AMD Benefits from SK Hynix’s $640 Million HBM Expansion as AI Demand Explodes
AMD•In 2014 AMD partnered with SK Hynix to launch the world’s first high-bandwidth memory (HBM) chips for GPUs. The ensuing AI boom driven by ChatGPT in 2022 sent HBM demand soaring, securing AMD a stable high-performance memory supply for its data-center GPU accelerators.
1. Historical Partnership on HBM Development
AMD collaborated with SK Hynix in 2014 to introduce the industry’s first high-bandwidth memory chips, targeting next-generation graphics and compute workloads. This early move positioned AMD as a pioneer in leveraging stacked memory technology for enhanced GPU performance.
2. AI Boom Spurs Surging HBM Demand
The launch of ChatGPT in 2022 triggered an unprecedented surge in AI model training, driving global demand for HBM in data-center accelerators. SK Hynix’s capacity, once underutilized, has become critical to meeting requirements from both Nvidia and AMD.
3. Impact on AMD GPU Product Line
AMD’s MI100, MI250 and upcoming MI300 accelerators incorporate HBM2 and HBM3 to deliver higher bandwidth and lower power per watt, strengthening its competitive stance in HPC and AI markets. Reliable memory supply supports tighter production schedules and potential margin improvements.
4. Future Outlook and Supply Stability
SK Hynix’s 880 billion won investment in a dedicated HBM packaging facility boosts production capacity and could drive down per-unit memory costs. AMD stands to benefit from this expanded supply ahead of planned GPU launches, underpinning revenue growth in AI and data-center segments.
