Nvidia is making another big move in AI infrastructure. This time, the money is heading to Taiwan.
The chip giant is investing $3.5 billion in MediaTek through convertible bonds. The agreement expands an existing partnership between the two semiconductor companies. But this is not simply a financial investment.
MediaTek will bring Nvidia technology into future custom AI chips. Those chips could sit inside the same data centers that already depend heavily on Nvidia hardware. That gives Nvidia another way to stay central to AI infrastructure. Even when customers start designing more of their own silicon.
Nvidia Is Making Its Biggest Direct Investment Outside the US
Nvidia will purchase $3.5 billion of convertible bonds issued by MediaTek. Bloomberg reported that it will become Nvidia’s largest direct investment outside the United States. The structure gives Nvidia financial exposure to MediaTek while strengthening their technology relationship. MediaTek has become increasingly important as demand grows for custom AI silicon. Its expertise stretches beyond smartphones into high-performance computing and advanced chip design.
MediaTek Will Bring NVLink Fusion Into Custom AI Chips
A major part of the deal revolves around Nvidia’s NVLink Fusion platform. MediaTek plans to use the technology when developing custom AI processors for customers. Those processors can then connect directly with Nvidia’s rack-scale infrastructure. That matters because cloud companies increasingly want chips built for their own workloads. Nvidia does not necessarily need to design every processor. It can still provide the infrastructure connecting everything together.
Nvidia Wants to Stay Inside the Data Center Even When Customers Build Their Own Chips
Big technology companies are spending heavily on custom silicon. Amazon, Google and Microsoft already have their own chip programs. AI companies are exploring similar strategies. The goal is often to reduce dependence on Nvidia’s expensive GPUs. Nvidia appears ready for that shift. By extending NVLink into third-party chips, it can remain part of the system even when another company designs the processor.
MediaTek Is Moving Deeper Into the AI Data Center Market
MediaTek is best known for processors used in smartphones and consumer electronics. That image is changing quickly. The Taiwanese company has been expanding into custom AI chips and data center hardware. Bloomberg reported that MediaTek expects around $2 billion in AI chip revenue this year. The company also wants a larger share of the custom data center silicon market. Nvidia’s investment gives that strategy much more weight.
The Partnership Also Covers Nvidia’s New NVHBM Technology
The expanded agreement goes beyond NVLink Fusion. Bloomberg reported that MediaTek will also work with Nvidia’s newly announced NVHBM technology. The broader goal is better communication between processors, memory and other components inside AI systems. AI data centers are becoming less about one powerful GPU. They increasingly depend on entire racks working as one computing system. Faster connections between components can make a major difference.
Nvidia and MediaTek Are Already Working Together on PCs
The two companies already have a relationship outside large data centers. MediaTek is helping Nvidia expand into AI-powered personal computing. Their collaboration includes multiple generations of RTX Spark and DGX Spark chips. These products combine Nvidia graphics technology with MediaTek system-on-chip expertise. The aim is to bring stronger local AI processing into PCs and workstations.
Automotive AI Is Another Part of the Deal
Cars are also part of the expanded partnership. Nvidia and MediaTek plan to continue developing platforms for software-defined vehicles. Modern vehicles increasingly rely on AI for driver assistance, infotainment and automated functions. Both companies see that market as another major computing platform. MediaTek brings power-efficient chip design. Nvidia brings its AI software and accelerated computing stack.
MediaTek Gives Nvidia Another Route Into Custom Silicon
MediaTek brings something Nvidia increasingly needs: custom chip expertise. Large cloud providers want processors designed around specific AI workloads. Building those chips takes engineering talent, packaging technology and advanced connectivity. MediaTek already works across those areas. Its growing credibility could make it a serious challenger to companies such as Broadcom and Marvell. Nvidia’s investment strengthens that position while tying MediaTek more closely to the Nvidia ecosystem.
Nvidia Is Expanding Beyond the GPU
Nvidia built its AI dominance around GPUs. The company is now spreading that influence across the rest of the data center. Networking, CPUs, interconnects, software and rack-scale systems have become equally important. The MediaTek deal fits that strategy. Nvidia does not need every AI chip to carry its name. It needs those chips to work comfortably inside its infrastructure. That may be the bigger play here.
The Custom AI Chip Boom Does Not Necessarily Hurt Nvidia
Custom chips are often described as a threat to Nvidia. The reality could be messier. More hyperscalers are designing their own accelerators and processors. Yet those chips still need networking, memory systems and data center infrastructure. Nvidia is positioning itself to provide those layers. NVLink Fusion gives outside chipmakers a path into Nvidia-based systems. MediaTek now becomes one of the most important partners in that approach.
Nvidia Is Investing Across the AI Ecosystem
The MediaTek deal also continues another Nvidia trend. The company has been investing across the wider AI industry. Some investments involve customers. Others involve suppliers, startups or potential competitors. Critics have raised questions about whether these relationships create circular demand. Nvidia rejects that argument. Its leadership says the investments help accelerate AI adoption and expand the overall market.
The Bigger Story Is Nvidia’s Grip on AI Infrastructure
The $3.5 billion figure grabs attention. The strategic part matters more.
AI companies increasingly want custom processors. Nvidia cannot stop that trend. Instead, it is making its technology harder to avoid.
MediaTek can design custom silicon. Hyperscalers can build specialized processors. Nvidia can still provide the connections tying those chips into massive AI systems.
That’s a very different kind of dominance.
Nvidia may eventually sell fewer GPUs into some custom-chip workloads. Yet it could still control much of the infrastructure surrounding them. The MediaTek investment shows how seriously Nvidia is preparing for that possibility.
Sources
Bloomberg — Nvidia to Invest $3.5 Billion in Chipmaker MediaTek
https://www.bloomberg.com/news/newsletters/2026-08-31/nvidia-to-invest-3-5-billion-in-taiwan-s-mediatek
Nvidia Newsroom — NVIDIA and MediaTek Deepen Long-Standing Partnership to Build AI Edge to Cloud Computing Platforms
https://nvidianews.nvidia.com/news/nvidia-and-mediatek-deepen-long-standing-partnership-to-build-ai-edge-to-cloud-computing-platforms
