Samsung Joins Core Ecosystem of Nvidia's 'AI Factory'

Samsung Electronics has officially joined Nvidia's next-generation artificial intelligence (AI) infrastructure ecosystem, 'NV Link Fusion.' This move is expected to mark a turning point, expanding beyond Nvidia’s GPU-centric exclusive structure toward an open AI factory ecosystem centered on custom silicon.

On October 13 (local time), Nvidia announced via its blog that Samsung Foundry and Intel have joined the NV Link Fusion ecosystem. NV Link is a high-speed communication technology that allows GPUs to exchange data with each other without a CPU. Previously, it only connected Nvidia GPUs. The Fusion architecture expands this externally, enabling the integration and operation of various chips such as CPUs, GPUs, and XPUs within a single system.

Samsung, in collaboration with Nvidia, will provide full-stack foundry capabilities from design to manufacturing to meet the growing demand for custom CPUs and XPUs. This is considered the first case in which Samsung is participating directly in the core structural design of AI factories, going beyond simple contract manufacturing. Intel also plans to produce x86 CPUs that will be integrated into Nvidia's infrastructure platform.

NV Link Fusion is central to Nvidia’s strategy of transforming its previously closed, GPU-centric ecosystem into an open one. Jensen Huang, CEO of Nvidia, first unveiled this technology during his keynote speech at Computex in Taiwan last May, emphasizing, "AI is converging with every computing platform. NV Link Fusion opens up the ecosystem so partners can build specialized AI infrastructure."

Samsung’s participation not only disrupts the TSMC-dominated foundry landscape but also signals the beginning of an era of heterogeneous chip integration beyond a single GPU structure.

According to Nvidia, utilizing NV Link Fusion allows companies to freely configure combinations of hardware and Nvidia platforms, enabling AI factories to process high-load tasks such as model training and agent-based AI inference at high speed. Currently, Nvidia’s NV Link Fusion ecosystem includes custom silicon designers such as MediaTek, Marvell, Astera Labs, Synopsys, and Cadence, as well as CPU and intellectual property (IP) partners.