Top 5 AI Chips Featured at Computex

 

AI photo 

Top 5 AI Chips Featured at Computex

At Computex 2025, major tech companies presented their bold visions for the future of artificial intelligence, both in the cloud and at the edge.

But the focus wasn't limited to the usual companies like NVIDIA and Intel.

From MediaTek's AI ambitions to a neuromorphic chip developed by German researchers, this year's exhibition revealed a diverse and evolving AI landscape, according to a report published by Gizmochina and reviewed by Al Arabiya Business.

Computex 2025 demonstrates one clear fact: the AI ​​chip market is not only booming, but rapidly evolving.

Here's a closer look at the most prominent AI chips and architectures showcased at the show:

NVIDIA Chips:

NVIDIA introduced its Grace Blackwell NVL72 architecture at Computex 2025, setting a new standard for AI-based high-performance computing.

This hybrid architecture combines GPUs and CPUs in an integrated package and is capable of running models with trillions of transactions, real-time inference, and large-scale scientific workloads.

The company also launched NVLink Fusion, a next-generation interconnect technology designed to enable high-speed inter-chip communication.

NVIDIA confirmed that companies such as MediaTek, Qualcomm, and Marvell will adopt NVLink Fusion to build semi-custom AI computing solutions.

This move opens NVIDIA's ecosystem to external partners and expands the reach of its AI infrastructure technology.

MediaTek's 2nm Chip:

MediaTek announced plans to produce its first 2nm chip in collaboration with TSMC in September 2025.

According to the company, the new chip will offer up to 15% better performance and 25% higher power efficiency compared to current 3nm chips.

MediaTek aims to use this new chip in both mobile devices and high-performance AI applications.

The company also highlighted its collaboration with NVIDIA on the development of DGX Spark, a small AI supercomputer for developers.

MediaTek has provided a custom 20-core CPU architecture for the system and is leveraging its ASIC expertise to design chips specifically designed for training and inferencing AI models.

This initiative supports MediaTek's intellectual property in the areas of high-speed input/output, SerDes, and optical interfaces.