Nvidia invests $3.5bn in MediaTek to extend AI infrastructure
Nvidia is investing $3.5 billion in Taiwanese chipmaker MediaTek, pairing the financial deal with access to its NVLink Fusion ecosystem. The arrangement is intended to let MediaTek design custom AI chips for cloud providers and AI companies that can be deployed directly in Nvidia-based data centres.
MediaTek will be able to use NVLink, Nvidia’s technology for fast communication between chips, alongside the company’s scale-up, scale-out and rack-scale infrastructure technologies. Nvidia said the goal is to let MediaTek customers place custom silicon alongside Nvidia-based systems on a common platform.
A response to the custom silicon shift
The deal comes as Amazon, Google, Microsoft, OpenAI and Anthropic invest in their own chips, seeking to reduce reliance on Nvidia GPUs for particular workloads. Rather than attempting to prevent that trend, Nvidia is offering the connective infrastructure that can make third-party chips part of an Nvidia-centred data-centre deployment.
The strategic tension is illustrated by hyperscalers seek alternatives while needing interoperable infrastructure as hyperscalers seek alternatives while still requiring interoperable AI infrastructure. Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions, described Nvidia as an AI infrastructure company that had expanded beyond pure computing chips years ago.
Nvidia says MediaTek can extend the standard rack-scale platform to its customers while allowing them to deploy their own custom chips alongside other components. The approach keeps Nvidia involved in the architecture and ecosystem even where the accelerator itself is not made by Nvidia.
MediaTek’s data-centre ambitions
MediaTek has been building its custom data-centre ASIC operation and said in June that it expects the business to generate $2 billion in revenue in 2026. It has not publicly detailed the relevant customer base, but Nvidia’s investment and technology access support MediaTek’s effort to address more of the market.
The companies will also continue their collaboration on DGX Spark, Nvidia’s developer-focused desktop AI computer, and RTX Spark for consumer AI PCs. Their work also includes platforms for AI-powered, software-defined vehicles. MediaTek’s automotive platforms use Nvidia RTX graphics for intelligent vehicle cockpits and work with Nvidia Drive AGX for autonomous-driving workloads.
What technology buyers should examine
For businesses planning AI capacity, the agreement highlights an architectural choice beyond selecting a single accelerator supplier. Buyers using bespoke chips should assess how networking, rack design and software compatibility will operate across mixed hardware. A shared infrastructure layer may make custom silicon easier to introduce, but procurement teams still need to validate the fit of each component with their workloads and deployment model.