Ann: Megabit Success, page-1081

The main goal is reduction in energy use.

Data centres store a huge amount of "static" information (that is, information that doesn't change due to calculations or modification) in system memory (DRAM) where it can be accessed quickly. Think, all of Amazon, Google Maps, Facebook - heavily visited websites. In your mobile or PC, that sort of data is stored on FLASH memory which is energy efficient by comparison.

DRAM burns energy constantly to maintain that stored information. Energy that mostly dissipates as heat. Which all has to be pumped out of the data centre building using massive climate control systems. Which all consume energy.

So if you can use a memory that cuts your energy use by a significant figure - you can design the size of your energy connection to be smaller. You can make your climate management systems smaller.

That 's one use case.

Another: Self driving car systems - the self driving car builds a 3D model of the area around it from pre-existing 3D data (the base model) and then superimposes objects it detects from its own sensors - the amount of memory required to maintain that 3D model is monstrous, and consumes energy - electric cars are big on energy reduction - they also want that information in fast memory storage to minimise latency in self driving decisions. Nanoseconds that all add up - the less latency, the better the "reaction time" of the car.