During Computex 2025, in Taipei, Nvidia presented one of its most ambitious initiatives in the artificial intelligence sector: NVLink Fusion. It is a semi-professionalized infrastructure that allows CPUs and third-party AI accelerators to communicate directly with gpus nvidia through a very high performance interconnection.
Hitherto exclusive to the company’s own products, NVLINK technology will now be licensed to strategic partners – which includes giants like Qualcomm, Fujitsu, MediaTek, Marvell, Alchip, Synopsys e Cadence.
The opening represents a paradigm shift and expands the ecosystem to hybrid architectures in AI -optimized racks.
What is Nvlink Fusion?
O Nvlink Fusion combines Custom Asics, low latency interconnectivity and a horizontal and vertical scalability model. It allows systems architects to create complete structures that operate with GPUS Nvidia and with other high performance chips, such as personalized CPUS and AI accelerators.
By expanding NVLINK beyond the traditional server environment, NVIDIA enables clusters em rack With hundreds of units cooperating simultaneously with bandwidth far superior to the PCI standard.
In practical numbers, the company states that NVLINK offers until 14 times more band than traditional interfaces, maintaining electrical compatibility with PCIE.
Partners and Practical Applications
Among the confirmed partners are:
- Qualcommwhich is developing its own CPU for servers;
- Fujitsuwith the processor Monaca of 144 nuclei and 3D stacking;
- MediaTek and Marvellcreating optimized asics for AI;
- Astera Labswhich will contribute to the design of interconnection silicon;
- Synopsys e Cadenceproviding the design and IP tools.
All these players start to integrate the ecossistema NVLink Fusioncan combine its products with Nvidia architectures in highly scalable datacers.
Fujitsu and Qualcomm will integrate CPUs with the Nvidia ecosystem via NVLINK FUSION, with up to 14x more bandwidth than PCIE in scalable architectures
Compatibility, performance and technological sovereignty
NVLINK FUSION also fits the proposal of technological sovereignty from various countries and companies.
The possibility of using CPUS NOT NAVIDIA WITH DIRECT CONNECTION TO GPUS Through standardized chipples and interfaces, it makes room for more modular, regional and customized solutions without depending exclusively on a single supplier.
Connecting our technologies directly to Nvidia’s architecture is a decisive milestone for our view of efficient and expanding Sovereign AI systems
Vivek Mahajan, CTO da Fujitsu
With this approach, NVLINK FUSION stands as Direct response to the Ualink initiativeled by companies such as AMD, Intel and Broadcom, who advocate an open interconnection standard for data centers.
New Software Layer: Mission Control
To orchestrate this new infrastructure, Nvidia also announced the Mission Control (Mission Control), a rack management software platform, cargo validation and distribution automation.
The tool aims to accelerate the delivery of ready -made production systems and facilitate the use of large -scale computational resources.
Also read:
By opening NVLINK to external partners, Nvidia expands its mastery over the AI ​​ecosystem and creates bridges with companies that previously operated independently.
NVLINK FUSION’s proposal shows how the company is willing to lead the future of computing in IA through collaboration and strategic interoperability.
Source: NVIDIA
Join the Adrenaline offers group
Check out the main offers of hardware, components and other electronics we find over the internet. Video card, motherboard, RAM and everything you need to set up your PC. By participating in our group, you receive daily promotions and have early access to discount coupons.
Enter the group and enjoy the promotions
Source: https://www.adrenaline.com.br/nvidia/nvidia-anuncia-nvlink-fusion-cpu-gpu-ia/
