Nvidia announces a new technology to keep it at the AI ​​Development Center

Nvidia CEO Jensen Huang announced a lot of ads on Monday and revealed new products that aim to keep the company at the Center for Development and Artificial Intelligence Computers.

One of the most noteworthy ads was the new “Nvlink Fusion” program, which will allow customers and partners to use the central NVIDIA processing units and graphics processing units along with NVIDIA products and its NVLink.

Until now, Nvlink was closed on the tokens made by Nvidia. Nvlink is a technology developed by NVIDIA to connect and exchange data between the GPU and CPU processor.

“NV Link Fusion is so that you can build AI semi-non-standard infrastructure, not only for semi-use systems,” Huang said at Computex 2025 in Taiwan, the largest electronic conference in Asia.

According to Huang, the NVlink fusion allows AI infrastructure to combine NVIDIA processors with various processors and integrated systems specific to the application (ASICS). “In any case, you use the use of NV Link infrastructure and NV connection ecosystem.”

Nvidia Monday was announced that the partners with AI chips for nvlink fusion already include MediatekIN MarvellIN AlchipIN Asera LabsIN Abstract AND Rhythm. As part of Nvlink Fusion, Nvidia customers like Fujitsu AND Qualcomm Technologies He added that they will also be able to connect their own external processors from NVIDIA GPU in AI data centers.

Ray Wang, an analyst of semiconductors and technologies from Washington, told CNBC that Nvlink represents NVIDIA plans to capture the participation of ASIC data centers, which were traditionally perceived as NVIDIA competitors.

While NVIDIA has a dominant position in the GPU used for general AI training, many competitors see Room for expansion In systems designed for more specific applications. Some of the biggest competitors of NVIDIA in AI computers – who are also some of its largest clients – include cloud suppliers such as Google, Microsoft and Amazon, all of which are building their own custom processors.

Nvlink Fusion “consolidates NVIDIA as a new generation AI center center-nawet when these systems are not fully built with NVIDIA systems,” Wang said, noting that NVIDIA opens the possibilities of servicing customers who do not fully build NVIDIA systems, but they are looking for some of the GPU.

“If it is widely accepted, Nvlink Fusion may broaden the NVIDIA industry trace, supporting deeper cooperation with non -standard processor programmers and ASIC designers in building AI infrastructure in the future,” said Wang.

However, Nvlink Fusion risks to reduce the demand for the NVIDIA processor, enabling NVIDIA clients to use alternatives, according to Rolf Bulk, analytics of capital research at New Street Research.

Nevertheless, “at the system level, additional flexibility improves the competitiveness of NVIDIA GPU solutions compared to alternative emerging architecture, helping Nvidia to maintain positions in the AI ​​computer center,” he said.

Nvidia Broadcom, AMD and Intel competitors are so far absent in the Fusion Nvlink ecosystem.

“DGX Cloud Lentton helps to solve the key challenge related to securing reliable, high-performance GPU resources by unifying access to AI services in the cloud and GPU capacity in the NVIDIA calculation ecosystem,” said Wa Press message.

In his speech, Huang also announced the plans of the new office in Taiwan, in which he will also build a project Supercomputer AI from Taiwan Foxconn, officially known as the HON Hai Technology Group, the world's largest electronics producer.

“We are glad that we can cooperate with Foxconn and Taiwan to help build AI Taiwan infrastructure and support TSMC and other leading companies in the development of innovation in AI and robotics,” said Huang.