Weebit - 2023 and beyond, page-1516

I think it will establish itself as new product class once rram is manufactured for compute in memory - as we know the foundries are currently available to manufacture rram as embedded product and set up accordingly - in 3 to 5 yrs when wbt and partners have established the authenticity of the reproduceable manufacturing of rram across customer IC designs ( which is about to begin when the news drops) ...it will align with the development curve of the AI rram cim capability as I see it....wbts' work in parallel is a huge payday moat - to be working on rram memory training now in the lab is a huge cost mitigation as it is a generational development already being put in test and updated- and very exciting

sensor data is already fed to AI processors on an ever increasing level as robotics takes more focus in manufacturing and development - but they also rely on CPUs' atm so the AI processor is used as a CPU assist.....
this will alter once that inflection point of proven reproduceable manufacturability is shown of rram in multiple IC applications....wbt are right there and progressing in exactly that direction - again just my view of what we're seeing develop
the Skywater foundry is the perfect vehicle to realise this as is the 130nm process node tbh .....and to have a ready customer base to adopt the AI cim rram product in that 3 to 5 yr time frame is about 5 birds in the bush - not 2 ........

TI on sensor data / AI processor implementation in robotics.....happening on a very large scale -
https://www.ti.com/lit/wp/sszy036/sszy036.pdf?ts=1686375362152&ref_url=https%253A%252F%252Fwww.google.com%252F

the cost saving dimension of the AI cim rram process combined with the application capture at the edge will drive its own adoption

just my own thoughts - could be way off - but i don't think so....i think that's what we're seeing unfold in the AI rram big picture - albeit still a few years off