Story
September 1, 2026
Nvidia’s $3.5 Billion MediaTek Bet Keeps Custom AI Chips in Its Orbit
Nvidia’s investment gives MediaTek a bigger route into bespoke data-center chips while tying those designs to Nvidia’s infrastructure. Investors cheered the deal, even as it underscores the cloud industry’s push to reduce dependence on GPUs.
Nvidia is embracing the custom-chip shift rather than resisting it: MediaTek gains a deeper foothold in data centers, while Nvidia aims to ensure the alternatives cloud giants build still run through its infrastructure.
The arrangement was announced Monday, with Nvidia investing $3.5 billion in MediaTek convertible bonds and opening its NVLink Fusion platform to the Taiwanese chipmaker’s custom AI-chip customers.1 The technology is designed to let non-Nvidia chips communicate rapidly within Nvidia-based data centers—an answer to hyperscalers and AI labs developing silicon tailored to their own workloads.
For MediaTek, best known as the world’s largest smartphone-chip supplier by market share and Qualcomm’s principal rival, the partnership accelerates a diversification campaign into data-center ASICs. The company has said its custom AI-chip business could generate $2 billion in revenue, targeting a market it estimates could reach $80 billion by 2027.2
Investors treated the deal as validation: MediaTek shares closed 10% higher Tuesday, adding momentum to a rally of nearly 200% this year.2 The company is positioning itself against established custom-silicon provider Broadcom as cloud customers seek more options beyond Nvidia GPUs.
Nvidia’s message is that custom chips need not mean an exit from its ecosystem. “Nvidia is an AI infrastructure company,” said Dion Harris, the company’s senior director for HPC and AI hyperscaler infrastructure solutions. “We expanded beyond pure computing chips years ago.”1 NVLink Fusion, he argued, allows customers to place MediaTek-built chips beside Nvidia systems on a common rack-scale platform.
The deal also extends a wider relationship spanning AI PCs, developer desktops and automotive computing. Chief executive Jensen Huang framed the strategy as a push to carry Nvidia’s accelerated computing into new markets while allowing customers to build “differentiated AI systems at enormous scale.”1 In effect, Nvidia is conceding the chip design while seeking to retain the connective tissue.