Nvidia's NVLink Fusion is an IP and chiplet solution that facilitates high-speed, direct communication between CPUs, GPUs, and accelerators, reducing performance bottlenecks in large-scale systems.

At the 2025 Open Compute Project Summit, Nvidia revealed next-generation hardware for large AI data centers, including the NVL144 server with 144 GPUs, liquid cooling, and GDDR7 memory.

Samsung's participation will also support the manufacturing of chips with cutting-edge processes, further boosting data center performance.

This collaboration includes the expansion of Nvidia's NVLink Fusion ecosystem, which now allows outside companies to connect their chips directly to Nvidia's GPU infrastructure, helping Nvidia retain market share amid rising competition.

This hardware rollout underscores Nvidia’s push to integrate comprehensive solutions for AI infrastructure.

Samsung Foundry has joined Nvidia’s NVLink Fusion ecosystem, enabling Samsung-manufactured chips to interface with Nvidia platforms, which could attract new clients amid TSMC’s capacity constraints.

Nvidia plans to launch a GPU-based AI supercomputer called Kyber by 2027, which will incorporate 576 GPUs operating as a single system.

Samsung Foundry's full design-to-manufacturing capabilities strengthen Nvidia’s control over the AI hardware stack and expand its presence in semiconductor manufacturing.

Nvidia has partnered with Samsung Electronics to develop custom CPUs and XPUs, aiming to strengthen its position in the competitive AI hardware market.

Samsung's involvement will enable the production of advanced chips, potentially using 2nm processes, to improve data center efficiency through faster interconnects and lower latency.

Major tech companies like Meta and Oracle are adopting Nvidia's Spectrum-X interconnect technology to enhance communication within large AI data centers.

Nvidia CEO Jensen Huang highlighted that AI models now operate at trillions of parameters, transforming data centers into gigascale AI factories, with Spectrum-X acting as the 'nervous system' connecting millions of GPUs.

This strategic partnership with Samsung is seen as a move to reshape supply chains, leveraging Samsung’s scale to produce chips more cost-effectively despite legal and IP challenges.