This week, NVIDIA announced a $3.5 billion investment in MediaTek through convertible bonds. The money is only half the story. On the tech side, MediaTek will now offer NVIDIA's full NVLink Fusion platform – including the Fusion chiplet, NVLink-C2C, and the new NVHBM memory protocol – directly to its custom silicon customers.

Why this matters:

Hyperscalers like Google, Meta, and OpenAI are all building their own AI chips. NVIDIA's move is a smart response. Instead of fighting that trend, the company is embracing it. By putting NVLink Fusion into MediaTek's hands, NVIDIA makes it easier for any custom XPU designer to plug directly into NVIDIA's rack-scale hardware ecosystem.

The key enabler is the NVLink Fusion chiplet. It acts as a plug-and-play bridge, giving custom chips full NVLink fabric support without requiring designers to license the core IP from scratch.

For those following Spark devices, NVIDIA confirmed MediaTek will remain involved through 2030, covering the Vera Rubin and Rosa Feynman generations. That collaboration is set to continue.

The bigger picture:

NVIDIA is generating nearly $60 billion in quarterly profits. The company is using that financial firepower to lock in key partners and build the backbone of the AI industry – one strategic investment at a time.