Nvidia Invests $3.5 Billion in MediaTek as the AI Chip Race Moves Beyond GPUs

Nvidia is deepening its relationship with Taiwan’s MediaTek with a $3.5 billion investment through convertible bonds, according to Reuters. The transaction is another sign that the AI chip race is expanding well beyond the question of who sells the fastest GPU.

MediaTek is already a major designer of chips for smartphones and connected devices. Its growing work with Nvidia pushes the partnership toward a more strategic part of the AI market: custom silicon designed for large technology companies and hyperscale data centres.

Why Nvidia wants more than GPU dominance

Nvidia’s extraordinary position in AI has been built around accelerators and the software ecosystem surrounding them. But the largest cloud companies increasingly want specialised chips tuned to their own workloads, power budgets and infrastructure.

TechCrunch reports that MediaTek is adopting Nvidia technology as it develops custom chips for AI companies and hyperscalers that can integrate with Nvidia-powered data-centre systems. That turns Nvidia from a component supplier into something closer to an infrastructure platform.

Custom silicon changes the competitive map

Google, Amazon, Microsoft and other large buyers have invested heavily in their own accelerators or custom silicon. The logic is simple: at hyperscale, even modest improvements in performance per watt or cost per workload can translate into enormous savings.

Nvidia therefore has an incentive to participate in that custom-chip wave rather than treat it purely as a threat. A tighter MediaTek relationship could help Nvidia stay embedded in the systems surrounding those chips even when the processor itself is not a standard off-the-shelf GPU.

The AI stack is getting wider

The current infrastructure race includes processors, memory, networking, packaging, power delivery, cooling, data centres and increasingly custom chip design. The winners may be companies that control several layers at once or make their technology difficult to replace across the whole stack.

MaryChuks analysis

This investment is a reminder that AI scaling is becoming an ecosystem problem. Nvidia does not need every future AI chip to carry the same label if the surrounding architecture still depends on Nvidia technology. That is how a hardware lead can become an infrastructure moat.

Sources: Reuters and TechCrunch reporting on Nvidia’s MediaTek investment, August 31, 2026.