Three Tech Companies Control Majority of High Bandwidth Memory Supply for AI
High bandwidth memory (HBM), a critical component in AI infrastructure, is facing significant supply constraints. According to Morgan Stanley, three companies—Nvidia, Alphabet, and Advanced Micro Devices—have secured 85% of next year's HBM production capacity. This gives them a competitive edge as demand for HBM grows due to the shift from training to inference workloads in AI applications. The limited supply is driven by several factors: EUV lithography machines required for advanced chip manufacturing are scarce, wafer usage for HBM chips is three times higher than standard DRAMs, and building new clean rooms takes years. These constraints make access to HBM a key differentiator among tech firms competing in the AI space. Nvidia has secured 37% of next year's supply through partnerships like its $500 billion deal with SK Hynix. Alphabet controls 36%, using it for custom Tensor Processing Units (TPUs), while AMD also benefits from early access to this scarce resource.
Tickers: $AMD $ASML $GOOG $GOOGL $MU $NVDA $SKHYV