AWS raises GPU prices 15% on a Saturday, hopes you weren’t paying attention

Steven

AWS raises GPU prices 15% on a Saturday, hopes you weren’t paying attention

Summary

AWS quietly increased prices for its EC2 Capacity Blocks for ML by roughly 15% over the weekend. Specific examples include the p5e.48xlarge moving from $34.61 to $39.80/hr and the p5en.48xlarge from $36.18 to $41.61/hr in most regions, with steeper rises in US West (N. California). The change was reflected on AWS’s pricing pages and AWS said the adjustment reflects expected supply/demand patterns this quarter.

Capacity Blocks guarantee GPU capacity for scheduled training runs and are widely used by serious ML teams who need uninterrupted resources. The move breaks the long-held expectation that cloud prices only trend downwards and hands competitors a clear enterprise sales talking point. It also raises questions about enterprise discount arrangements tied to public pricing and sets a precedent for future increases on constrained resources.

Key Points

  • AWS increased EC2 Capacity Blocks for ML prices by ~15%, implemented over a Saturday.
  • Examples: p5e.48xlarge rose from $34.61 to $39.80/hr; p5en.48xlarge from $36.18 to $41.61/hr; higher increases in some regions.
  • Capacity Blocks provide reserved GPU capacity for scheduled ML training — mainly used by large-budget teams.
  • AWS says the change reflects supply/demand patterns; competitors (Azure, GCP) may use the hike as a sales advantage.
  • Enterprise Discount Program customers could see higher absolute spend even if discount percentages remain unchanged.
  • The price rise may signal a broader shift: once AWS raises prices on one constrained resource, others could follow.

Author’s take

Punchy and short: this isn’t just a billing line changing — it’s a paradigm shift. If you run ML at scale, this matters right now. If you negotiate cloud deals, expect awkward conversations. If you don’t use GPUs, it still matters because the precedent is the story.

Why should I read this?

Quick and crude: your GPU bill probably just jumped and AWS did it when fewer people notice. Read this so you can ask your account team the right questions, check your reserved/discounted agreements, and decide whether to shop your workloads around before the next quiet price change.

Context and relevance

This move comes amid global GPU supply constraints and a broader industry rush to ML workloads. Historically AWS has preferred to reshape pricing models rather than enact blunt increases; this straightforward rise breaks that pattern and may make future hikes easier. The change affects organisations running scheduled GPU-heavy training jobs, vendors negotiating enterprise discounts, and cloud purchasers watching total cost of ownership. It also hands Azure and GCP a clear talking point in competitive deals — though they may face similar supply limits.

Source

Source: https://www.theregister.com/2026/01/05/aws_price_increase/