Arteris Discusses Data Movement Challenges in Physical AI
Understand the story
On October 21, 2026, Semiconductor Engineering reported on the challenges physical AI systems face in data movement architecture. The report noted that physical AI systems mix multiple types of compute units, need to process data and make decisions in real time while managing battery life, and that the key lies in deciding where data is processed and how to minimize data movement. Ashley Stevens, Arteris's Director of Product Management and Marketing, explained in the report that coherent interconnects are better suited for CPUs while non-coherent interconnects better suit AI accelerators, and outlined other challenges engineers will face going forward. Currently there is only one report on this topic, with no further developments.
AI overview of reporting · Updated 2 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- Semiconductor Engineering Chips and PackagingScaling Data Movement For Physical AI
Physical AI systems mix multiple compute units and need to process data, make decisions in real time, and manage battery life; the key is deciding where data is processed and how to minimize data movement. Ashley Stevens, Arteris's senior director of product management and marketing, explains that coherent interconnects suit CPUs better while non-coherent interconnects suit AI accelerators better, and looks ahead at other challenges engineers will face.
Story heat trend
Not enough continuous observations to show a trend yet.