NVIDIA and MediaTek Expand AI Partnership With $3.5 Billion Investment
NVIDIA is investing $3.5 billion in MediaTek through convertible bonds as the companies deepen their long-standing partnership.
MediaTek will adopt NVIDIA NVLink Fusion, giving customers a path to connect custom AI accelerators with NVIDIA’s rack-scale AI infrastructure.
The collaboration also covers local AI computing, AI PCs and automotive platforms, extending from data centers to edge devices and vehicles.
What happens when companies want more control over the chips powering their AI workloads?
That question is becoming increasingly important as AI infrastructure expands beyond conventional GPU deployments. NVIDIA and MediaTek are now taking a significant step in that direction, announcing an expanded partnership that connects NVIDIA’s accelerated computing ecosystem with MediaTek’s expertise in custom silicon and power-efficient system-on-chip design.
Announced on August 31, 2026, the agreement includes a $3.5 billion NVIDIA investment in MediaTek through convertible bonds. More importantly for the AI infrastructure market, MediaTek will adopt NVIDIA’s NVLink Fusion platform to help customers develop custom AI accelerators that can work within NVIDIA-connected, rack-scale AI systems.
The partnership spans three areas: AI infrastructure, local AI computing and automotive technology.
What Did NVIDIA and MediaTek Announce?
At the center of the announcement is an expansion of the existing NVIDIA-MediaTek relationship.
MediaTek will work with the NVIDIA NVLink Fusion ecosystem, which is designed to help hyperscalers, cloud providers and AI developers integrate custom accelerators into NVIDIA-based infrastructure.
In simple terms, companies developing their own AI chips may not need to build the entire surrounding infrastructure from scratch. NVLink Fusion provides a foundation for connecting those custom processors to NVIDIA’s high-speed interconnect and rack-scale systems.
The partnership also continues work on AI computing closer to the user, including PCs, as well as AI-powered automotive platforms.
Why Is NVIDIA Investing $3.5 Billion in MediaTek?
The investment is a strong indication that NVIDIA views MediaTek as a strategically important technology partner.
The two companies bring different strengths to the relationship. NVIDIA has built a broad ecosystem around GPUs, accelerated computing, AI software and high-speed networking. MediaTek, meanwhile, has extensive experience designing SoCs, connectivity technologies, power-efficient chips, advanced packaging and custom silicon.
That combination becomes particularly relevant as AI systems become more specialized.
The investment itself does not guarantee future products, revenue or commercial success. Both companies also highlighted the risks and uncertainties surrounding future development and commercialization.
What Is NVIDIA NVLink Fusion?
NVLink Fusion is essentially a technology foundation for connecting custom AI accelerators to NVIDIA’s AI infrastructure.
An XPU is a general term for a specialized processor designed to accelerate particular computing workloads. Instead of relying exclusively on standard processors, large cloud and AI companies can design custom silicon optimized for their own workloads.
NVLink Fusion is intended to make integrating those processors into larger NVIDIA-based systems easier.
The platform brings together several technologies, including:
Technology | Purpose |
NVLink Fusion | Connects custom AI accelerators to NVIDIA’s NVLink ecosystem |
NVLink-C2C | Provides high-bandwidth chip-to-chip connectivity |
NVHBM | Supports high-bandwidth memory configurations |
NVIDIA MGX | Provides a modular approach to rack-scale infrastructure |
The significance is not simply the custom chip itself. Building a production-ready AI system also requires advanced packaging, memory, networking, high-speed SerDes, I/O and extensive system-level engineering.
By providing more of that surrounding infrastructure, NVIDIA and MediaTek aim to reduce some of the complexity involved in deploying custom AI accelerators at scale.
Why Would Companies Want Custom AI Chips?
Custom AI chips can be attractive when a company has workloads that justify specialized hardware.
For example, a cloud provider could design an accelerator specifically around the AI workloads it runs most frequently. Potential advantages include workload-specific performance, power efficiency, greater hardware control and differentiation.
However, developing custom silicon is expensive and technically demanding. The challenge doesn't end when the chip is designed; it must work reliably with memory, networking, packaging and the rest of the data-center infrastructure.
That is where a platform such as NVLink Fusion could become useful.
Importantly, this does not mean custom XPUs are expected to replace NVIDIA GPUs. Instead, the announcement points toward an AI infrastructure environment where GPUs and specialized accelerators can coexist within larger systems.
Is the Partnership Only About Data Centers?
No. The NVIDIA and MediaTek collaboration extends beyond large AI factories.
Area | What the companies are working on |
AI Infrastructure | Custom AI accelerators and NVLink-connected rack-scale systems |
Local AI | AI PCs, developer systems and workstations |
Automotive | AI-powered, software-defined vehicles |
MediaTek previously collaborated with NVIDIA on the GB10 Grace Blackwell Superchip used in NVIDIA DGX Spark, combining an NVIDIA Blackwell GPU and Grace CPU through NVLink-C2C.
The companies are also extending their work into the next generation of consumer PCs through NVIDIA RTX Spark.
In automotive, MediaTek’s Dimensity Auto platforms integrate NVIDIA technologies for AI and graphics, while the companies continue developing solutions alongside NVIDIA DRIVE AGX for increasingly software-defined vehicles.
What Could This Mean for the AI Hardware Market?
The partnership reflects a broader shift in AI infrastructure toward specialized and heterogeneous computing.
As AI workloads become more diverse, organizations may increasingly look beyond a single type of processor. Custom accelerators, GPUs, high-bandwidth memory, chiplets and high-speed interconnects can all play different roles within the same infrastructure.
For NVIDIA, working with MediaTek could broaden its ecosystem around custom AI silicon while keeping NVIDIA connectivity and infrastructure at the center.
For MediaTek, the partnership creates a deeper connection to one of the most important AI computing ecosystems while allowing it to bring its custom silicon and SoC expertise into higher-performance AI infrastructure.
The actual commercial impact will depend on customer adoption and successful product development, but the direction is clear: AI infrastructure is becoming more interconnected and specialized.
What Does This Mean for Businesses?
For organizations running large AI workloads, the development could eventually mean more options for how computing infrastructure is designed.
Custom accelerators can potentially be optimized around specific workloads, while established infrastructure technologies can help connect those accelerators into larger AI systems.
For businesses, the important takeaway isn't simply the $3.5 billion investment. It is the possibility of more flexible AI infrastructure combining general-purpose accelerated computing with specialized silicon.
Conclusion
The NVIDIA and MediaTek partnership is becoming broader than a conventional chip collaboration.
With NVIDIA’s $3.5 billion investment, MediaTek’s adoption of NVLink Fusion, and continued work across AI PCs and automotive systems, the two companies are targeting multiple layers of the AI computing stack.
The bigger question is whether the next generation of AI infrastructure will be built around a single dominant processor type, or around systems that combine powerful GPUs with increasingly specialized custom accelerators. This partnership suggests NVIDIA and MediaTek are preparing for the latter.
