Nvidia is investing $3.5 billion in Taiwan’s MediaTek as the chip giant deepens a strategy of embracing custom artificial intelligence processors that could otherwise threaten its dominance of the data center market.
The investment will give MediaTek access to Nvidia’s NVLink Fusion technology, enabling chips developed by MediaTek and its customers to connect directly with Nvidia-powered AI infrastructure. The partnership is designed to allow cloud providers and AI companies to deploy their own custom processors alongside Nvidia GPUs within the same data center systems.
The deal comes as some of Nvidia’s largest customers, including Amazon, Google, Microsoft, OpenAI and Anthropic, are developing proprietary AI chips to reduce their dependence on Nvidia’s expensive and highly sought-after GPUs. Rather than resisting that shift, Nvidia is positioning itself as the infrastructure layer that connects different types of processors.
The strategy allows customers to develop specialized chips for particular AI workloads while continuing to rely on Nvidia’s networking, software, and rack-scale systems.
“Nvidia is an AI infrastructure company,” Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions, told reporters on Monday. “We expanded beyond pure computing chips years ago.”
The partnership therefore matters beyond the size of Nvidia’s investment. It represents an attempt to make Nvidia’s technology the common platform around which competing processors operate.
NVLink Fusion allows non-Nvidia chips to communicate at high speed with Nvidia processors. That could give hyperscalers greater flexibility to introduce custom accelerators without having to abandon Nvidia’s broader data center architecture.
“Basically, every cloud, every model builder is deploying our platform in some shape, form, or fashion,” Harris said. “So by MediaTek being able to offer this extension to its customers, it allows them to standardize on the rack-scale infrastructure across their AI factories.”
For Nvidia, that is an important distinction. The company may lose some of the computing workload to custom ASICs, or application-specific integrated circuits, but it can still capture value from the surrounding infrastructure.
The approach also helps explain Nvidia’s expanding investment activity across the AI ecosystem. The company has increasingly backed businesses that could become customers, partners or complementary suppliers within its platform, creating a model in which Nvidia’s capital helps expand demand for the infrastructure it sells.
The MediaTek investment comes shortly after Nvidia announced a similar expansion of its relationship with Amazon Web Services. AWS plans to deploy an additional 2 million Nvidia GPUs and integrate NVLink Fusion into its infrastructure, although that agreement does not include a direct Nvidia investment in AWS.
MediaTek brings a different advantage to the relationship. The Taiwanese company has extensive experience designing specialized chips for smartphones, automobiles, wireless communications and consumer electronics, and has been expanding into custom data center silicon.
MediaTek said in June that it expects its custom data center ASIC business to generate $2 billion in revenue in 2026. Nvidia’s technology could help the company compete for a larger share of the rapidly expanding market for custom AI accelerators, particularly as hyperscalers seek processors tailored to their own workloads.
The opportunity is growing because AI companies increasingly need different types of computing for training, inference, and specialized applications. Instead of relying entirely on a single general-purpose accelerator, large technology companies can combine GPUs with custom silicon designed for specific workloads.
That creates a potential opening for MediaTek while also giving Nvidia a way to remain central to systems in which its own chips are no longer the only processors.
“This is really about opening up this ecosystem to the entire MediaTek customer base,” Harris said.
The partnership extends beyond data centers. Nvidia and MediaTek will continue working together on DGX Spark, Nvidia’s developer-focused desktop AI computer, as well as the RTX Spark initiative aimed at consumer AI PCs.
The companies are also collaborating on software-defined vehicles and automotive AI. MediaTek’s automotive platforms use Nvidia RTX graphics technology for intelligent vehicle cockpits and work alongside Nvidia’s Drive AGX platform for autonomous-driving workloads.
The broader strategy reflects Nvidia CEO Jensen Huang’s effort to turn the company from a supplier of high-performance GPUs into a much broader computing platform spanning data centers, PCs and vehicles.
“AI is transforming every computing platform, from the world’s largest AI factories to the PC and the car,” Huang said. “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.”
The challenge for Nvidia is that the same strategy designed to preserve its dominance could also accelerate the development of alternatives to its GPUs. If custom silicon becomes more capable and widely adopted, Nvidia will need to ensure that its networking, software and rack-scale architecture remain indispensable even when customers use processors supplied by competitors.
Industry analysts see the MediaTek deal as an indication that Nvidia is betting that controlling the architecture around AI computing could ultimately prove as important as controlling the processors themselves.






