, packaging, manufacturing, and rack-scale pieces. Customers can bring XPU designs to MediaTek and tailor connectivity, memory, packaging, performance, and power. Partners can also use the NVIDIA MGX rack-scale architecture.
NVIDIA founder and CEO Jensen Huang said MediaTek’s SoC, connectivity, and power-efficiency work, combined with NVIDIA accelerated computing, is meant to “give customers the freedom to create differentiated AI systems at enormous scale.” MediaTek vice chairman and CEO Rick Tsai said NVIDIA’s investment “strengthens a collaboration that spans cloud AI infrastructure, local AI computing and automotive in the era of physical AI.”
PCs, workstations, and vehicles
The companies said they will keep collaborating on multiple generations of NVIDIA RTX Spark and DGX Spark PC chips that pair NVIDIA GPUs with MediaTek SoCs for consumer PCs, developer systems, and enterprise workstations. MediaTek previously worked with NVIDIA on the GB10 Grace Blackwell Superchip in DGX Spark, which combines a Blackwell GPU and Grace CPU over NVLink-C2C.
On automotive, MediaTek Dimensity Auto platforms integrate NVIDIA technologies for AI and NVIDIA RTX graphics in vehicle cockpits and can work alongside NVIDIA DRIVE AGX. The companies said they will continue that work across future generations.
The release includes standard forward-looking-statement language: commercial outcomes are not guaranteed, and NVIDIA and MediaTek both flagged execution, supply-chain, and market risks.
For teams training or serving large models, the practical news is the combination of a large convertible-bond investment and a documented path to run custom XPUs on NVIDIA’s NVLink rack fabric rather than qualifying an entire scale-up stack in-house. NVIDIA has not, in this announcement, published independent silicon benchmarks for NVHBM-enabled designs.
Source: NVIDIA Newsroom, NVIDIA and MediaTek Deepen Long-Standing Partnership.