TOGETHER WITH CLOUDZERO
Why Finance Caps What It Can't See
CloudZero's FY2027 AI spend report surveyed 252 finance leaders, 79% of them CFOs. 87% are increasing AI budgets next year. 61% say they already spend more than they can justify. When the bill lands, most reach for the one lever that works without visibility: friction.
Inside:
Approval gates are the top FY2027 change at 27%. Teams leading with gates killed initiatives at 20% vs. 7% and delayed hiring at 33% vs. 19%
CFO-led AI programs sit at the two weakest allocation tiers 65% of the time. Owning the budget is not the same as seeing it
The 23% who can allocate spend and prove ROI plan the same increases and absorb half the consequences
CLOUD COST OPTIMIZATION
Migrate First, Optimize Second: How Our AWS Migration Paid Off Early
Your Amazon Cognito bill can grow even when customers are not using your app. The cause may be a default setting, a background job, or how often your code asks for tokens. This guide shows where to check first:
Check your plan. New user pools default to Essentials. At 100,000 monthly active users, the article puts Essentials at $1,350 a month versus $495 for Lite. Keep the higher plan only if you need its added features.
Check what counts as use. A job that reads every user through AdminGetUser can make idle accounts billable. Background token refreshes and bulk user creation can also raise costs.
Cache service tokens. For 200 workers, the article shows monthly token costs falling from $3,888 to $324 when token life rises from five to 60 minutes. Set the plan in code, add cost alerts, and check charges for text messages and backup Regions.
Do not assume each AWS account gets its own free allowance. Check current AWS prices before using these figures in a budget.
Control what your system counts, not just how many customers sign in.
CLOUD PROVIDERS
Google Cloud adds spend caps, AWS tracks Bedrock costs and cuts ECS metric fees
AWS
Bedrock costs now show up in Cost Explorer, Budgets, and Dashboards. You can see spend by model, provider, and type of use.
Claude Haiku 5.5 is now on AWS. Anthropic says it costs about 75% less than Haiku 4.5 for most tasks.
ECS no longer charges for the RunningTaskCount metric. The new LiveTaskCount metric also helps you scale better.
Read All AWS Updates
Google Cloud
Spend cap budgets are here for Cloud Billing. They can pause usage when spend goes over your limit, so you don't only get an alert.
Cloud Run now supports budget spend caps (GA). Workloads can pause on their own when a budget is hit.
C4N machine types are now GA with Titanium SSD. They give you more storage speed to weigh against price.
Read All GCP Updates
Azure
No major FinOps updates this week.
Read All Azure Updates
EVENTS
FinOps Weekly Summit 2026
Join FinOps professionals at the FinOps Weekly Summit 2026 and discover how to:
Transform FinOps from reactive cost control into strategic business value
Build scalable unit economics for the AI era
Learn proven strategies from organizations managing billions in cloud spend
📅 October 20 & 21, 2026 - 9AM ET / 3PM CEST
VIDEOS & PODCASTS
How to Control AI Costs, Token Spend and ROI
We discuss the similarities between AI adoption and the cloud era, tokenomics, AI gateways, model routing, caching, hybrid AI architectures, local models, frontier models, data sovereignty and the importance of measuring business outcomes rather than simply counting tokens.
ARCHITECTURE & WORKLOAD OPTIMIZATION
Going Event-Driven? Read the Bill First.
A faster checkout can still come with a bigger cloud bill. Event-driven systems let tasks like emails and sales reports run without making customers wait, but each added task can raise costs. AWS cost model shows why teams should check the numbers before they build:
One million events sent to 12 small Lambda functions cost about $5.90.
At one billion events, that becomes about $5,900.
With 256 KB messages, the same billion events cost about $8,900.
These are model results, not a real bill. They exclude logs, data transfer, extra database reads, and other services.
Three checks matter most: how many services receive each event, how much data each message holds, and how often failed tasks try again.
Send only what each task needs, limit repeat attempts, and catch loops before launch.
Smaller messages may cut event fees but add database costs when services fetch more details. At low volume, staff time may cost more than these fees.
Use events when work can wait or traffic spikes—not as the default—and price the full flow before scaling.
COST ALLOCATION AND REPORTING
Scaling Out FinOps with a Grok bot Agent
Cloud cost reports mean little if no one knows what to do next. Frugal tested Grok bot to turn cost data into reports and tasks for the people who need them.
In one morning, the author set up three bots with seven scheduled jobs:
Finance gets a monthly email with cloud costs, customer counts, and the five customers with the highest AI token spend.
Engineers get cost-saving tasks in GitHub after a human review in Slack.
Open tasks get a reminder after two quiet weeks and a Slack alert after four.
One team gets a monthly report showing which parts of its code drive AI costs.
The bots also track task results and draft praise for cost cuts, subject to approval. There are limits. The bot could not find owners for the AWS suggestions.
This is a useful test of less manual reporting, not proof of lower bills. Start with one report, check the data, and keep a person in charge of task approval.
🎖️ MENTION OF HONOUR
Find the AWS Resources Nobody Is Using, and What They Cost You
Every cloud bill hides a few resources that nobody remembers turning on.
A storage volume outlives the server it was attached to. A spare IP address sits unused after a migration. A network gateway keeps running long after the project it served is gone.
Each one bills every hour and shows up nowhere clearly on your invoice. A new open source tool called zombiescan tackles this problem directly.
- It scans an AWS account for 22 types of forgotten resources, covering storage, compute, networking, and data services.
- It checks 17 regions in about 20 seconds and prices every finding using real AWS rates, not flat averages.
Region matters a lot here. A network gateway can cost $32.85 a month in Virginia but $67.89 in São Paulo, so accurate per-region pricing changes the real savings number.
The tool only reads your account by default. It never deletes anything until you run a separate cleanup command.
Even then, it backs up each resource first and stops if a backup fails. You get a clear report, a cost per item, and a safe path to clean up.
This kind of tool turns vague cloud waste into a specific list with dollar signs next to it, making cleanup an easy decision instead of a guessing game.
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:
Discounts on FinOps Certifications
Use code: FINOPSWEEKLY_20 to get an instant 20% discount on the most prestigious certification bundles:
FinOps Certified Practitioner
FinOps AI Value
FinOps Technology Value
FinOps Certified Engineer
FinOps Certified FOCUS Analyst
FinOps Weekly Community
Join the Slack community where the real FinOps conversations happen
FinOps Jobs
Join our FinOps Weekly Jobs Newsletter to have a weekly email with the most relevant updates every Monday
FinOps Hub
Find the right FinOps solution for your company in our Hub
Want to appear in FinOps Weekly? Talk with us