AWS Will Reportedly Raise GPU Compute Prices in October
The projected 15% increase for reserved cloud capacity follows earlier price hikes driven by high market demand.
Updated on Sept. 30, 2026 in Data Centers
Live Poll
Do you plan to reduce your usage of cloud-based AI services due to rising costs?
Amazon Web Services is expected to implement a 15% price increase for reserved GPU capacity on October 7, 2026. This development, which remains unconfirmed by the company, affects EC2 Capacity Blocks for machine learning.
Why it matters
Rising costs for reserved GPU compute reflect intense industry demand for specialized cloud resources. AWS reportedly attributes these adjustments to ongoing supply and demand dynamics within its cloud infrastructure.
Costs for H200-based p5e.48xlarge instances rose from $34.61 to $39 per hour following the July 2026 increase, with rates reaching $49.74 per hour in the US West region.
The players
Amazon Web Services
A subsidiary of Amazon that provides cloud computing platforms, including the EC2 infrastructure used for large-scale machine learning training.
NVIDIA
A semiconductor manufacturer specializing in the graphics processing units (GPUs) essential for accelerating AI model training and data center compute.
The details
Capacity Blocks enable users to book GPU compute capacity up to eight weeks in advance, ensuring resource availability for machine learning workloads. These reservations allow the contracted compute power to be distributed across multiple accounts. The current pricing structure specifically targets these reserved blocks rather than standard EC2 instances.
Timeline
January 2026: AWS increased Capacity Block prices by 15%.
March 2026: AWS announced a goal to deploy one million NVIDIA GPUs.
July 1, 2026: AWS increased Capacity Block prices by 20%.
October 7, 2026: Expected implementation date for the reported 15% price hike.
The Tech Race
Access to advanced NVIDIA H200 chips remains the primary bottleneck for cloud providers competing to offer the most capable AI training infrastructure. AWS plans to deploy over one million NVIDIA GPUs by March 2027 to scale its capacity against rival cloud service providers.
Organizations relying on EC2 Capacity Blocks for machine learning development face higher operating expenses if this projection holds. Users should prepare for potential budget reallocations before the October 7 implementation date.
The takeaway
The sustained rise in reserved GPU pricing signals that compute scarcity is likely to persist through late 2026. Developers should track AWS pricing notifications and the company's progress toward its one million GPU deployment goal by March 2027 to forecast future infrastructure costs.
What happens next
Market analysts and enterprise customers will monitor official AWS announcements on October 7, 2026 to confirm if the reported 15% price increase takes effect.
Further reading
For broader trends in infrastructure, explore the Data Centers section.
Live Poll
Do you plan to reduce your usage of cloud-based AI services due to rising costs?
