Nvidia Pours $3.5B Into Mediatek Convertible Bonds As Ai Tie-Up Expands

Deal at a glance

Nvidia has invested $3.5 billion in convertible bonds issued by MediaTek, the companies announced on Monday, 31 August. The move comes alongside an agreement to deepen a long-running partnership as both chipmakers push into artificial intelligence infrastructure, personal computing and automotive systems.

Nvidia described the investment as part of a broader collaboration that gives MediaTek a larger role in its AI ecosystem while enabling Nvidia to extend its architecture into customised computing platforms for cloud providers, hyperscalers and AI developers.

What NVLink Fusion brings to the table

MediaTek will adopt Nvidia’s NVLink Fusion platform to help customers build custom AI accelerators that can connect with Nvidia-powered rack-scale systems. Nvidia says NVLink Fusion provides a prebuilt, system-validated foundation for multi-die XPU development, bringing together high-speed connectivity, memory and advanced packaging capabilities.

The goal is to reduce the engineering lift required to move a bespoke AI chip from prototype to deployment in rack-scale environments, letting customers tailor connectivity, memory, packaging, performance and power characteristics to specific workloads.

"This partnership helps make advanced AI computing more accessible across the cloud, local devices and the car — letting partners move faster from chip design to full rack-scale systems," said Jensen Huang, Nvidia's founder and CEO.

How NVLink Fusion works in practice

NVLink Fusion is designed as a validated starting point for companies building specialised processing units (XPUs). By providing a tested connectivity and packaging stack, it aims to shorten development time for custom accelerators that must interoperate with Nvidia’s GPUs and rack architectures.

For customers, the platform is a way to combine MediaTek’s system-on-chip expertise with Nvidia’s interconnect and system-level engineering to meet demanding AI workloads more quickly.

Financials – the $3.5 billion convertible bonds

Nvidia said its $3.5 billion purchase of MediaTek convertible bonds is part of the expanded collaboration, but the announcement did not include further details on the securities’ terms. Convertible bonds typically start as debt and can be converted into equity under specified conditions, allowing the holder to participate in upside if the issuer’s stock rises.

The timing of the investment comes as semiconductor firms and cloud providers compete to secure the building blocks for generative and agentic AI systems. For Nvidia, the deal reinforces a strategy of extending its technology ecosystem beyond its own processors; for MediaTek, it opens a path into custom AI infrastructure markets.

AI PCs and automotive plans

The companies will continue co-developing multiple generations of chips for local AI computing, extending work on systems that pair Nvidia GPUs with MediaTek system-on-chip technology. Their prior collaboration included MediaTek’s contribution to the GB10 Grace Blackwell Superchip used in Nvidia’s DGX Spark systems, and the partnership will now extend to Nvidia RTX Spark and future consumer and workstation products.

They are also advancing platforms for AI-powered, software-defined vehicles. MediaTek’s Dimensity Auto platforms already integrate Nvidia technologies for AI and RTX graphics and are designed to operate alongside Nvidia DRIVE AGX. The companies said they will keep building on that architecture in future automotive platform generations.

Why this matters

The expanded tie-up signals how major chipmakers are reshaping alliances to meet the infrastructure needs of advanced AI models. By combining Nvidia’s interconnects and system validation with MediaTek’s SoC design strengths, the partnership aims to offer customers more turnkey routes from custom chip design to deployment at scale.

For the broader market, the deal highlights a trend toward modular, collaborative ecosystems where companies share validated building blocks – from cloud racks to local PCs and cars – to accelerate AI rollouts.

About the author

Sayantani Biswas is an assistant editor at Livemint. She has covered geopolitics, international relations and global power dynamics for several years and joined Mint in 2021 after reporting for other publications.

Biswas holds an MPhil in Comparative Literature from Jadavpur University and specialises in historically grounded analysis of contemporary issues. Outside the newsroom, she follows cultural history and critical theory.