IBM develops and sells IP. Not to mention having the tech and getting yield and working at scale are completely different thing. I doubt it make much of a different.

VTFET is sub 2nm, on TSMC time table, assuming they use it, you get Nanowire ( GAA ) with 2nm in 2025, it normally get another generation of fine tuning and cost amortisation. So you are looking at 2029 ( 1nm ) for volume production at the earliest.

TSMC is working on similar things.

At one time this research was shared amongst members of Sematech (which is the origin of the current NanoTech center in Albany where this work was done). But about 10 years ago, Sematech broke up as companies decided it was cheaper and more secure IP-wise to "go their own way" (Multinationals Going Their Own Way - or MGTOW ;-) so TSMC started doing their own R&D. This is a large part of why they are currently ahead of everyone else.

https://www.sunypoly.edu/research/albany-nanotech-complex.ht...

The NanoTech Center currently is primarily a boutique R&D center for Global Foundries + IBM with other specific companies invited in as needed.

But there is a "convergent evolution" to this kind of work because the laws of physics are not arbitrary nor are random directions of successful R&D possible - it's a constrained problem so even in independence, solutions will be similar because the physics problems that arise are the same and thus the solutions are going to "rhyme". TSMC is already doing something similar but the process design (the only part patentable) is different.

IBM are slightly weird in that they do ground breaking research and pump out some genuinely impressive products but then they don't seem particularly bothered selling these products to anyone with anything less than a board of directors.

I worked at ibm research last century. They spent 6 billion on research. They had a chip fab in the building and still were making chips at the time. They had a wide range of researchers (mathematicians, physicists , software). working on all sorts of stuff (very motivated). Lots of cooling gasses.

The problem was trying to turn those inventions into something they could sell. Chip tech was sold to whoever they could.

They don't need to make money selling chips if they can license their IP to people who run the foundries. IBM's prodigious patent portfolio is a huge revenue source for the company.

I wonder what is percentage? Is it still a mainframe shop? Last time I heard abd it failed is to become a consultancy firm with Pwc. What is ibm now?

IBM just spun out a $19B annual revenue managed infrastructure business, Kyndryl. [1]

Doing some back of the napkin math on the remaining business, by revenue it is roughly 45% consulting, 40% software, and 15% hardware. IP revenue represents 1% ($626M) of its revenue. [2]

What does it actually do? At a high level, help companies deal with the complexities created in a world of heterogeneous infrastructure (data and AI across environments, common management/security, automation, etc.).

[1] https://techcrunch.com/2021/11/04/kyndryl-officially-launche...

[2] https://www.ibm.com/investor/att/pdf/IBM_Annual_Report_2020....

Disclaimer: I'm an IBM employee, but my opinions are my own and I'm just using public data.

I don't know the numbers, but I do know they still do significant work in mainframes. They're even still designing their own Power processors for them.

Just a point of clarification, POWER and z (mainframe) are different architectures, both of which IBM still develops.

[1] https://en.wikipedia.org/wiki/Power_ISA

[2] https://en.wikipedia.org/wiki/Z/Architecture

Ah, ok. I didn't know if the IBM relationship with Samsung was exclusive. Sounds like no.