Hyperscaler

A hyperscaler is a cloud infrastructure provider operating at massive global scale — specifically AWS (Amazon), Azure (Microsoft), and GCP (Google). These three companies control an estimated 65-70% of the global cloud infrastructure market and have committed over $200B in combined capital expenditure for 2025 alone, much of it directed at GPU infrastructure. Hyperscalers serve as both GPU cloud providers (competing with neoclouds) and as major customers for data centre capacity — leasing 10-100MW+ facilities from colocation providers and data centre developers.

Hyperscalers design custom data centres, develop proprietary networking fabrics, and increasingly develop their own AI accelerators (Google TPUs, AWS Trainium/Inferentia, Microsoft Maia). Their GPU offerings are typically virtualised, more expensive per-GPU-hour than neoclouds, but offer broader ecosystem integration (managed ML platforms, data pipelines, storage). The hyperscaler pricing premium for B200-class GPUs ranges from 2-3x neocloud rates — AWS on-demand B200 pricing reaches $14.24/hr versus $4-5/hr at independent neoclouds.

Hyperscaler behaviour sets market dynamics. Their capex plans determine NVIDIA's production allocation. Their pricing sets the ceiling for the market. Their colocation demand drives data centre development. We track hyperscaler GPU strategy as a fundamental input to all our advisory work.