Nvidia is investing $3.5 billion into Taiwanese chipmaker MediaTek. As a part of the deal, MediaTek will undertake Nvidia’s know-how that can assist it design customized chips for AI corporations and hyperscalers that may be plugged instantly into Nvidia-based information facilities.
The deal comes as a rising variety of AI corporations and main cloud suppliers like Amazon, Google, Microsoft, OpenAI and Anthropic spend money on constructing their very own chips to allow them to be much less reliant on Nvidia’s GPUs. Offers like this one with MediaTek permit Nvidia to cede floor to customized silicon whereas nonetheless sustaining its lead because the dominant information heart scaffolding.
“Nvidia is an AI infrastructure firm,” Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure options, mentioned on Monday on a name with reporters. “We expanded past pure computing chips years in the past.”
The partnership underscores the round nature of Nvidia’s financing efforts over the previous few years, which have usually seen the chip large investing in corporations that move again into its personal ecosystem. The deal will give MediaTek entry to Nvidia’s NVLink Fusion ecosystem, together with NVLink, the know-how that lets any chips — even when they’re not Nvidia’s chips — talk with one another rapidly.
Final week, Nvidia introduced the same partnership with Amazon Net Providers, albeit with out the direct funding. AWS will deploy an extra 2 million Nvidia GPUs throughout its infrastructure, and likewise combine NVLink Fusion.
With the Nvidia partnership, MediaTek can proceed to construct customized chips designed by cloud corporations or AI labs tailor-made to their workloads, leveraging Nvidia’s “confirmed scale-up and scale-out know-how stack and ecosystem” and “rack-scale structure.”
“Principally, each cloud, each mannequin builder is deploying our platform in some form, kind, or style,” Harris mentioned. “So by MediaTek with the ability to supply this extension to its prospects, it permits them to standardize on the rack-scale infrastructure throughout their AI factories… and likewise deploy their [custom chips] proper alongside these utilizing the identical commonplace platform.”
“That is actually about opening up this ecosystem to the whole MediaTek buyer base,” Harris continued.
MediaTek hasn’t publicly shared details about its buyer base because it pertains to customized AI chips, nevertheless it’s clear Nvidia goals to assist it enhance that aspect of the enterprise as long as it could actually come alongside for the experience. MediaTek has been slowly constructing its customized information heart ASIC (application-specific built-in circuit) operations, saying in June that it expects the enterprise to generate $2 billion in income in 2026, and hopes to focus on extra of the market within the coming years.
Given MediaTek’s experience in creating customized chips that energy smartphones, sensible houses, vehicles, and wi-fi communications, the corporate may also proceed to collaborate with Nvidia on DGX Spark, the corporate’s small developer-focused desktop AI laptop. The businesses have prolonged their collaboration with the RTX Spark, Nvidia’s push to place its AI tech into client AI PCs.
MediaTek and Nvidia may also proceed “creating platforms for AI-powered, software-defined vehicles within the period of bodily AI,” per a press launch. MediaTek’s auto platforms depend on Nvidia’s RTX graphics for clever car cockpits and work alongside Nvidia Drive AGX, the corporate’s in-car computing platform for autonomous driving workloads.
“AI is remodeling each computing platform — from the world’s largest AI factories to the PC and the automobile,” Jensen Huang, founder and CEO of Nvidia, mentioned in a press release. “Collectively, we’re constructing platforms that convey NVIDIA accelerated computing to new markets and provides prospects the liberty to create differentiated AI techniques at monumental scale.”
Whenever you buy by way of hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.
