FinOps for AI: How CIOs are navigating tokenomics

AI pushes FinOps into a new era

For years, FinOps focused primarily on managing cloud infrastructure costs. As organizations adopted and matured their FinOps practices, many became able to forecast cloud spending with a high degree of accuracy and implement consistent optimization strategies. Some organizations also began applying FinOps principles to SaaS and other technology spending categories.

The mature FinOps teams on the cloud are usually forecasting in 1 to 3% of where they're going to land. [With AI] it's totally blown out.

Now, practitioners say AI is forcing the next major evolution -- one far less predictable than cloud. Organizations are already seeing AI budgets exceed forecasts by two to three times within a single year as adoption scales faster than expected.

"We solved this problem in cloud FinOps. The mature FinOps teams on the cloud are usually forecasting in 1 to 3% of where they're going to land. [With AI] it's totally blown out," Storment said.

Unlike cloud spending, which was largely driven by engineering teams, AI spending is spreading across entire organizations. Employees in sales, marketing, legal and other business functions can now consume AI services directly, making demand harder to predict and costs more difficult to forecast.

Additionally, FinOps teams are being pulled into a more technically complex environment. AI cost drivers now include model usage, tokens, GPUs, orchestration frameworks and agentic workflows -- each with different measurement and optimization challenges.

"It's going to change the daily life of the FinOps team and move them into needing to be a lot more technically aware," said Richard Hoyer, co-founder and CEO of FinOptik, a FinOps consulting and services firm.

However, FinOps for AI is still in its early stages. Some describe it as a return to the "crawl" phase after years of maturity in cloud cost management.

"We're so early in AI. People were like, 'We're runners in FinOps -- we figured it out,' and now they're back to the very beginning with AI," Storment said.

Despite the complexity, the direction of travel is clear: FinOps is expanding beyond cloud and SaaS into AI cost management because business leaders care about total technology spend, not individual spending categories.

FinOps ownership gets messy

Organizations have not settled on a clear owner for FinOps for AI. Historically, FinOps has lived under the CIO or CTO, with finance playing a close advisory role.

"In our data last year, over 80% of the teams in our community were CIO or CTO aligned," Storment said. "There are some giant Fortune 10 to 20 FinOps teams that live in finance that we're aware of, but it's an outlier."

Some organizations use a shared model that spans IT and finance. However, FinOps has traditionally remained a technology discipline because it sits closest to the engineers making infrastructure decisions.

We're seeing a scramble in all these companies about not only who owns the cost but who owns the strategy.

"Engineers don't like it when finance people tell them how to run their infrastructure," Storment said.

That long-standing ownership model is becoming less clear as organizations expand AI initiatives across the business.

"We're seeing a scramble in all these companies about not only who owns the cost but who owns the strategy," Storment said.

Chief AI officers, business leaders and technology leaders are all claiming a stake in AI initiatives, making ownership less clear than it was during the early cloud era. As a result, organizations are increasingly using discussions around AI economics and tokenomics to bring together technology, finance and business leaders.

"What FinOps did six to 10 years ago was build bridges between engineering leaders and finance leaders," Storment said. "Now we're seeing a new set of bridge building for tokenomics."

Accountability becomes more distributed

While ownership remains unsettled, some practitioners see organizations moving toward a distributed accountability model. Under that approach, individual teams remain accountable for their spending decisions, while FinOps teams provide governance, visibility and guidance across the organization.

Not everybody is an expert in finance, but everybody can be accountable for their area and decisions.

"Not everybody is an expert in finance, but everybody can be accountable for their area and decisions. However, you'll still need a FinOps team to be the responsible party," said Udam Dewaraja, founder and CEO of StitcherAI and co-founder of the FinOps Foundation's FOCUS billing specification.

What CIOs can do now          

Tokenomics remains an emerging discipline, and organizations are still working to understand the fundamentals. While industry groups and vendors continue to develop standards, CIOs can start by improving visibility into how teams are using AI across the organization.

"Folks are just trying to get their arms around how to budget for it for next year," Storment said.

Recent reports about Uber's AI spending show what can happen when AI adoption outpaces budgeting and forecasting. The company reportedly exhausted its projected annual AI budget by March, underscoring the challenges many organizations face as AI usage scales across the business.

Beyond understanding how much they are spending, organizations increasingly need to understand who is generating those costs and where they originate. FinOps leaders must look for ways to allocate AI costs to individual users, developers and projects, rather than treating AI as a single line item, Storment said.

That level of visibility could become increasingly important as AI usage spreads across business units. Without it, organizations may know costs are rising but have limited insight into which teams, applications or initiatives are driving spending.

Vendors are beginning to respond to this need with more granular usage visibility. Anthropic provides enterprise analytics that enable organizations to export per-user usage and spend data for Claude, while AWS has introduced more detailed cost-attribution capabilities for Amazon Bedrock. These tools improve visibility into AI usage and spending across users, models and workloads.

Ultimately, organizations want to connect AI spending to business value. However, the industry is still laying the groundwork by improving visibility, allocation and governance before moving on to more advanced optimization efforts.

Tim Murphy is a site editor and writer for the IT Strategy team at TechTarget.