TL;DR
Over a weekend in January, Amazon Web Services increased EC2 Capacity Block prices for ML GPU instances by roughly 15 percent. AWS said the adjustment reflects expected supply-and-demand patterns for the quarter; the move alters cost math for customers relying on reserved GPU capacity.
What happened
Over the weekend in early January, AWS implemented a roughly 15% increase to EC2 Capacity Blocks for machine-learning GPU instances. Instance examples reported include the p5e.48xlarge moving from $34.61 to $39.80 per hour and the p5en.48xlarge rising from $36.18 to $41.61; in US West (N. California) the p5e rate increased from $43.26 to $49.75. The company’s pricing page had indicated that current prices were scheduled to be updated in January 2026 but did not specify the direction of change. An Amazon spokesperson said the Capacity Block pricing varies with supply and demand patterns and that the adjustment reflects what AWS expects this quarter. Capacity Blocks are a product that lets customers reserve guaranteed GPU capacity for defined training windows. The increase follows earlier AWS announcements last year about price reductions for some GPU offerings, though those cuts applied to On-Demand and Savings Plans rather than Capacity Blocks.
Why it matters
- Organizations that reserve guaranteed GPU capacity for ML training will see higher absolute costs, even if negotiated percentage discounts remain the same.
- The timing—implemented over a weekend with limited advance detail—affects perception and supplier trust for enterprise buyers.
- Competitors may use the change as leverage in sales conversations for ML workloads.
- The move may set a precedent for future price adjustments in other constrained cloud resources if supply pressures persist.
Key facts
- AWS raised EC2 Capacity Block prices for ML GPU instances by approximately 15%.
- Reported examples: p5e.48xlarge increased from $34.61 to $39.80 per hour; p5en.48xlarge from $36.18 to $41.61 per hour.
- In US West (N. California) the p5e rate rose from $43.26 to $49.75 per hour.
- AWS’s pricing page noted that current prices were scheduled to be updated in January 2026 but did not state the direction of the change.
- An Amazon spokesperson said the adjustment reflects supply-and-demand patterns the company expects this quarter.
- Capacity Blocks are reserved, time-windowed GPU capacity for ML training jobs that guarantee availability for scheduled runs.
- Seven months earlier AWS had announced price reductions for some GPU instances—but those applied to On-Demand and Savings Plans, not Capacity Blocks.
- Historically AWS has more often altered pricing structures or dimensions rather than issuing straight increases to published rates.
What to watch next
- Whether AWS will announce similar price adjustments for other constrained resources (not confirmed in the source).
- If enterprise customers with Enterprise Discount Programs request renegotiations or raise disputes over higher absolute costs (not confirmed in the source).
- How Azure and Google Cloud respond in pricing or sales outreach to court ML workloads after this change (not confirmed in the source).
Quick glossary
- EC2 Capacity Blocks for ML: A reserved offering that lets customers book guaranteed GPU instance capacity for a defined time window to ensure uninterrupted ML training runs.
- GPU instance: A cloud virtual machine equipped with graphics processing units (GPUs) used to accelerate workloads such as machine learning and high-performance computing.
- On-Demand pricing: Cloud pricing model where customers pay for compute capacity by the hour or second with no long-term commitment.
- Savings Plans: A cloud billing model that provides lower rates in exchange for committing to a consistent amount of spend over a term.
- Enterprise Discount Program (EDP): A negotiated agreement between a cloud provider and a customer that offers discounts off public list prices, usually tied to committed spend.
Reader FAQ
Did AWS raise GPU prices?
Yes. The company increased EC2 Capacity Block rates for ML GPU instances by about 15% in early January.
Why did AWS change prices?
AWS said the adjustment reflects supply-and-demand patterns it expects this quarter.
Were customers warned in advance?
AWS’s pricing page noted prices were scheduled to be updated in January 2026 but did not specify the direction; broader advance notice details are not confirmed in the source.
Will my Enterprise Discount Program protect me from the increase?
EDPs typically offer discounts off public pricing, so a higher public rate can increase absolute costs even if the discount percentage remains the same; specific contract effects are not confirmed in the source.
PAAS + IAAS 4 AWS raises GPU prices 15% on a Saturday, hopes you weren't paying attention An anomaly or the beginning of a new trend? My bet's on the…
Sources
- AWS raises GPU prices 15% on a Saturday, hopes you weren't paying attention
- 45% Price Reduction for Amazon EC2 NVIDIA GPU Instances
- Announcing up to 45% price reduction for Amazon EC2 …
- Announcing Capacity Blocks support for AWS Parallel …
Related posts
- Show HN: Replaced Beads with tk – a faster, Markdown-based task tracker
- Bringing Cló Gaelach’s insular Irish script into the modern web
- GoGoGrandparent (YC S16) hires backend and full-stack engineers remotely