Excellent deep dive on memory! The section on Western Digital's NAND partnership with Kioxia is particularly interesting - that 50/50 JV structure has been both a strength and a weakness for them. On one hand, it allows WDC to share capex burden and maintain scale, but on the other hand, they're dependent on Kioxia's fab operations and technology roadmap. The point about them being the weakest players without a profitable DRAM business to fund them through cycles really resonates - that's exactly why the proposed spin-off of SanDisk makes sense strategically. I'm curious whether the endgame is a merger between SanDisk and Kioxia, or if one of the bigger players (Samsung/SK Hynix/Micron) eventually consolidates the NAND market further. The transition from floating gate to charge trap and the move toward 1000-layer 3D NAND is also fascinating - it shows how the industry keeps finding ways to scale without hitting the same litho bottlenecks as DRAM. Great anlysis overall!
First, thanks for another great post, and hope your family is doing well!
As a sidenote for memory that spans/spanned working memory and storage, 3D XPoint (Optane) was IMHO a great example for Intel walking away from one of their interesting developments just before it would have likely taken off. The ability to use 3D XPoint - non-volatile memory - just like DRAM (also as sticks in DRAM slots!), lower latency than regular NAND, and its great endurance (read/write cycles) should be really attractive now in the age of AI. Maybe someone will revive that tech?
Excellent deep dive on memory! The section on Western Digital's NAND partnership with Kioxia is particularly interesting - that 50/50 JV structure has been both a strength and a weakness for them. On one hand, it allows WDC to share capex burden and maintain scale, but on the other hand, they're dependent on Kioxia's fab operations and technology roadmap. The point about them being the weakest players without a profitable DRAM business to fund them through cycles really resonates - that's exactly why the proposed spin-off of SanDisk makes sense strategically. I'm curious whether the endgame is a merger between SanDisk and Kioxia, or if one of the bigger players (Samsung/SK Hynix/Micron) eventually consolidates the NAND market further. The transition from floating gate to charge trap and the move toward 1000-layer 3D NAND is also fascinating - it shows how the industry keeps finding ways to scale without hitting the same litho bottlenecks as DRAM. Great anlysis overall!
Thank you for such great article
First, thanks for another great post, and hope your family is doing well!
As a sidenote for memory that spans/spanned working memory and storage, 3D XPoint (Optane) was IMHO a great example for Intel walking away from one of their interesting developments just before it would have likely taken off. The ability to use 3D XPoint - non-volatile memory - just like DRAM (also as sticks in DRAM slots!), lower latency than regular NAND, and its great endurance (read/write cycles) should be really attractive now in the age of AI. Maybe someone will revive that tech?