AI Infrastructure Bottlenecks Position AMD and Micron for Long-Term Growth

The artificial intelligence sector continues to attract massive capital inflows, with the five largest hyperscalers projected to spend over $700 billion on infrastructure. This spending spree is evolving from general model training toward inference and agentic AI, which is fundamentally altering the hardware requirements for modern data centers. Advanced Micro Devices (AMD) is positioned to capitalize on this transition. While GPU-to-CPU ratios were previously 8-to-1 for AI model training, the emergence of agentic AI could shift this ratio toward 1-to-1, or potentially favor CPUs. This shift addresses a current bottleneck in high-performance CPU availability, which is expected to drive gross margins for the company. To maintain its edge, AMD is introducing high-core CPUs specifically designed for agentic AI and utilizing a chiplet design for its GPUs to increase high-bandwidth memory (HBM) capacity. The company has already secured strategic GPU agreements with Meta Platforms and OpenAI to fuel growth. Simultaneously, a critical shortage of HBM is driving prices higher across the broader DRAM supply chain. Micron Technology, one of the three global leaders in DRAM, is benefiting from this imbalance. Because HBM can require more than three times the wafer capacity of ordinary DRAM, scaling is difficult, leading to surging revenues and expanding gross margins. To mitigate the historical cyclicality of the memory market, Micron and its competitors are increasingly signing long-term contracts ranging from three to five years. These agreements lock in minimum HBM commitments, signaling a structural shift in the memory market driven by the relentless demand for AI accelerators.