Nvidia Projects AI Compute Shortages to Persist Until FY2028
During its Q2 FY2027 earnings call, Nvidia announced that the shortage of AI compute supply will persist until at least the end of fiscal year 2028. CEO Jensen Huang provided a strong forecast of 70% revenue growth for FY2028, stating that supply capacity is the only constraint to even higher growth. This projection highlights that the global AI boom is far from slowing down, with demand outstripping supply across multiple sectors including cloud providers, sovereign AI, and enterprises. It also underscores the critical dependencies of the AI industry on physical infrastructure like wafer manufacturing, specialized memory, and power grids. The supply bottlenecks are driven by comprehensive shortages in wafer capacity, High Bandwidth Memory (HBM), and data center power. Meanwhile, demand remains highly robust, with traditional hyperscalers holding over $2 trillion in orders, and non-cloud segments (such as sovereign AI and NeoCloud providers) growing over 100% year-over-year.
## BACKGROUND
High Bandwidth Memory (HBM) is a 3D-stacked memory architecture designed to feed data to processors like GPUs at extremely high speeds, making it essential for AI workloads. NeoCloud providers are specialized cloud platforms built exclusively to offer GPU-as-a-Service, providing a cheaper alternative to traditional hyperscalers like AWS for AI training and inference.