Nvidia's Rubin expected to use mostly 12-layer HBM4 despite 8-layer shift reports

Nvidia's next-generation Rubin AI accelerators are expected to rely predominantly on 12-layer HBM4 memory, despite reports that the chipmaker asked Samsung and SK hynix to boost supplies of 8-layer products.

• Rubin will predominantly use 12-layer HBM4
• Nvidia asked Samsung and SK hynix for more 8-layer HBM4
• Industry tests HBM stacks beyond 16 layers