This website uses cookies

Read our Privacy policy and Terms of use for more information.

Presented by

cloudzero logo

Want to appear here? Talk with us

TOGETHER WITH CLOUDZERO
A Provider's Cap Only Covers That Provider

CloudZero's smart quotas and caps set a single spend limit on a person, team, project, or agent, and count spend from every connected AI provider against it. Someone can stay under the cap in each tool and still blow past the total finance intended. This closes that gap.

How it works:

  • Each quota shows spend to date, remaining budget, and the projected date the limit is reached. Consumption counts within seconds of the call

  • Choose a soft quota that flags the overage or a hard quota that stops usage

  • Budget requests arrive with the person's spend by tool, activity, and repo, plus a peer comparison. Approve or decline in one place

  • Agents get quotas too, which matters when one loops and burns a month of budget in a day

CLOUD COST OPTIMIZATION
How A Simple S3 Design Decision Turned Into A $7M Cost

In this case study, a single AWS S3 bucket held 5.6 PB across 1.56 trillion objects, averaging about 3.5 KB each.

The team reported about $400,000 a month in storage costs and another $50,000 in request charges.

Object size changed the cost of storing the same data. In the team’s consolidation experiment, grouping snapshots into objects averaging 2.5 MB cut modeled monthly storage costs from $457,000 to $11,900 without compression or deletion.

A lifecycle rule can cost millions before it saves anything. At an earlier point, the team estimated that transitioning 720 billion objects at $0.01 per 1,000 would cost $7.2 million, before ongoing storage charges.

The transition did not run because most objects were below S3’s default 128 KB minimum for lifecycle transitions.

The team left the design unchanged, so the modeled savings were never realized.

Before approving an S3 design or lifecycle rule, model stored volume, object count, and the one-time cost of changing course.

CLOUD PROVIDERS
AWS cuts Kafka and Kinesis costs; Azure launches Cobalt 200 VMs and cheaper storage

AWS

MSK Express Brokers now run on Graviton4 M8g instances, cutting compute costs by 15% and cross-AZ transfer charges by 50%.

Kinesis Data Streams adds queues, which can cut stream processing costs by up to 43%.

Lambda Managed Instances can now scale to zero when idle, so you stop paying for unused capacity.

Read All AWS Updates

Azure

Cobalt 200 Arm-based VMs are now generally available, with up to 50% better performance than Cobalt 100, so you can run on fewer cores.

Azure Compute Spot Plus is a new Spot tier with steadier runtimes and earlier eviction warnings, making cheap capacity easier to use.

Azure Files now supports NFS shares on HDD (public preview), a lower-cost option for Linux shared storage.

Read All Azure Updates

Google Cloud

No major FinOps updates this week.

Read All GCP Updates

PODCAST
How to Control AI and LLM Costs: Align Technology's FinOps Strategy

Discover how Align Technology manages LLM costs with FinOps, team budgets, AI dashboards, token tracking and cost optimization strategies.

EVENTS
Hit the FinOps plateau? Your next big savings opportunity is hiding in your code.

You’ve optimized infrastructure, secured commitments, and eliminated idle resources. So where do the next cloud savings come from?

Join us for an exclusive session with Inon Henig, Co-founder & CTO at Ascend, to discover how inefficient application code creates hidden, recurring cloud costs and how AI-generated code is making the problem even bigger.

📅 November 24, 2026 | 6:00 PM CET / 12:00 PM EST

🎙️ Featuring Inon Henig, Benjamin Linares, Larry Advey, and Victor García.

Don't stop at infrastructure optimization. Discover your next FinOps savings opportunity.

COST ALLOCATION
Breaking Down AWS Costs with CUR, Athena & SCAD

AWS Cost Explorer shows where spend goes by service, account or region, but those views may not tell you which bucket or workload drove the bill.

Writing for AWS Community Builders, Aarush Luthra shows how to export CUR 2.0 with resource IDs to S3 and query it in Athena.

Separate S3 storage costs from request costs. Luthra’s query groups charges by bucket and API operation, with GB-months and request counts alongside cost.

It excludes charges without bucket IDs, and a lifecycle rule won’t fix a bill driven by API calls.

Include idle capacity in ECS service costs. For ECS on EC2, Luthra adds the allocated task cost and its share of unused capacity from Split Cost Allocation Data (SCAD) to reconcile with the instance’s amortized cost.

For Parameter Store, he uses naming paths to attribute API charges where resource IDs are available and keeps batch and path-based calls unattributed.

Save these Athena queries as recurring reports, join results to your service ownership map, and keep unattributed spend visible.

COST ALLOCATION & TAGGING
Maximizing return on token spend: How we're approaching FinOps for AI

Cheaper tokens do not guarantee a smaller AI bill.

Stanford University’s 2025 AI Index reports that inference costs for GPT-3.5-level performance fell more than 280-fold between November 2022 and October 2024, while Microsoft Digital says bills continue to rise as more complex agent workflows consume more tokens.

Meter before you cap. Microsoft Digital recommends tracking token use by app, workflow, agent, and user, then tagging usage to a team, project, or cost centre as it happens.

It also recommends setting budgets, alerts, and rate limits before workloads scale, using your worst month rather than your average month to plan for variance.

Track cost per outcome. Microsoft Digital defines return on token spend as business value delivered divided by tokens consumed.

It recommends cutting low-value consumption from runaway agent loops and repeated regenerations.

Start with one metered workflow: assign it an owner and track what each completed task costs before you expand it.

🎖️ MENTION OF HONOUR
A monthly AWS cost report that lands in Slack

Furkan Gungor built an AWS Lambda job that posts a cost report to Slack on the first of each month.

It compares the latest 30 days with the prior 30, then puts service changes and commitment details in threaded replies.

Rank unused commitments by dollars, not utilization.

In Gungor’s example, a Reserved Instance subscription at 46.9% utilization wasted $30.25, while one at 78.4% wasted $215.54.

The report shows coverage and utilization for Reserved Instances and Savings Plans.

Set thresholds for your account size.

For a $2 million organization, Gungor suggests reporting service changes only when they meet both a $500 minimum change and a 25% minimum change.

He estimates about $0.12 in Cost Explorer calls per run for an account with both commitment types.

Use the Slack report to prioritize follow-up, and recheck recent charges in Cost Explorer before treating the totals as final.

We Are More than a Newsletter

FinOps Learnings (30% off only for members)

Hands-on on-demand workshops led by FinOps Experts, with a limited discount for members. Learn about:

30% OFF FinOps Certs: Cyber Monday

Cyber Monday promotion was just announced, and you can now save up to 30% on FinOps certification courses and bundles, now through December 4, 2026.

Code: cyber2026finopsweekly

Individual courses 30% off (Does not apply to recertification or exam-only purchases.)

FinOps Certified Practitioner
FinOps Certified: AI Value
FinOps Certified Professional
FinOps Certified: Technology Value
FinOps Certified FOCUS Analyst
FinOps Certified Engineer
FinOps for Containers

Certification bundles 10% off: Applies to every bundle on the FinOps Foundation bundled certifications page.


Community

Join the Slack community where the real FinOps conversations happen


Read more from our Newsletters


FinOps Hub

Find the right FinOps solution for your company in our Hub

Rate Today's Newsletter

Feedback = Better Newsletter for You

Login or Subscribe to participate

Want to appear in FinOps Weekly? Talk with us

Recommended for you

View all
caret-right