|Son G|

Talking out loud on memory (of which I'm bullish) for the sake of open discussion and iterative thought - I actually do agree with the "bears" in that memory is a somewhat a commodity so you have to treat it like one - why commodity ? well standards are set so the outputs are fungible for the most part - does commodity means something is not valuable? not int the slightest - what this does mean is you should have an opinion on the value of the commodity if you hold the commodity producer - and you should be aware of cost of production on value of the commodities: - I think @bubbleboi might have much more sophistication on this but I think the $/gb of NAND has more upside than HBM and DDR. on a relative basis I see HBM and DDR fully priced. HBF is one reason why theres upside optionality for nand but I would say HBF represents the vector of thinking. Which is how to optimize the cost structure of a system -- and I think the ecosystem is only starting to try to tune the optimization. which I don't fault them for because you don't start optimizing until you realize its the long term path. So how to utilize more % NAND and DDR whilst still maintaining performance. While DDR has utility upside I *think* DDR prices are extreme because of the reallocation to HBM so I dont think theres as much $ upside. - I said relative basis because I actually think all tiers of memory unlock another demand bump at some point when agents become truly 1/ easy to use 2/ somewhat autonomous - Kioxia / SNDK seem to have best in class cost of production of NAND Why do I still own HBM companies? Well it's partly out of admitting my own ignorance to be candid. The big 3 have capabilities across all 3 tiers and are allocating their time and capital on where they see the best ROI. Samsung has a logic foundry that makes it less clean so I'll just reference SK and Micron from here.. I think Sanjay is a good capital allocator and SK has had tech leadership for a while. So to some degree I am admitting to Sanjay hey you see things that I don't see and you're telling me it's worth the focus on HBM etc. I don't take this ignorance lightly. Some random thoughts: - in the semis space some historical references of valuation rewritings was TSMC and TXN in. the 2010s - and similarly big oil in 2020s - management has to prove capital allocation wisdom and restraint, its not just "LTAs". And I think we're seeing memory management take advantage of the moment to do so. If they start being reckless, even with LTAs then guess what people like me will sell and the multiple will drop. The street likes to act like reratings are simple but I disagree. It's complex and it comes down to trust. See Walmart and Costco trading at 40-50x. Costco will be an "expensive" stock until I die. $MU $SNDK $DRAM