Apple are partly to blame for the chip shortage. Getting deals on previous chips priced to the bone meant that the chip firms didn't have money to invest in new plants. Memory use has expanded massively. it's only recently Apple were selling an 8gb Mac now 96gb and 128gb is quite common for PC's. These factors together.
Apple should build/buy a memory FAB, they are too exposed. Apple have left themselves vulnerable.
Something the new guy has to be all over. It already looks like Macs have been held back due to lack of ram.....
I would have thought Tim would have seen this coming?
Even if Apple and other DRAM purchasers hadn't been so cutthroat with the DRAM manufacturers in 2023, and this had given the "Big Three" manufacurers enough money to build two or three extra mega-fabs, those factories wouldn't be producing consumer RAM today anyway. Because HBM (High-Bandwidth Memory) sells at a 70%+ profit margin compared to the very thin margins of consumer PC RAM, memory executives would have immediately stuffed those hypothetical new factories with specialized HBM-manufacturing equipment that would use even more silicon waferage for HBM. They would still be selling 100% of that capacity to Nvidia, Microsoft, and Google. The consumer market would still be starving for standard DRAM because currently a wafer used to make DRAM for a laptop or desktop computer is a wafer wasted in the eyes of a chipmaker's shareholders.
The average consumer buying a 16GB Mac or a 32GB PC, or one with the higher RAM configurations you cite, provides RAM manufacturers with a profit level that's now a drop in the bucket compared to what they make selling HBM RAM to companies making AI data centers. A single AI server cluster requires terabytes of HBM, and because HBM requires stacking chips vertically, it physically consumes 3 to 4 times the raw silicon wafer space of standard consumer RAM. The current DRAM shortage isn't because people have been buying computers recently with more RAM, or even because RAM manufacturers didn't build more manufacturing capacity for them to keep up with that demand. It's because AI data centers are devouring 20% to 25% of the entire world's raw silicon wafer output.
Building a cutting-edge memory fabrication plant takes 2 to 3 years and costs upwards of $20 billion per facility, and you need a massive, global network of them to achieve economies of scale. And RAM manufacturing isn't Apple's part of Apple's business anyway. They don't even make their own A-series or M-series processors, and instead they outsource that to TSMC.
If Apple sunk $100 billion into building a global memory foundry network just to supply its own devices, it would significantly deflate its corporate margins, and we know shareholders don't like that.
Tim Cook did see this coming. His solution hasn't been to build a RAM factory, but to use Apple's geopolitical leverage to qualify alternative Chinese memory suppliers like CXMT to try to break the Western memory cartel's pricing chokehold.