NVIDIA dominates the artificial intelligence (AI) data center computing industry with estimates that they will hold over 80% of the market by the end of 2024. Their AI CPUs and AI GPUs work using the NVLink connection to enable large amounts of data to be sent quickly.
However, NVLink only supports NVIDIA chips. Yesterday in conjunction with Computex 2025, Jensen Huang announced NVLink Fusion which will finally allow NVIDIA AI data center GPUs to be paired with CPUs from other companies.
MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys and Cadence have stated their commitment to producing AI chips that support the NVIDIA NVLink Ecosystem. Meanwhile, Fujitsu and Qualcomm have plans to produce data center chips that support NVIDIA GPUs.
The NVLink Fusion system can be seen as a response to UALink, which was established by Intel, Google, Microsoft, Meta AMD, Hewlett Packard Enterprise, Broadcom and Cisco as an open source data center format to compete with NVLink.