Weebit - 2023 and beyond, page-940

"We not on their radar?"

I think perhaps they already realise that WBT is not likely to ever be a direct competitor. They are talking about multi layering two types of III-V compound semiconductors and building this on a Si substrate. Very difficult. The diametric opposite of the simplicity of WBT's technology.

It will be very expensive for sure, which means they would target the highest value applications, at least initially. They might start with CPU cache (perhaps chiplets, given how difficult it is going to be to integrate their technology into Si wafers) then followed by DRAM. That would keep them well occupied for at least the first 5 to 10 years of their commercialisation. Only then would they even consider challenging WBT on their home turf of embeded NovRAM, memory sticks, and probably by that stage, SSDs.

Here is the link to the paper for those who are interested:
https://onlinelibrary.wiley.com/doi/10.1002/aelm.202101103

They are currently working at 10 mm and 20 mm and extrapolating performance to smaller geometries. This from the paper:
"Assuming ideal capacitive scaling[33] down to state-of-the-art feature sizes, the switching performance would be faster than DRAM, although testing on smaller feature size devices is required to confirm this."

A lot of ifs there. They have to scale down by 3 orders of 10 to verify their claims. If you look at their cross-section:

their "triple-barrier resonant-tunneling (TBRT) structure" looks to be about 15nm thick. I think the operative word here is 'resonant'. this implies (to me) that the thickness is important to the function and cannot be scaled. This would then logically limit how far down they can scale this device. My intuition is that they will be facing a similar limiting scale to flash i.e. 40nm, perhaps 22nm if they are really lucky.

Given the advantages of this tech (assuming it can be made to work), the cost and limits to scaling should still not be barrier's to adoption in some applications, such as cache. However, I'm confident that ReRAM will be safe from challenge in it's market niches.

Another couple of things: This tech won't be naturally radiation resistant like WBT's ReRAM, nor will it be any good for building synaptic neural networks.

I'm not losing any sleep.