Tech news in 3 minutes

Nvidia’s $3.5B MediaTek bet reveals its plan for tackling Big Tech’s AI chip buildout

4 d ago

Nvidia is investing $3.5 billion into Taiwanese chipmaker MediaTek, enabling MediaTek to adopt Nvidia’s NVLink Fusion ecosystem for designing custom AI chips that plug directly into Nvidia-based data centers. This SEO-focused summary targets readers searching for semiconductor, AI chip, data center, and technology business news, highlighting Nvidia’s strategy to maintain dominance amid growing competition from cloud providers building their own silicon. The deal allows MediaTek to leverage Nvidia’s rack-scale architecture and NVLink interconnect technology, letting custom chips—even non-Nvidia ones—communicate rapidly. This helps Nvidia cede ground to custom silicon while keeping its data center scaffolding lead. “Nvidia is an AI infrastructure company,” said Dion Harris, Nvidia’s senior director of HPC and AI hyperscaler infrastructure solutions. “We expanded beyond pure computing chips years ago.” The partnership follows Nvidia’s circular investment pattern, funding companies that feed back into its ecosystem. Last week, Nvidia announced a similar collaboration with Amazon Web Services (without direct investment), deploying an additional 2 million Nvidia GPUs and integrating NVLink Fusion. MediaTek, which expects its custom data center ASIC business to generate $2 billion in revenue by 2026, will also continue working with Nvidia on DGX Spark (desktop AI computer), RTX Spark (consumer AI PCs), and automotive platforms for software-defined vehicles and autonomous driving. “AI is transforming every computing platform,” said Jensen Huang, Nvidia CEO. “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 investment underscores Nvidia’s push to standardize its infrastructure across AI factories, allowing cloud companies and AI labs to deploy custom chips alongside Nvidia hardware.