Presented by

Want to appear here? Talk with us

AI is changing the FinOps mandate. Discover FinOps+, a modern framework for navigating growing complexity with greater visibility and scalable financial controls that support your innovation agenda.

Learn how FinOps+ extends traditional FinOps to:

Improve financial accountability across cloud, SaaS, and AI spend
Reduce manual effort through automation and AI-driven workflows
Move teams from reactive cost management to proactive financial governance

“We didn’t just need FinOps, we needed FinOps Plus. Kion gave us automation, forecasting, and policy in one platform.”
— Matt Cofran, Director of Cloud Operations, Insulet

AI bills are starting to look like early cloud bills: confusing, fast-growing, and hard to explain after the fact.

A new open framework called Big-T Notation wants to fix that.

Created by Dan Neff at Adobe, it borrows an idea from computer science, Big-O notation, and applies it to AI tokens instead of code.

Not all AI tasks cost the same as they scale. Some stay cheap and steady. Some quietly multiply because of hidden steps like reasoning loops, retries, or AI agents calling other AI agents.

Big-T sorts workloads into classes, from constant cost all the way up to unbounded loops that can run forever if nobody sets limits. The framework includes a real example.

Summarizing ten documents cost $3.04 the sloppy way, and just $0.09 once the team fixed the architecture, a 34 times drop, with no discount involved.

The fix came from separating chats, using the right-sized model, and limiting output length.

Credit bundles and flat AI subscriptions often hide true usage, making it impossible to know what is actually driving cost.

Get visibility first, classify the biggest spenders, then ask if the cost matches the value.

This looks a lot like the cloud spend story from years ago, just showing up again for AI tokens.

Read More

Claude Code now shows gateway spend limits before they become a problem, warning users of spend caps, reset times, and operator messages so teams can act before hitting budget limits.

Claude Code cuts prompt-cache costs and improves token tracking, reusing cached conversation prefixes for permission checks and breaking down token usage by type in the Stats panel.

Gemini Enterprise adds direct controls for overages and monthly spend, letting admins set spend limits and view usage costs directly in the billing console to avoid surprise charges.

Gemini Enterprise adds tracing for connector-driven workloads, giving teams visibility into which data connectors cost the most so they can optimize spend.

OpenAI adds API key views for clearer cost attribution, letting teams filter usage and cost dashboards by API key for easier chargeback across apps and teams.

GPT-5.6 Fast mode now handles requests above 272K tokens, running up to 2.5 times faster than Standard tier to improve efficiency for large-context workloads.

OpenAI updates chat-latest to the newest ChatGPT model version, a snapshot change teams should track when planning production model costs and performance.

Read All AI Provider Updates

AI agents can burn through cloud budgets fast.

The article gives an example of an AI agent stuck in a loop that spent $10,000 in Bedrock tokens in just 8 hours.

That kind of speed means old-school monthly budget checks are too slow to catch problems.

The piece lays out a hands-on plan for tracking and controlling AI costs on AWS, built around Amazon Bedrock.

Teams should tag each squad or project with its own IAM role so AWS bills show who is spending what.

Every AI request should carry small tags showing which team and task caused it, so costs can be traced back to the source.

Logs get sent to cheap storage and queried with simple tools, but teams need to watch logging costs too, since heavy tracking can add up.

The most useful part is the layered alarm system.

It checks spending instantly at the app level, watches for unusual spikes within minutes, and can automatically cut off a runaway agent's access before real damage is done.

A slower daily check acts as a final backup.

Speed and safety must move together when machines are spending the money.

Cost control now needs to live inside the infrastructure itself, not just in a report someone reads once a month.

Read More

Join this FinOps for AI webinar to learn how to control AI costs, improve visibility, implement governance, and justify AI investments with confidence.

📅 August 27, 2026
🕚 6:00 PM Spain / 12:00 PM ET

Discover how Cost Goblin enables local AWS cost analysis without additional infrastructure, third-party tools, or per-user dashboard costs. In this Summit Series interview, Victor speaks with Etienne, the creator of this open-source tool powered by DuckDB.

AI is getting cheaper by the token, but company AI bills keep climbing. That sounds backwards, but there is a simple reason.

New AI agents use far more tokens per task than old chatbots did.

One customer service chat that cost four cents in 2023 can cost over a dollar today.

Most companies are billed one of three ways.

- Per seat, which hides how much people actually use.

- Per token, which shows usage but not results.

- Per query on dedicated capacity, which is accurate but harder to manage.

The real fix is knowing your cost per unit of work, like cost per case solved or document processed.

That number lets you compare vendors, spot rising costs early, and judge if the spend is worth it.

Five ways to cut cost per query stood out:

1. Track every request from day one.

2. Send easy tasks to cheaper models.

3. Reuse answers for repeat questions.

4. Keep prompts short and focused.

5. Pick the right billing method for each task, not just whatever a vendor offers.

Cheaper AI does not mean cheaper bills unless someone is watching the actual cost per task.

Read More

AI bills are starting to look a lot like early cloud bills. They grow fast, and it is hard to tell if that growth is smart investment or waste.

OpenAI shared five steps for managing AI spend as companies move from simple chat tools to longer, multi-step AI workflows.

Know who is spending and why. Track usage by user, team, product, and model. This helps leaders spot waste, healthy experimentation, or a workflow that is quietly becoming critical to the business.

Look past the price tag. A cheap model that fails or needs retries can cost more in the end than a pricier model that gets it right the first time. Track cost per successful result, not just cost per token.

Set rules before scaling up. As AI tools connect to more systems and take more actions, leaders need clear approval steps and spending limits before usage grows out of control.

Fund AI like a portfolio. Give broad access for everyday tasks, but save bigger investment for workflows that repeat often and clearly pay off.

Match your capacity to real demand. Just like reserved cloud capacity versus pay-as-you-go pricing, AI usage should be matched to the right plan once a workflow proves its value.

The message is simple. AI spend needs the same discipline that FinOps teams already apply to cloud costs: visibility, accountability, and proof of value before scaling further.

Read More

Right now, most AI spending looks a lot like the early days of cloud waste.

Teams send every prompt, even simple ones like fixing a typo, to the biggest and most expensive AI model available.

That is like driving a Ferrari to check your mailbox.

For years, FinOps teams have asked a simple question: does this workload really need this much power.

Now that same question needs to apply to AI models. Two companies are already building tools for this.

Cursor built a router that sends simple coding tasks to cheap models and hard tasks to powerful ones.

Ramp built a smarter system that watches model performance in real time and adjusts routing based on speed and reliability.

Both tools aim for the same goal.

Send each task to the cheapest model that can still do the job well.

The problem is that AI lacks the mature tracking tools cloud computing already has.

There are no clear dashboards yet to measure if a model's output quality matches its cost.

Stop paying for AI power you do not actually need.

Read More

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: