Nvidia is spending $3.5 billion to ensure its interconnect standard becomes the backbone of custom AI silicon built by hyperscalers and cloud providers.
返回
Nvidia invests $3.5 billion in MediaTek, deepens NVLink Fusion tie-up
Nvidia is spending $3.5 billion to ensure its interconnect standard becomes the backbone of custom AI silicon built by hyperscalers and cloud providers.
Nvidia Corp. is investing $3.5 billion in MediaTek Inc. through convertible bonds, deepening a partnership that will see the Taiwanese chipmaker offer Nvidia's NVLink Fusion platform to customers building custom AI accelerators for data centers.
"AI is transforming every computing platform — from the world's largest AI factories to the PC and the car," Nvidia founder and CEO Jensen 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."
MediaTek will offer customers the NVLink Fusion platform, which combines NVLink connectivity, high-bandwidth memory, advanced packaging, and rack-scale integration. This includes NVLink-C2C, the high-bandwidth fabric also used in Nvidia's upcoming "Rosa" CPUs, and NVHBM, which integrates the memory controller into the 3D HBM stack. Nvidia claims NVHBM delivers up to 30 percent more memory bandwidth, 15 percent lower HBM power consumption, and 25 percent more usable compute die area compared with standard HBM4E.
The deal extends Nvidia's push to keep its interconnect technology at the center of AI data center build-outs as hyperscalers increasingly design their own silicon. Amazon.com Inc. recently agreed to deploy 2 million additional Nvidia components and pledged to run Nvidia's connection technology alongside its in-house chips. MediaTek, whose market value has roughly tripled in recent months, is competing with Broadcom Inc. and Marvell Technology Inc. for custom ASIC design business.
NVLink Fusion opens the custom silicon door
MediaTek will offer the NVLink Fusion platform to its customers, which include numerous ASIC clients like Google, enabling them to connect their XPUs to the NVLink scale-up fabric using photonic or classical copper interconnects. Customers can hand off the entire ASIC design and manufacturing process to MediaTek, which will use Nvidia and NVLink technologies for connectivity, memory architecture, packaging, manufacturing, and rack-scale integration to expedite accelerator development. Shorter design cycles will lead to faster prototypes and production runs, and customers can bring their own XPU designs to MediaTek for integration with NVLink Fusion technologies.
For Nvidia, the agreement is the latest effort to ensure the future of AI is built around its hardware platform and standards. The company has been building data center interconnect systems like NVLink Fusion and partnering with companies like MediaTek to maintain a central role in the entire hardware chain, not just with its AI accelerators.
Beyond the data center
The collaboration extends across three areas. In AI infrastructure, MediaTek will work with the NVLink Fusion platform to help customers develop custom AI infrastructure designed for integration with Nvidia's rack-scale systems and AI factories. In local AI computing, the two companies will continue developing multi-generation Nvidia RTX Spark and DGX Spark PC chips that integrate Nvidia GPUs with MediaTek SoCs for consumer PCs, AI development supercomputers, and enterprise workstations. In automotive, MediaTek and Nvidia will continue building AI-driven software-defined vehicle platforms based on MediaTek's Dimensity Auto line.
MediaTek had previously partnered with Alphabet Inc.'s Google and collaborated with Nvidia on the GB10 Grace Blackwell Superchip that powers the DGX Spark developer computer. MediaTek vice chairman and CEO Rick Tsai said the investment "strengthens a collaboration that spans cloud AI infrastructure, local AI computing and automotive in the era of physical AI."
The $3.5 billion convertible bond purchase could become a significant equity stake in MediaTek if converted, giving Nvidia a direct financial interest in the success of its custom silicon partners. For Nvidia, the investment is the latest in a series of strategic bets across the AI supply chain, including power infrastructure developer Lancium and chip designer Marvell Technology. Concern about the circular nature of these investments has drawn scrutiny from investors, though Nvidia has disputed the idea that it is creating false demand.
This article is for informational purposes only and does not constitute investment advice.
