AWS GPU reservation price increase
Amazon Web Services announced that, effective July 1 2026, the hourly rates for its EC2 Capacity Block reservations for machine‑learning GPU instances will rise by approximately 20 %. The company cited supply‑and‑demand dynamics on its pricing documentation page.
New hourly rates
The updated rates for the most powerful Nvidia‑based instance families are as follows: P6‑B300 will be billed at $14.04 per hour, P6‑B200 at $12.355, the P5 family in U.S. regions at $5.191, the P5 family in non‑U.S. regions at $4.72, P5e at $5.97, P5en in U.S. regions at $6.865, P5en in non‑U.S. regions at $6.241, and the P4de in U.S. regions at $2.214. All other EC2 pricing remains unchanged.
AWS statement
Amazon noted that “Amazon EC2 Capacity Blocks for ML reservation prices are updated periodically based on supply and demand.”
Market context
The price hike arrives amid sustained, surging enterprise demand for GPU compute. AWS reported revenue growth of 28 % year‑over‑year to $37.6 billion in the first quarter of 2026, marking the fastest growth rate in more than three years and giving the cloud unit considerable pricing leverage with customers locked into AI training and inference workloads. Amazon has committed roughly $200 billion in capital expenditure for AI infrastructure in 2026, and Reuters reported in March 2026 that Amazon is slated to receive 1 million Nvidia GPU chips by the end of 2027 under a cloud supply agreement, underscoring the tight supply of high‑end GPUs.
Capacity Block product details
Capacity Blocks for ML are a reserved‑capacity offering that allows enterprises to secure scarce GPU instances for future, time‑bound workloads such as large‑scale model training. Because the product is reservation‑based, customers typically pay a premium over spot‑market rates for guaranteed availability. Prior to the change, the on‑demand rate for a P6‑B200 eight‑GPU node was roughly $14.24 per hour, according to pricing analysis from Spheron Network published on June 20.
Implications for Nvidia and competitors
For Nvidia, the pricing action signals robust end‑market demand for its Blackwell (B200, B300) and Hopper (H100) GPU architectures, but the higher reservation costs could lead some AWS customers to evaluate alternatives, including Nvidia‑powered offerings on rival clouds or Google Cloud’s TPU‑based instances, which Alphabet has been promoting as cost‑competitive.
Competitive landscape and customer considerations
The move will be closely watched to see whether Microsoft Azure or Google Cloud follow with comparable GPU reservation price increases. Azure, as AWS’s nearest rival in enterprise cloud infrastructure, could either match the hike or use the situation to attract cost‑sensitive AI workloads. It remains unclear whether Capacity Block reservations placed before July 1 will be honored at the prior rates or billed at the new schedule from that date forward. With the increases taking effect in less than a week, enterprise buyers face an immediate decision: lock in any remaining capacity at current rates before July 1 or absorb the higher costs as a structural feature of the AI infrastructure landscape.