NVIDIA and MediaTek expand AI partnership with $3.5bn bond investment
NVIDIA and MediaTek have announced a broad expansion of their existing partnership, covering rack-scale AI infrastructure, edge computing and automotive platforms. Alongside the technical collaboration, NVIDIA has taken a $3.5 billion position in MediaTek convertible bonds, marking one of the largest strategic investment commitments NVIDIA has publicly disclosed in a chip industry peer.
The centrepiece of the technical agreement is MediaTek's adoption of the NVIDIA NVLink Fusion platform, which provides hyperscalers, cloud service providers and frontier model developers with a prevalidated path to design custom XPUs (accelerator chips) that connect into NVIDIA's NVLink-based, rack-scale AI factories. Rather than engineering NVLink connectivity, memory architecture and packaging from scratch, customers can work through MediaTek to bring differentiated compute designs to production while relying on the shared infrastructure the two companies are building around NVLink Fusion.
Three pillars: infrastructure, edge and automotive
The collaboration is structured across three distinct workstreams. In AI infrastructure, MediaTek will serve as a design and integration partner for custom XPU development under the NVLink Fusion ecosystem. The platform bundles NVLink chiplet interconnects, NVLink-C2C die-to-die connectivity, and NVIDIA's NVHBM customised memory architecture, giving customers a prequalified stack they can tune for performance and power without building chip-to-rack integration from scratch.
In local AI computing, the partnership extends existing joint work on the NVIDIA DGX Spark product line. MediaTek contributed to the GB10 Grace Blackwell Superchip at the heart of DGX Spark, combining an NVIDIA Blackwell GPU and Grace CPU over NVLink-C2C. The companies are now extending that collaboration to the RTX Spark line, targeting consumer PCs and enterprise workstations being repositioned for on-device generative and agentic AI workloads.
Automotive represents the third workstream. MediaTek's Dimensity Auto SoC line already integrates NVIDIA RTX graphics and can work alongside NVIDIA DRIVE AGX platforms for AI-driven vehicle cockpits. Both companies signalled they intend to extend this architecture through multiple future generations.
Jensen Huang, founder and chief executive of NVIDIA, described MediaTek as "one of the world's great semiconductor companies" and said the collaboration would "bring NVIDIA accelerated computing to new markets and give customers the freedom to create differentiated AI systems at enormous scale." Rick Tsai, vice chairman and chief executive of MediaTek, said NVIDIA's investment "strengthens a collaboration that spans cloud AI infrastructure, local AI computing and automotive."
Market and competitive context
The NVLink Fusion announcement positions NVIDIA to compete more directly with alternative fabric and chiplet ecosystems, notably the UALink consortium backed by AMD, Google, Intel, Broadcom and others, which is developing an open interconnect standard intended to reduce dependence on NVLink for scale-up AI clusters. By bringing a major fabless SoC house into the NVLink Fusion ecosystem, NVIDIA is enlarging the gravitational pull of its proprietary interconnect at the moment when hyperscaler custom silicon programmes are accelerating rapidly.
MediaTek's positioning is also notable. The company is a dominant supplier of SoCs for mobile handsets, smart televisions and Wi-Fi chipsets, but its profile in data-centre and AI-factory silicon is considerably smaller than peers such as Broadcom or Marvell, both of which have established XPU design partnerships with major hyperscalers. The NVIDIA investment and joint NVLink Fusion go-to-market could accelerate MediaTek's entry into that tier.
The $3.5 billion convertible bond structure means NVIDIA gains exposure to MediaTek's equity upside without a controlling stake, a structure that avoids the antitrust scrutiny a full acquisition would attract given the concentration already present in the AI accelerator market. Regulatory bodies including the UK CMA and the European Commission have signalled heightened interest in semiconductor consolidation; a bond rather than a share acquisition sidesteps that review threshold for now.
Both parties included standard forward-looking statement disclaimers in their respective sections of the release, noting that commercial outcomes depend on customer adoption, manufacturing capacity and broader macroeconomic conditions.