TOGETHER WITH KION
Which AI investments should you scale or stop?
Knowing what you’re spending on AI is only the beginning. The bigger challenge is understanding what to do with that information once you have it.
At AWS re:Invent 2026, we’ll be showing what FinOps for AI looks like when visibility actually leads to action.
If you’re going this year and FinOps is on your agenda, make sure Kion is too.
🎤 Catch our Lightning talk:
FinOps for AI: Scale Investments That Work. Stop Ones That Don't.
Learn how FinOps teams can move from cost visibility to action with:
AI cost attribution that connects spend to specific use cases
Automated guardrails that help teams respond to spending patterns
Value tracking that ties AI investments to business outcomes
📅 Wednesday, December 2 | 1:00–1:20 PM
📍 AWS Industries Zone, Expo Floor
And if you want to talk through your own AI spend challenges while we’re there, visit us at Booth #622 or grab time with our team
AI COST OPTIMIZATION
Convert Anthropic, OpenAI and Together AI bills to FinOps FOCUS 1.4 CSV, and check any CSV against FOCUS 1.2-1.4. Runs locally
If your team uses Claude Code on AWS, a new update could make costs easier to track and control.
Claude Code v2.1.318 adds three changes worth a look from FinOps teams.
Track costs by developer: New
assume_rolesupport helps link AWS costs to each developer. This can support chargeback, where teams pay for the resources they use.Spend less on side jobs: An
effortoption for model forks lets low-effort side jobs run at a lower cost. Check which tasks need less work before using this setting.Reduce wasted work: Fixes address failed runs that repeat and sessions that get stuck.
These changes can help cut waste, though the article gives no savings figures.
For your next review, check how developer roles map to cost reports and which side jobs could use less effort.
The main benefit is clearer cost ownership, cheaper side work, and less waste.
AI PROVIDERS
Anthropic adds cost APIs and budgets, OpenAI doubles plan usage, Google adds prepaid credits
OpenAI
Pro and Business Premium plans now include more usage. Pro 100 and Pro 500 get twice the Work and Codex usage as before, so you get more for the same plan.
Anthropic
Claude apps gateway adds a cost analytics API. You can see spend by date, developer, day, or group. New usage details also make chargeback easier.
Claude Code can track AWS costs per developer. You can see who is driving spend. Small side jobs can also run at low effort, which costs less.
Pooled group budgets are here. Set a shared monthly limit for a group, on top of limits for each person.
Gemini Enterprise now supports prepaid credits. Buy credits, set auto-reload, and add spend limits. They cover Pay-as-you-go use and use above quota.
Salesforce federated data store is now generally available. Ask questions about Salesforce data without copying it first. This means less data movement and less storage.
EVENTS
FinOps Weekly Summit 2026
The first sessions from FinOps Weekly Summit 2026 are now available!
Catch the key insights, expert perspectives, and actionable strategies from our speakers.
Watch the sessions now and stay ahead in the evolving world of FinOps.
AI COST MANAGEMENT
The Token Time Bomb. AI makes software delivery cheaper…
A small AI test can turn into a large, hard-to-plan bill when thousands of staff start using it.
For federal agencies with fixed yearly budgets, the key is to track what AI gets done, not just what each token costs.
Tokens are units of data an AI model processes, but they are only part of the bill.
One staff request may trigger searches, tool calls and retries, plus costs for security, data movement and system links.
A cheaper model can cost more if it needs several tries to finish the job.
FinOps teams can put these checks in place:
Link each AI call to a team or program, and share its costs.
Set spend alerts and limits on how often AI agents can repeat steps.
Test models on real tasks, and track cost per case resolved.
Compare cloud services with in-house systems using full costs and expected use.
Use a simple script when the task does not need AI.
Judge AI spending by the cost of a successful result, not the price of a token.
EVENTS
100th Edition FinOps Weekly Event
We’re bringing the FinOps community together for a special end-of-year event filled with inspiring talks, meaningful conversations, and new connections.
This is your chance to connect with fellow FinOps professionals, share ideas, exchange experiences, and celebrate everything our community has achieved.
Be part of this special milestone and help us make our 100th edition an unforgettable experience!
AI INFRASTRUCTURE OPTIMIZATION
Claude API Pricing 2026: What 100,000 Requests Cost With and Without Caching
Your GPU dashboard can show 100% use while much of the chip sits idle. That can hide cloud waste from your FinOps team.
Ben de Mora explains that the common GPU-Util measure tracks time when any kernel, or GPU task, is running, not how much of the chip is doing work.
In NVIDIA’s example, a task can use just 20% of the chip yet score 100% on this measure.
AWS Compute Optimizer uses this signal and GPU memory use to help size instances. Its lack of a warning is not proof that your GPU spend is sound.
Before your next cost review:
Track SM activity, which measures activity across the GPU’s compute units.
Check memory use too; it answers a different sizing question.
Compare both measures over a period that reflects your normal workload.
Tell finance which measure your GPU report uses.
The article builds its case from vendor docs, not a test on the author’s own production fleet.
Treat it as a reason to check your data, not a promise of savings.
A full timesheet does not prove a full day’s output, and a busy GPU reading does not prove good value.
AI COST ALLOCATION
The Use of AI Tokenomics for Federal AI Deployment
Gísli Guðmundsson’s guide explains how to pick a payment model and track each team’s use.
Start with pay-per-token until you can measure steady demand. For work that can wait, Batch costs 50% less than Global Standard, with a 24-hour target turnaround.
For teams with steady use and strict response-time needs:
Provisioned Throughput Units (PTU) charge by the hour, even when idle.
PTU alone is not a cost-saving move; savings depend on high use and reservation discounts.
Size PTU for normal demand, then send extra traffic to pay-per-token.
Create the deployment before buying a reservation, since reservations do not secure capacity.
To keep teams accountable:
Track prompt and response tokens by team at the API gateway.
Set monthly token caps and use token counts to estimate each team’s costs.
Watch actual use, not just quota, which is not a bill.
Budget alerts do not stop spending. Use gateway limits to control token use, and buy reserved capacity only when measured demand supports it.
🎖️ MENTION OF HONOUR
AI Value Data Layer: Joining AI Cost to Value With FOCUS
Prepaid credits give FinOps teams another way to manage Gemini Enterprise spending. Google Cloud now offers these credits to customers with self-serve billing accounts.
Customers can buy credits and set up auto-reload with spend limits. The credits cover pay-as-you-go use and usage above quota across:
The Gemini Enterprise app.
Agent Platform.
AI developer tools.
For a team testing AI agents, this could mean adding funds up front and setting limits on automatic top-ups. That can help teams plan funding and keep a closer watch on costs.
One key distinction: limits on credit reloads are not the same as a cap on total usage.
Track actual spending alongside credit balances, and check what happens when funds run out.
Prepaid credits add billing control, but teams still need to watch how much AI they use.
Save 20% on AI Value & FinOps Certifications
The job market is hungry for certified professionals who can prove results. Don't let your company's budget leak because of a lack of specialization.
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
AI Economics Community
Join the Slack community where the real AI Money conversations happen
