Nvidia invests $3.5 billion in MediaTek to secure AI infrastructure dominance
The strategic partnership integrates MediaTek's custom chip design with Nvidia's NVLink Fusion to hedge against Big Tech's in-house silicon trend.
Nvidia has invested $3.5 billion into Taiwanese chipmaker MediaTek via convertible bonds to integrate custom silicon capabilities into its data center ecosystem. The move signals a strategic shift for Nvidia as it seeks to remain the primary infrastructure provider for the AI era.
Under the terms of the partnership, MediaTek will adopt Nvidia's NVLink Fusion platform. This technology allows custom AI chips, often referred to as XPUs, to communicate rapidly and integrate directly into Nvidia-based rack-scale data center systems. Beyond the data center, the collaboration extends into local AI computing through the RTX Spark and DGX Spark lines, as well as the automotive sector.
The shift to AI infrastructure
This investment comes as major cloud providers and AI labs—including Amazon, Google, Microsoft, OpenAI, and Anthropic—increasingly develop proprietary AI chips to reduce their heavy reliance on Nvidia's GPUs. By opening its ecosystem to third-party custom silicon, Nvidia is repositioning itself from a pure hardware vendor to a comprehensive infrastructure company.
"Nvidia is an AI infrastructure company," said Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions. This transition allows Nvidia to provide the essential "scaffolding" of the data center, ensuring that even when Big Tech firms design their own compute silicon, those chips must still operate within Nvidia's architectural framework to function efficiently at scale.
Industry implications
The deal represents a calculated hedge against the trend of in-house chip development. By controlling the interconnects and the rack-scale architecture via NVLink Fusion, Nvidia ensures it remains the indispensable layer of the AI stack. If custom chips are designed to fit Nvidia's standards, the company maintains its grip on the data center's physical and logical layout, regardless of which company designs the actual processor.
What to watch
Industry analysts will now look for the first deployments of MediaTek-designed XPUs within Nvidia racks to see if the NVLink Fusion integration delivers the promised performance gains. While the strategic framework is set, the long-term success of this pivot depends on whether hyperscalers are willing to adhere to Nvidia's infrastructure standards rather than building entirely independent, end-to-end proprietary stacks.