Nvidia Pours $3.5B Into MediaTek as AI Chip Alliance Deepens

Nvidia Corp. has made a $3.5 billion investment in MediaTek Inc. through convertible bonds, a move announced alongside an expanded technical collaboration between the two chipmakers. Convertible bonds function as debt instruments that investors can later exchange for equity stakes. Such financial commitments are typical for Nvidia when deepening ties with technology partners; the company invested $2 billion in Marvell Technologies Inc. shares earlier this year.

MediaTek, headquartered in Taiwan, primarily manufactures processors for smartphones and internet-connected devices. The company also operates a custom chip design division focused on artificial intelligence accelerators, which represents the initial thrust of this expanded partnership with Nvidia.

AI Accelerator Integration

Customers commissioning MediaTek to design AI accelerators will gain access to a selection of on-chip networking solutions. These technologies enable processors to communicate with other chips within the same equipment rack or transmit information across multiple racks. Going forward, Nvidia's NVLink Fusion networking technology will be available as an interconnect choice for MediaTek's custom designs.

Data centers running AI workloads house not only graphics processing units but also central processing units that handle auxiliary functions like data preparation for AI models. NVLink Fusion allows Nvidia's GPUs and CPUs to communicate with third-party GPUs and CPUs respectively. A key advantage of this technology lies in its performance: NVLink, which forms the foundation of NVLink Fusion, transfers data at speeds exceeding standard PCIe 6.0 implementations by more than tenfold.

Nvidia recently enhanced NVLink Fusion by incorporating support for NVHBM, a novel memory architecture. In conventional designs, a small memory controller chip manages the HBM memory housed on graphics cards. NVHBM relocates this controller from the GPU's primary logic die to the die containing HBM memory itself. According to Nvidia, this architectural shift improves data transfer rates while reducing energy demands.

Personal Computing Expansion

The partnership's second pillar targets the personal computing sector. Nvidia introduced the DGX Spark workstation in 2025, a system designed for developers performing tasks like AI model fine-tuning. MediaTek contributed to designing the system-on-chip powering this device. The collaboration subsequently yielded the RTX Spark, a processor for laptops and desktops that launched in June.

Both companies have committed to collaborating on successive generations of the RTX Spark and DGX Spark processors. While MediaTek's specific responsibilities in future iterations remain unspecified, the current generation provides insight into their contributions. MediaTek engineered the onboard CPUs and additional components, including the RTX Spark's memory controller.

Automotive Sector Focus

The expanded partnership also encompasses the automotive industry. MediaTek distributes a lineup of systems-on-chip branded Dimensity Auto, which automotive manufacturers integrate into in-vehicle display systems. These processors incorporate an Nvidia graphics rendering chiplet among other components. The companies announced plans to maintain their collaboration on software-defined vehicle chips.

Together, we're building platforms that bring Nvidia accelerated computing to new markets and give customers the freedom to create differentiated AI systems at enormous scale

Nvidia Chief Executive Officer Jensen Huang

For MediaTek, this deepened alliance reinforces one of its most dynamic business segments. The company recently raised its annual revenue projection for its data center chip design division to $2 billion, doubling the previous forecast. Management envisions expanding this figure substantially over the coming years.