As demand for AI computing power continues to surge, Amazon Web Services (AWS) has once again raised prices for select AI cloud services. AWS recently announced that starting this July, pricing for EC2 Capacity Blocks for Machine Learning will increase by approximately 20%. This marks the second price adjustment within six months for this service, following an earlier hike of about 15% in January. EC2 Capacity Blocks for Machine Learning enables enterprises to pre-book GPU compute resources, securing stable computational capacity during specified timeframes and preventing training job interruptions due to resource shortages. The service primarily targets enterprise customers requiring large-scale GPU resources, such as those involved in training foundational models or fine-tuning generative AI models. Regarding the price adjustment, AWS stated that reservation fees for EC2 Capacity Blocks for Machine Learning will be adjusted periodically based on market supply and demand dynamics. The current changes reflect sustained growth in global demand for AI GPUs. However, AWS emphasized that this price increase applies solely to this reserved procurement method; clients retain access to other cloud purchasing options, and the company remains committed to maintaining fixed pricing for related products. Industry observers widely believe this round of hikes underscores ongoing tightness in AI infrastructure supply and demand. In recent years, rapidly growing needs for large-model training and inference have driven persistent shortages of GPUs and associated hardware. Meanwhile, constrained supplies of critical components like High Bandwidth Memory (HBM) have further elevated costs for AI servers. According to Peter Berezin, Chief Economist at BCA Research, limitations in HBM production capacity restrict advanced GPU manufacturing volumes, ultimately slowing data center construction rates. Against a backdrop of scarce GPU resources, major cloud providers including AWS, Microsoft, Google, and Oracle possess stronger pricing leverage, enabling them to pass some of their rising infrastructure costs onto customers. Concurrently, robust demand for AI infrastructure continues to benefit upstream memory chip manufacturers. Driven by positive market expectations, shares of leading HBM suppliers such as Micron Technology Inc. and SK hynix Inc. have reached record highs recently. Investors broadly anticipate that heightened demand fueled by AI adoption will sustain a seller's market for high-performance storage solutions over the coming years.