AWS says that beginning October 2, it will start charging EC2 customers by the second, not hour; competitors Azure and Google Cloud charge by the minute
Jordan Novet / CNBC :
Context & Ripple Effects
This lands mid-arc in a decade-long AWS pricing offensive: the company had already cut S3 standard storage prices by up to 28% the previous November, and now extends that pressure from storage into compute metering. The wrinkle is timing — Google Cloud moved first, adding per-second billing effective September 27, five days before AWS's October 2 start date, turning what looked like a leadership move into a response.
The competitive frame is explicit in the coverage: Azure and Google Cloud already billed compute by the minute, so AWS jumping straight to per-second granularity is both catch-up and one-upmanship on the same battleground.
First-order effects
- EC2 customers running short-lived or bursty workloads stop paying for unused fractions of an hour starting October 2, with savings concentrated on jobs that run minutes rather than hours.
- AWS converts a pricing gap against Azure and Google Cloud into a granularity lead — per-second versus their per-minute — forcing the comparison onto its home turf.
Second-order effects
- Azure and Google Cloud face pressure to match per-second metering or cede the fine-grained-billing argument, since Google had just claimed that exact position days earlier.
- Finer metering makes short-duration compute cheaper at the margin, sharpening price competition for bursty workloads — the same cost-down dynamic AWS applied to storage with the S3 cuts.
Third-order effects
- Metering granularity becomes a recurring competitive lever: the same playbook later shows up when regulatory pressure over vendor lock-in pushed AWS to follow Google in dropping egress fees, suggesting cloud pricing keeps ratcheting toward customer-favorable terms as compute commoditizes.
- If per-second billing normalizes, pricing power migrates from raw instance hours to differentiated capacity — a shift visible years later in AWS charging premiums for reserved Nvidia GPU blocks while keeping its own Trainium chips priced lower.
The trend: Cloud providers are competing on ever-finer metering and customer-favorable pricing terms as commodity compute turns into a race to the bottom on unit economics.