Memory in physical AI: LPDDR6 and the data hierarchy in robots, vehicles and drones

The content retrieved from this Semiconductor Engineering link is only a very brief teaser of the article. The full piece, which appears to be a video or interview, was not available, so this summary is limited to what appears in the listing and does not cover how the conversation unfolds.

The subtitle argues that LPDDR6 is becoming an essential ingredient as intelligence moves into autonomous systems and mobile edge systems. The article is dated November 9, 2026, and is by Ed Sperling.

The starting point is that humanoid robots, autonomous vehicles and drones run on batteries. When the battery runs out, they have to return to a charger. For humanoids, useful runtime may currently be limited to a few hours per charge. The listing notes there is no simple answer for extending that time, because no single component solves every problem. The fundamentals remain threefold: efficient processing, denser and faster memory, and optimized data movement between processors and memory.

According to the listing, Frank Ferro, group director of product management at Cadence, explains why data hierarchies matter so much for physical AI. He also covers what can be done on the memory side to speed up read and write times, and how to maximize performance without driving up costs. The article's tags are cache, Cadence, HBM, LPDDR6, memory, memory hierarchies and SOCAMM.

No further details are available, such as specific power consumption figures, comparisons between memory types or performance data. For those aspects, the original article should be consulted.