Nvidia’s $3.5B MediaTek bet reveals its plan for tackling Big Tech’s AI chip buildout

Nvidia is spending $3.5 billion on MediaTek, and the money comes with a clear trade: the Taiwanese chipmaker will use Nvidia’s technology to help design custom AI chips for cloud companies and model builders. Those chips are meant to plug straight into Nvidia-based data centers. That’s not a small detail. It’s the whole point.

The timing makes sense for Nvidia, even if the message is a little ironic. Amazon, Google, Microsoft, OpenAI and Anthropic are all pushing to build more of their own silicon so they don’t have to lean so hard on Nvidia’s GPUs. Instead of fighting that shift head-on, Nvidia is trying to sit in the middle of it. Let the customers build custom chips, sure. Just keep Nvidia’s hardware and networking fabric underneath the whole stack.

That fabric now includes NVLink Fusion, which MediaTek will get access to. NVLink is the part that lets chips communicate quickly, even if they aren’t Nvidia chips. Nvidia says this helps customers standardize on rack-scale infrastructure across their AI factories, while still deploying custom chips alongside Nvidia systems. Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions, put it bluntly: Nvidia is an AI infrastructure company, not just a chip maker.

The partnership also shows how circular Nvidia’s business has become. It has spent years investing in companies that end up feeding back into its own ecosystem, and this deal fits that pattern neatly. Last week, Nvidia announced a similar partnership with Amazon Web Services, though without the direct investment. AWS will deploy another 2 million Nvidia GPUs and also integrate NVLink Fusion.

For MediaTek, the upside is straightforward. The company has been building its custom data center ASIC business and said in June that it expects that line to generate $2 billion in revenue in 2026. It already has experience making custom chips for smartphones, smart homes, cars and wireless gear, and Nvidia wants that reach applied to AI. The two companies will also keep working together on DGX Spark, RTX Spark and AI-powered software-defined vehicles.