Custom CPUs & XPUs Made By Samsung Foundry Could Now Leverage NVIDIA’s NVLink Ecosystem
At the recent OCP Global Summit, one of the most notable announcements came from NVIDIA, revealing that Samsung Foundry has officially joined its expanding NVLink Fusion ecosystem, a development that underscores both NVIDIA’s strategic vision for scalable AI infrastructure and Samsung’s growing role in next-generation chip manufacturing.
In a blog post on OCP, NVIDIA confirmed that Samsung Foundry, alongside Intel, is joining the NVLink Fusion ecosystem, which connects custom silicon designers, CPU and IP partners to help AI factories scale efficiently for intensive workloads such as model training and agentic AI inference. This marks a milestone for Samsung, signaling its deeper integration into NVIDIA’s AI strategy and positioning it as a key player in the emerging AI hardware race.
According to NVIDIA, Samsung Foundry will support the growing demand for custom CPUs and custom XPUs, offering its full design-to-manufacturing expertise. The collaboration is not limited to production alone, as it encompasses end-to-end support, including silicon design, verification, integration, and tape-out. This partnership empowers Samsung’s customers to develop custom processors and accelerators that can seamlessly leverage NVIDIA’s NVLink interconnect technology, without necessarily being tied to NVIDIA’s own product lineup.
Samsung’s inclusion in the NVLink ecosystem also reflects the tightening partnership between NVIDIA and Samsung, which has strengthened notably following the recent meeting between NVIDIA CEO Jensen Huang and Samsung Chairman Lee Jae-yong. The outcome of that meeting included Samsung gaining approval to supply HBM3E memory for NVIDIA’s future AI accelerators, a critical component in enabling high-bandwidth performance for next-generation AI computing.
With the NVLink Fusion ecosystem now open to custom chip developers, firms such as OpenAI and other AI-focused enterprises could potentially turn to Samsung Foundry to produce specialized silicon optimized for NVIDIA’s networking and interconnect technologies. This collaboration also indirectly hints at possibilities for Intel, though the report suggests it is unlikely that Intel would rely on Samsung’s foundry services for its own x86 server CPUs.
The broader implication of this move lies in Samsung Foundry’s readiness to serve Big Tech’s custom silicon needs, backed by its advanced process technologies such as SF2 (2nm), which have already drawn significant industry attention. As AI demands surge and semiconductor designs grow increasingly complex, Samsung’s integration with NVIDIA’s NVLink ecosystem could establish it as a pivotal supplier for the next wave of AI hardware innovation.
What are your thoughts on Samsung’s growing role in NVIDIA’s AI ecosystem? Could this partnership reshape the balance of power in global semiconductor manufacturing?
