Nvidia Bets $3.5 Billion on MediaTek to Anchor Data Center Ecosystem
AI-generated image · US National Wire
By integrating custom silicon into its rack-scale architecture, Nvidia aims to keep hyperscalers within its infrastructure even as they build their own chips.
Nvidia is investing $3.5 billion into Taiwanese chipmaker MediaTek in a strategic move to maintain its dominance over AI data center infrastructure, as TechCrunch first reported.
As major cloud providers and AI labs—including Microsoft, Google, Amazon, OpenAI, and Anthropic—invest in proprietary silicon to reduce their reliance on Nvidia GPUs, the chip giant is shifting its focus toward the underlying "scaffolding" of the data center. Through this partnership, MediaTek will utilize Nvidia's technology to design custom chips for hyperscalers that plug directly into Nvidia-based data centers.
A central component of the deal is access to the NVLink Fusion ecosystem. This technology allows disparate chips, including those not manufactured by Nvidia, to communicate rapidly. Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions, told TechCrunch that this allows customers to standardize on Nvidia's rack-scale infrastructure while deploying custom chips alongside the platform.
This strategy mirrors a recent partnership with Amazon Web Services (AWS), which will integrate NVLink Fusion and deploy 2 million additional Nvidia GPUs, though that deal did not include a direct investment.
Beyond the data center, the two companies are expanding collaborations on consumer and automotive hardware. This includes the RTX Spark for AI PCs, the DGX Spark desktop AI computer, and software-defined vehicle platforms utilizing Nvidia Drive AGX and RTX graphics. MediaTek previously stated in June that it expects its custom data center ASIC business to generate $2 billion in revenue in 2026.