More than half of Samsung’s 4nm output is going to HBM4 base dies
Samsung Electronics has allocated more than half of its 4nm foundry capacity to HBM4 base dies. According to DigiTimes, the production lines are currently running at full load, and the company is accelerating shipments to major AI chip customers.
The report said 50% to 60% of Samsung’s 4nm output is being used for HBM4 base dies. Production is based at the second and third campuses in Pyeongtaek, using the S5 and S6 foundry lines. Combined 4nm capacity at those sites is about 30,000 wafers per month, with roughly 15,000 wafers assigned to HBM4 base dies.
Why HBM4 base dies consume foundry capacity
The base die sits at the bottom of the HBM stack. It communicates with the main chip and controls the full memory package. In sixth-generation HBM4, that layer has moved from a memory manufacturing process to a logic foundry process, which means it now takes up foundry capacity directly. The report said this is why Samsung’s foundry and memory divisions can end up competing internally for the same resources.
Less capacity left for other 4nm customers
When more than half of a production line is occupied by a single product family, the amount of capacity available to other 4nm customers becomes tighter. For chip design companies that need to place orders on this node, delivery schedules and allocation are likely to be the next issues to confirm.
Samsung is using its in-house SF4 node
Samsung is using its own 4nm node, internally known as SF4, for HBM4 base dies. ABMedia said this differs from the approach taken by some other memory makers that outsource base-die production. It also noted that ChainNews had previously reported SK hynix was considering having Intel manufacture HBM4E base dies, showing that different suppliers are taking different approaches to the same part of the stack.