Amazon Web Services (AWS) has increased pricing for its EC2 Capacity Blocks for ML service amid rising demand for AI computing infrastructure, according to a recent announcement.

The service, which enables customers to reserve GPU capacity in advance to ensure uninterrupted AI workloads, will see hourly rental rates rise by approximately 20% beginning in July, following an earlier 15% price increase introduced in January.

EC2 Capacity Blocks for ML are widely used by organizations training large AI models and fine-tuning foundation models that require guaranteed access to high-performance GPUs.

AWS said the pricing update reflects changing supply and demand conditions and noted that the increase applies only to this reservation-based offering, while other AI cloud purchasing options continue to maintain fixed prices.

The move comes as rising hardware costs ripple throughout the technology industry. Apple and Xbox have recently implemented price increases tied to higher memory costs, while Elon Musk has also warned about unprecedented increases in memory chip pricing.

Analysts point to shortages of high-bandwidth memory (HBM) as one of the industry’s biggest constraints. Because HBM is essential for advanced AI processors, limited memory production restricts GPU manufacturing and slows data center expansion.

Meanwhile, the ongoing supply imbalance has boosted memory chip makers such as Micron and SK Hynix, with investors expecting AI-driven demand to support elevated prices and tight supply conditions for years to come.