Nvidia Invests $3.5 Billion to Connect MediaTek Chips to NVLink
Nvidia is investing $3.5 billion in MediaTek as it expands a partnership aimed at bringing MediaTek’s custom AI chips into Nvidia’s data-center infrastructure. Chief Executive Jensen Huang says MediaTek’s SoCs and XPUs will connect to NVLink Fusion, Spectrum-X networking and Nvidia system designs, allowing customers to use custom compute without building the surrounding architecture independently. MediaTek CEO Rick Tsai argues the combination gives customers a faster, more flexible way to scale AI systems across hyperscale, enterprise and desktop deployments.
Nvidia wants custom-chip wins to pull through its own infrastructure
Nvidia’s deeper relationship with MediaTek is designed to make Nvidia’s systems relevant even when a customer’s core compute is custom silicon. The central commercial logic, as Jensen Huang describes it, is not simply that MediaTek can connect its chips to Nvidia GPUs. It is that a MediaTek win can bring Nvidia’s NVLink, networking and system architecture into a deployment—and Nvidia’s existing presence in clouds and data centers can give MediaTek’s chips a path into those environments.
The on-screen announcement says Nvidia will invest $3.5 billion in MediaTek. Huang characterized the agreement as a major expansion of an existing partnership, which began with MediaTek’s system-on-chips, or SoCs. Nvidia integrated its NVLink chip-to-chip interface with those SoCs, Huang said, allowing a MediaTek SoC to connect to an Nvidia GPU.
The expansion applies the same premise to MediaTek’s XPU business. Huang said Nvidia will connect MediaTek XPUs to NVLink Fusion, Nvidia’s scale-up NVLink system, Spectrum-X switches, networking portfolio, and system chassis. The aim, in Huang’s account, is to let MediaTek compute join Nvidia-based data-center systems rather than operate as an isolated alternative.
Huang said Nvidia’s footprint in clouds and data centers means a MediaTek XPU can connect into those environments “in a really seamless way.” The partnership is therefore presented as reciprocal distribution: MediaTek’s custom-compute projects can pull Nvidia networking and systems into an account, while Nvidia’s own wins can create openings for MediaTek.
The stated customer case is speed and flexibility across the AI stack
Ed Ludlow framed the customer question directly: an ASIC customer has built its own accelerator, so why use Nvidia’s NVLink architecture rather than independently assemble the rest of the system?
Rick Tsai answered by placing the collaboration across what he called the full AI-computing stack, from hyperscalers’ ASICs through enterprise data centers and new clouds to desktop supercomputers for AI applications. MediaTek’s XPU capability combined with Nvidia’s NVLink Fusion, he said, can serve customers across those environments.
Tsai’s emphasis was not on replacing a customer’s custom accelerator. It was on enabling customers to build and scale data centers quickly while preserving flexibility in an AI market he described as rapidly changing. His account makes the partnership a joint system proposition: MediaTek supplies XPU capability, while Nvidia supplies the interconnect and broader infrastructure around it.
For Tsai, speed has both a technical and operational meaning. He said the companies aim to help customers scale “with high speed”—invoking Huang’s phrase “speed of light”—while working with customers with “tremendous flexibility.” The value claimed is not simply chip-to-chip connectivity, but a route for customers with varied architectures to assemble and expand AI systems.
Desktop AI and robotics broaden the partnership’s reach
The earlier collaboration produced DGX Spark, a small desktop computer that Huang described as a one-petaflop system capable of running advanced agentic AI on a desk rather than in the cloud.
Huang said Nvidia and MediaTek will develop multiple generations of the processor used in that system. He also said it will be the foundation of a new computer from Microsoft and Nvidia, which he described as a reinvention of the Windows PC for the age of agents.
He grouped those desktop systems with work in robotic systems, including autonomous vehicles. Huang did not offer technical detail on the robotics effort, but his examples position the relationship beyond data-center integration: across local AI computing, larger-scale infrastructure, and autonomous or embedded systems.