Nvidia invests $3.5 billion in MediaTek to expand their partnership into custom AI infrastructure using NVLink Fusion.
Nvidia and MediaTek have announced a major expansion of their partnership, with Nvidia investing $3.5 billion in convertible bonds issued by MediaTek. In return, MediaTek will adopt Nvidia’s NVLink Fusion platform across its custom AI accelerators, local AI systems, and automotive platforms. This integration allows MediaTek to design accelerators compatible with Nvidia’s fully developed rack-scale architectures, while giving Nvidia a strategic foothold in the expanding market for custom AI hardware.
As the world’s largest supplier of AI accelerators, Nvidia faces no direct competitors of comparable scale. Yet, as enterprises increasingly turn to bespoke solutions, Nvidia must hedge against the rise of custom AI chips and ensure its addressable market grows regardless of whether it wins every accelerator design. Spreading the NVLink Fusion platform beyond its own silicon and popularizing it among custom-hardware developers is a key part of that strategy, and the MediaTek agreement directly supports this objective.
Hyperscalers and tech giants—including AWS, Google, Meta, Microsoft, and most recently OpenAI—are developing proprietary AI accelerators to reduce reliance on costly commercial GPUs. While Nvidia cannot halt this trend, NVLink Fusion offers a counter-strategy: if customers substitute some Nvidia GPUs with their own XPUs, Nvidia ensures those accelerators remain connected via NVLink, integrated with Nvidia CPUs where applicable, and deployed within Nvidia’s networking and rack architectures. Under the new framework, clients can introduce their own XPU designs to MediaTek, after which both companies will provide the surrounding ecosystem, including connectivity, memory, packaging, and manufacturing technologies.
MediaTek will leverage NVLink Fusion as the foundational layer for custom accelerators designed to scale alongside future Nvidia architectures. The platform comprises the NVLink Fusion chiplet, linking custom XPUs to Nvidia’s NVLink scale-up fabric via electrical or photonic interconnects; NVLink-C2C, delivering high-bandwidth, energy-efficient connections between XPUs, Nvidia Rosa CPUs, and other compatible processors; and Nvidia NVHBM, which supports customizable memory configurations while freeing additional silicon area for compute logic.
By adopting NVLink Fusion, MediaTek’s clients can focus exclusively on differentiating their core compute architectures while relying on Nvidia and MediaTek for the surrounding infrastructure. This delivers a ready-made, rack-scale deployment environment for custom AI accelerators, a capability unavailable through alternative routes. Given Nvidia’s broader focus on supplying complete AI infrastructure ecosystems rather than standalone chips, the partnership aligns seamlessly with its long-term strategy.
“MediaTek is one of the world's great semiconductor companies, with exceptional expertise in system-on-chip design, connectivity, leading performance and power efficiency,” said Jensen Huang, founder and CEO of Nvidia. “Together, we are building platforms that bring Nvidia accelerated computing to new markets and give customers the freedom to create differentiated AI systems at enormous scale.”
Beyond custom accelerators, the partnership extends to local AI computing. Building on a successful prior collaboration involving the GB10 Grace Blackwell Superchip that powers the DGX Spark, Nvidia and MediaTek will now co-develop multiple generations of RTX Spark and DGX Spark processors. These next-generation chips will target client devices, AI developer supercomputers, and enterprise workstations.
The agreement also solidifies a multi-generation automotive partnership. MediaTek’s Dimensity Auto platforms already integrate Nvidia AI technologies and RTX graphics to power intelligent vehicle cockpits, operating in tandem with Nvidia Drive AGX. Subsequent iterations will further combine MediaTek’s automotive system-on-chip expertise with Nvidia’s accelerated computing, AI, graphics, and software stacks to develop increasingly advanced software-defined and AI-driven vehicles.