Question of the Day
One question per day to look beyond the headlines.
Why does AMD’s 2027 AI-chip ramp hinge on securing memory makers, not just TSMC wafers?
Take-away AI accelerators are gated by HBM, not just logic wafers: HBM supply is capped by EUV/wafer-intensive memory fab capacity, so ramp requires locking memory allocations.
AMD's 2027 AI-chip ramp relies heavily on securing memory makers because high-bandwidth memory (HBM) is a critical component for AI accelerators, which are essential for AI inference workloads [1]. While TSMC provides the necessary wafer manufacturing for chip production, the supply of HBM is constrained due to limited EUV lithography machines and high wafer usage requirements, making it a bottleneck in AI infrastructure [1]. AMD, along with Nvidia and Alphabet, is expected to control about 85% of 2027 HBM production, highlighting the importance of securing memory access to support its AI-chip strategies [1].