TOGETHER WITH KION
What does data sovereignty mean for FinOps?
Your FinOps platform is part of your data sovereignty strategy.
FinOps platforms sit at the intersection of financial, operational, and infrastructure data. So when enterprises evaluate sovereignty requirements, where the platform runs and what happens to that data deserve a closer look.
A SaaS platform may introduce another vendor, another infrastructure layer, and potentially another jurisdiction into your environment. That doesn't automatically make SaaS incompatible with sovereignty requirements, but it does make the architecture worth understanding.
Our latest blog breaks down:
✓ What sovereignty actually means beyond data residency
✓ Why your FinOps platform architecture should be part of the conversation
✓ How self-hosting changes the control model
✓ What AI adoption adds to the sovereignty equation
✓ Eight questions to ask when evaluating vendors
Sovereignty shouldn’t stop cloud and AI adoption. The right architecture can help you move forward with greater control.
COST ALLOCATION
Cost Attribution in Discord’s API
Discord runs one giant codebase across hundreds of cloud deployments. That setup makes a classic FinOps problem even harder: figuring out which product feature is actually driving the cloud bill.
Standard cloud billing tools can split costs by deployment, but not by feature. Discord needed to know exactly how much it costs to run things like chat messages, streaming, or friend requests.
What they found and built:
Counting requests per feature does not work well, since some actions take much more computing power than others.
Using request time as a stand-in for cost also fails, because servers often wait on other systems instead of doing real work.
Discord compared this to getting a speeding ticket for a car someone else was driving.
Their fix was to tag CPU profiling data with the feature and endpoint being used at that exact moment.
They then matched this profiling data with real cloud billing numbers to get true costs per feature.
WEBINAR
Your Cloud Commitments Are Changing. Are They Still Saving You Money?
Join us on October 14th for a live online session exploring how to build a cloud commitment strategy that continuously adapts to changing workloads and discover:
How to balance savings with flexibility
How rolling purchases can reduce lock-in risk
Why manual commitment optimization doesn't scale
📅 October 14th
⏰ 12:00 PM ET / 5:00 PM UK
📍 Online
CLOUD PROVIDERS
AWS Cuts AI Costs 80%, Azure Adds Auto-Pause Databases, GCP Hits 48TB Memory
AWS
Security Hub now bills GuardDuty Runtime Monitoring inside its Threat Analytics plan. You no longer get a separate GuardDuty charge for it.
Aurora and RDS add new AMD-based R8a and M8a instances. You get more choices to match your database to your budget.
AWS Well-Architected Agent is in preview. This AI tool checks your setup for cost, security, speed, and reliability, then lists what to fix first.
Read All AWS Updates
Google Cloud
Spot VMs can now give a 120-second warning before shutdown. This helps cheap, interruption-friendly jobs finish safely.
BigQuery jobs explorer is easier to use. It adds status counts, resource grouping, and timeline charts so you can spot heavy workloads faster.
Read All GCP Updates
Azure
DCsv3 and DCdsv3 VMs will retire in 2029. Start planning now for right-sizing and a move to new VM types.
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 Measure AI ROI & Control AI Spending
Discover how to manage AI costs, allocate AI spending, set effective budgets, and measure AI ROI. Rick Haggard from nOps explains the AI cost journey, governance, optimization, and how to connect AI spending with real business outcomes.
CLOUD COMMITMENT MANAGEMENT
Build a Savings Plans Coverage Agent: Size AWS Compute Commitments
Savings Plans are one of the few cloud cost decisions you cannot undo. They lock you into an hourly bill for one to three years, no cancellations. That is why most companies only commit to 30 to 50 percent of their compute, leaving savings on the table out of fear.
The agent reads your AWS Cost Explorer data, past spending, plan usage, and expiration dates.
It then calculates a safe commitment number based on real daily history, not just a guess.
It accounts for things AWS tools miss, like workloads moving to spot instances, planned shutdowns, and plans about to expire.
The AI writes a clear memo explaining what to buy and why, broken into smaller chunks, so a mistake is never too costly.
CLOUD RESOURCE OPTIMIZATION
From 1.3 TB to 66 GB: On-Disk Vector Search in OpenSearch
Your vector search may need less RAM than you think—but the savings come with slower queries.
Amazon OpenSearch Service can keep compressed vectors in memory and full-size vectors on disk, then check the best matches before returning results.
For 100 million vectors with 1,536 dimensions and one replica, the article estimates RAM use falls from about 1.3 TB to 66 GB. That is roughly 95% less memory, not a promised 95% cut in your cloud bill. The main choices for FinOps teams:
Match spending to speed needs: keep fast customer searches in memory and use disk for overnight or background jobs.
Test under real load: a separate benchmark found disk-based search took about 100 milliseconds at p90, versus 24 milliseconds in memory.
Cut disk waste too: derived source removes a second stored copy of vector data.
Start small if speed matters: FP16 cuts RAM use in half with little change in search quality.
Track total cost, search quality, and response times before choosing a setup. Pay for the speed each workload needs, rather than giving every query the most costly option.
🎖️ MENTION OF HONOUR
Prometheus on AWS: A 96% Cost Cut From One Line
Your monitoring tools can run up your cloud bill even when the data barely changes. Ahmad Babaei’s team found that two Prometheus copies checked each daily S3 storage metric 5,760 times a day.
Those paid reads made up roughly 5–6% of the team’s production AWS bill. A setting marked “daily” set the data window, not how often Prometheus called AWS.
Adding interval: 15m changed checks from every 30 seconds to every 15 minutes. The cost of those CloudWatch metric reads fell about 96.5%, while total CloudWatch spend fell about 87%.
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