Connecting memory, foundry, and packaging
Samsung Electronics and Broadcom announced a memorandum of understanding to expand collaboration on next-generation AI infrastructure. The memory work is intended to include HBM supply for future Broadcom AI accelerators. The foundry track focuses on 2-nanometer and below processes, with potential 2.3D and 2.5D advanced packaging built around Samsung's 2nm technology.
The companies said they expect the collaboration to be estimated at more than $200 billion across memory and foundry over five years through 2030. That figure is a projection in the announcement, not recognized revenue or a disclosed firm order backlog.
Why packaging matters for AI accelerators
Large AI accelerators are constrained by more than logic-die speed. Bandwidth between HBM and compute, die-to-die interconnect density, power delivery, and thermal removal can limit the complete system. Co-designing memory, leading-edge logic, and packaging creates an opportunity to optimize performance and energy efficiency at the system level.
Limits of the MOU
An MOU does not establish final product volumes, customer qualification, yields, pricing, or production dates. The stated value is an estimate of potential collaboration. Follow-up supply agreements, tape-outs, packaging validation, customer adoption, and recognized revenue will provide stronger evidence of execution.