Chris Bergey, executive vice president of Arm's Edge AI Division, recently said that the release of NVIDIA RTX Spark based on Arm architecture marks that the PC industry has officially entered a new stage of intelligent AI."For decades, PCs have been designed around how users use applications," Bergey said.Agent AI (Agentic AI) is changing this landscape: applications are becoming more autonomous, and users only need to issue instructions, and agents can complete observation, reasoning, planning and execution operations in the workflow."
Bergey pointed out that intelligent AI has given rise to two broad types of computing power requirements: one is excellent energy efficiency to support all-day battery life and efficient productivity; the other is extreme performance to meet the needs of advanced AI workloads, content creation, software development and local reasoning.
Bergey believes that the new NVIDIA RTX Spark based on Arm architecture is the key product to respond to these needs.
It is understood that the RTX Spark released at COMPUTEX 2026 integrates the NVIDIA Grace CPU and NVIDIA Blackwell RTX GPU based on the Arm architecture, and adopts a completely unified memory architecture.
Kaustubh Sanghani, vice president of product management at NVIDIA, said that to achieve a responsive end-side AI experience, the agent needs to deeply integrate and optimize the GPU, CPU and memory architecture, and RTX Spark is the practice of this concept.
Bergey further emphasized that as the number of tokens used by AI models continues to increase, the key to reducing single-task costs is to adopt large-scale end-side models to utilize tokens more efficiently. This not only brings lower, predictable costs, but also provides better data privacy protection for users and enterprises.
