Samsung Electronics is newly joining Nvidia's "custom artificial intelligence (AI) infrastructure" ecosystem. With this, Samsung Electronics is expected to emerge as a key partner in Nvidia's "AI factory" (AI data center) strategy.
According to the industry on the 14th, Nvidia said on the 13th (local time) via its official blog that Samsung Foundry and Intel have joined the "NVLink Fusion" ecosystem.
NVLink is a high-speed consolidation technology that enables direct communication between GPUs (graphics processing units) and has previously been applied only to Nvidia chips. The newly expanded NVLink Fusion is an architecture that broadens this technology to external central processing units (CPUs), GPUs, and custom AI chips, enabling integrated data exchange between different semiconductors.
Samsung Foundry will work with Nvidia to meet growing demand for custom CPUs and XPUs (accelerated processing units) and provide one-stop capabilities from chip design (IP) to manufacturing. Intel will produce x86-based CPUs that integrate into Nvidia's AI infrastructure platform.
The addition of Samsung and Intel is seen as part of Nvidia's strategy to build an open collaborative ecosystem with a variety of semiconductor corporations rather than designing AI factories exclusively. Nvidia CEO Jensen Huang said in a keynote at "Computex 2025" in May that "AI is permeating every computing platform," and emphasized that "NVLink Fusion is a technology that opens Nvidia's ecosystem so partners can build their own specialized AI infrastructure."
Nvidia said that by leveraging NVLink Fusion, corporations can freely configure hardware combinations and the Nvidia platform, allowing AI factories to perform high-performance computing tasks such as model training and agentic AI inference faster.
Currently, major semiconductor design and IP corporations including MediaTek, Marvell, Astera Labs, Synopsys, and Cadence are participating in the NVLink Fusion ecosystem, and the ecosystem's scope has expanded further with the addition of Samsung Electronics and Intel.