NVIDIA, 2025 Open Compute Project Global Summit in the scope of Samsung Foundry started a new collaboration with. Company, NVLink Fusion by expanding the ecosystem based on the chiplet structure called special CPU And XPU is moving to a stronger position in production. With this development, NVIDIA is not only a GPU manufacturer but also assumes a decisive role in the field of chip design and production in artificial intelligence-supported data centers.
Recently, processing power and communication speed between components have gained great importance in data centers. NVIDIA’s NVLink Fusion architecture offers a solution to this need by providing direct and fast data flow between processors. Thanks to this structure, CPU, GPU and accelerator components can be efficiently integrated in a single rack system. This technology is especially in high-performance artificial intelligence workloads It aims to reduce latency and increase system efficiency. In addition, NVLink Fusion, the company’s existing MGX And OCP are being developed in line with their infrastructure. This compatibility facilitates scalable system installations in data centers.
Samsung Foundry’s participation in this structure was an important turning point for NVIDIA. Now the company will be able to be active not only on the design side but also on the production line. Samsung will provide full support from design to production in the production of custom processors and accelerators. This situation is reflected in NVIDIA’s production processes. without depending on a single source allows it to move. On the other hand, Production diversity in different geographies Its creation means strategic assurance for the global supply chain. In addition, Samsung’s advanced 3-nanometer manufacturing process will enable high-efficiency custom chip production.
Samsung Foundry expands NVIDIA’s influence in custom chip space
Previously included in the NVLink Fusion ecosystem Intel And Fujitsu Manufacturers such as were involved. Now, with the participation of Samsung, the scope of the system has expanded. The company will work with NVIDIA to connect custom CPU and accelerator chips directly via NVLink. But this structure is not a completely open system. NVIDIArequires chips developed under NVLink Fusion to work with its own products. Communications controllers, physical link layers, and NVLink Switch licenses remain under NVIDIA’s control. This situation reduces the space of action of third-party manufacturers, while allowing the company to maintain its dominance over the ecosystem.
Despite this, new producer initiatives are also increasing on a global scale. OpenAI, Google And Amazon Web Services Organizations such as have turned to developing their own hardware solutions. This trend is intensifying competition in the AI hardware space dominated by NVIDIA. NVIDIA, with its NVLink Fusion, provides a solution that extends from hardware to software against this competition. an integrated structure It creates. Thus, it can directly integrate its own IP infrastructure into both the production chain and data center architecture. In the long run, this approach will make NVIDIA not just a GPU provider, full-scale AI hardware provider moves it to position.
The main purpose of NVLink Fusion is to eliminate the performance bottleneck between components. Thanks to its high-speed communication, this system reduces inter-processor delay and increases energy efficiency. In addition, this structure, built on NVLink architecture, can be used in data centers. not based on a single supplierhowever providing integrated performance enables the development of solutions. This new collaboration between NVIDIA and Samsung Foundry is preparing to transform the way artificial intelligence-focused hardware is produced in the future. While the company expands its ecosystem with new partnerships in the production chain, it also develops artificial intelligence equipment. A structure that covers all layers continues to establish.