AWS Plans October Update to EC2 Capacity Block Pricing
AWS has not specified the update date or percentage. Capacity Block prices are set when purchased and do not change afterward.
Amazon Web Services is scheduled to update pricing for its EC2 Capacity Blocks for machine learning in October, adding uncertainty for companies trying to secure high-end computing capacity for artificial-intelligence workloads.
AWS's public pricing page does not specify an Oct. 7, 2026, date or a percentage increase. The page says Capacity Block reservation prices are updated based on supply-and-demand trends and charged at the rate in effect when customers make a purchase.
Capacity Blocks allow customers to reserve GPU-based EC2 instances for future training, fine-tuning and inference workloads. Reservations can begin up to eight weeks in the future, and AWS charges the reservation fee upfront. Once purchased, the price does not change.
The current pricing page lists the p6-b300.48xlarge at $112.32 per hour, or $14.04 per accelerator, in US West (Oregon) and US East (N. Virginia).
Any increase would raise costs for organizations that prioritize guaranteed GPU access over flexible, on-demand capacity. Customers could respond by booking capacity before the update, shifting workloads to other regions or instance families, or using alternative cloud and in-house infrastructure.
The pricing review comes as AWS expands its AI infrastructure. AWS and NVIDIA plan to deploy more than 1 million NVIDIA GPUs beginning in 2026 and an additional 2 million NVIDIA GPUs in 2027–2028. The expansion may improve long-term availability, but the October pricing update will determine whether reserved capacity becomes more expensive in the near term.
