Presented by

Want to appear here? Talk with us

Take the 2-minute benchmark and find out. Five questions on who owns AI spend, how fast the numbers reach finance, and what your board expects. Then see your cohort among 260 senior finance leaders.

Your results show:

AI companies are racing to slash token prices, but that price war will not shrink your AI bill the way you might think.

Meta, OpenAI, and other providers just cut prices sharply on new models. Sounds like good news for budgets. But the article makes a sharp point: cheaper tokens do not mean cheaper AI agents.

AI agents do not answer a question once. They plan, search, call tools, check their work, and sometimes retry the whole thing. Each of those steps burns tokens.

A workflow can use 20 times more tokens even as the price per token drops 75 percent. Do that math, and your total bill can still go up fivefold.

This is the same pattern FinOps teams already know from cloud computing. Storage and compute got cheaper every year for a decade.

Total cloud bills still climbed, because usage grew faster than prices fell. The real fix is tracking cost per completed task, not cost per token.

A pricier model that finishes the job on the first try can beat a cheap model that needs retries and human cleanup.

And token charges are just one line on the invoice. Storage, search tools, and human review all add up too.

Before your next AI vendor call, ask for real numbers on retries and task success rates, not just the rate card.

The team with its own cost-per-outcome data will win the negotiation.

Read More

GPT-5.6 pricing cut dramatically, with Luna down 80% and Terra down 20%, giving teams much more room in their token budgets.

New Fast mode replaces Priority Processing, offering quicker responses without the old overhead, so you get speed without extra cost.

Gemini Enterprise pay-as-you-go edition is now generally available, letting you pay only for actual feature usage instead of buying fixed license pools.

Built-in spend limits and usage monitoring come with the new edition, helping teams cap costs before they get out of hand.

Expanded tracing for Gemini Enterprise data connectors adds detailed spans for tool execution and connector calls, making it easier to spot where AI budget is being spent.

Read All AI Provider Updates

AI projects cost more than most teams expect, and this article breaks down where the money actually goes.

Four cost buckets to track:

Some gains are easy to count, like lower costs or faster sales. Other gains are harder to measure, like better decisions or happier employees. Both types matter for the full picture.

One good example comes from Pernod Ricard. The spirits company mixed outside experts with its own team. This let them move fast while still building skills in-house.

Their new AI tool helped marketing teams see real return on ad spend for the first time, instead of guessing.

The main advice is simple: start small, test results, then grow.

Companies that rush big AI spending without proof often waste money. Companies that wait too long miss real gains. The middle path works best.

Treat AI spending as an ongoing investment, not a one-time purchase. Track costs closely and tie every dollar back to a clear business goal.

That is how AI turns from a risky bet into a real advantage.

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 20, 2026
🕚 6:00 PM Spain / 12:00 PM ET

Learn how FinOps is evolving to tackle AI cost management and tokenomics. Discover actionable strategies for prompt engineering, model right-sizing, and cost governance across non-technical teams.

AI agents can run up big bills without anyone noticing. They can retry the same task, call the wrong model, or send bloated prompts, all while looking perfectly healthy on paper.

A developer ran into this problem and built a tool called BurnRate to fix it.

- It adds real dollar costs to the technical data that teams already collect on AI usage.

- It shows which agent, model, or service is spending the most money right now, not just how many tokens were used.

- It connects to SigNoz, a monitoring tool, so teams can watch spending in real time and dig into what caused a spike.

- It goes one step further with a feature called Cost Guard, which investigates cost spikes on its own and can automatically slow down or stop an agent that is burning money.

In one test, an agent got stuck repeating the same work over and over.

The system caught it, found the cause, and shut it down automatically, bringing costs back to normal without a person stepping in.

Cost needs to be tracked just as closely as speed or uptime.

As AI agents get more independent, teams will need systems that do not just show spending problems, but actually help stop them.

Read More

Who should set the AI budget? That question is causing real headaches for FinOps teams right now.

A group of practitioners met to talk about AI cost estimation. They found something bigger than a math problem. Nobody agrees on who owns the budget number.

Engineers cannot predict a tenfold jump in AI use.

Their planning tools were never built for that kind of leap. Finance cannot set the number alone either. They see the total bill but not where the waste hides.

Much of today's AI work is still experimental.

Tools bought last year are already being replaced. That means most teams can only plan about twelve months ahead.

Inside engineering, AI value is easy to track. Code output, incidents, and cycle time all leave a paper trail. Outside engineering, in marketing or HR, the value is real but invisible.

There are no tickets or hours to count. Waste hides in familiar places. Costly models get used for simple tasks. Idle tools and repeated retries quietly burn cash.

One team found a coding tool secretly picking an expensive model just to check out a branch.

The fix is not more human review. It is shared ownership. Finance sets the spending limit. Engineering builds the controls that enforce it.

FinOps connects the two, showing where money leaks and where it can be saved.

A budget without a quota is just a number, not a control.

AI spending needs teamwork, not guesswork.

Read More

The Linux Foundation just launched the Tokenomics Foundation.

Its goal is simple: help companies figure out what AI actually costs and what it's really worth.

Thirty big names joined at launch, including IBM, Oracle, SAP, JPMorganChase, Accenture, and FinOps tool makers like Flexera, DoiT, and Vantage.

AI spend is growing fast, with token use expected to jump 24 times by 2030.

But there is no shared way to measure the true cost of AI; tokens are just one part of a much bigger bill that includes compute, storage, data, and people.

Many leaders cannot answer a simple question: What does one AI task actually cost, and what do we get back for it?

The new group plans to build common definitions, a full cost model for AI, and ways to measure value beyond just token counts.

It will also add token cost tracking to FOCUS, the billing standard many finance teams already use.

Education and certification programs are on the roadmap too, so practitioners get trained on these new rules.

This effort could give finance and tech leaders a shared language for AI spend, much like FinOps did for cloud costs.

That shared language may be the difference between AI spending that pays off and AI spending that just adds up.

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: