The Growing Compute Shortage [pdf]
The document examines the emerging shortage of compute power, driven by surging demand for AI and machine learning workloads. It highlights that supply constraints for advanced chips and data center capacity are creating a bottleneck, potentially slowing innovation and economic growth. The paper discusses implications for businesses and the need for strategic investment in compute infrastructure.
Background
- Apollo Global Management is a giant alternative asset manager ($700B+ in assets). This is an Apollo research paper, not an independent report — it reflects their investment viewpoint.
- "Compute" = the processing power (GPUs, data centers, cloud infrastructure) required to run AI. The "shortage" thesis: AI demand is far outstripping supply of chips, data centers, and power.
- Key bottlenecks: NVIDIA's near-monopoly on AI chips; limited TSMC manufacturing; 12-24+ month data center build times; strained power grids.
- This matters because it frames AI's growth as constrained by physical infrastructure, arguing the shortage is structural, not temporary — with big investment implications for data centers, utilities, and chip companies.