WBT are establishing themselves as a thought leader, leaving no...

WBT are establishing themselves as a thought leader, leaving no stone unturned in exploring the potential applications of reram. Firstly on AI with the Indian Uni now Technion for this, they are making the most of their momentum.

I wont engage in the nearology of the players in the space and their relationships, been burnt too many time in the past because confirmation bias is a very powerful thing. This arrangement can stand on its own right without leaping to end manufacturing partnerships. Suffice to say all of Leti's friends are around and WBT is working its way solidly into that ecosystem of these players.

Being able to do computation at the secondary storage level in the memory hierarchy is a really big thing. If you dont need to move data to transform it, that is a huge thing. The really slow part is moving it around and managing the hierarchy effectively. I wouldnt expect overtly complex calculations or particular speed but if the only thing you are delivering to the CPU is value added data that augers a huge increase in efficiency. It also screws with the traditional CPU, L1 L2 L3 cache, RAM, NVM model. If one of the outer tiers that is NVM can change data it may mean it becomes sensible to use that insetad of the faster volatile tiers and push it directly to the CPU afterwards. This could change the economics of the memory tiers considerably.

If we think what database platforms like Exadata do in terms of database focused storage cells which do huge amounts of database specific processing offload away from the main processors, leading to huge efficiencies, I can possibly see this being the outcome for WBT reram for a more general case of data.