How to manage AI token costs: FinOps for AI before the bill gets out of hand

North, which unifies cloud and AI spend, raised $60 million, and 98% of FinOps practitioners now manage AI spend. How to manage AI token costs: the hidden pricing factors, approaches and a checklist.

As companies use more AI, a new line item keeps growing: fees per model, per seat and, above all, per token. As Webrazzi reported on October 16, 2026, U.S.-based North, which brings cloud and AI spending into one platform, raised a $60 million Series B ^[1]. The news shows that the question of how to manage AI token costs is no longer a side task for engineers but a shared agenda for finance and operations. This article covers FinOps data, the hidden factors in token pricing and a checklist for companies, including those in Türkiye.

The news: North raises $60 million

According to Webrazzi, PeakSpan Capital led North's Series B; the company was founded in 2023 by Matt Biringer and Yassine Açoine ^[1]. The platform pulls AWS, Microsoft Azure and Google Cloud spend into one place, flags waste and tries to estimate the extra cost of a new feature before it goes live ^[1].

By the company's own account, North has helped customers save more than $400 million across nearly $2 billion in managed cloud spend; the North v3 release added Azure and native integrations with OpenAI, Anthropic and Snowflake ^[2]. These figures are company claims without independent verification.

Its AI product is TokenFlow, which brings model usage, attribution, budgets and financial controls into the same system used for cloud infrastructure ^[2]. The product page says TokenFlow tracks token spend by model, provider and team, can cap usage, and is in beta with select users ^[3].

Biringer's line to SiliconANGLE sums up the shift: the rest of the market still treats optimization as an engineering problem; he argues it is a business problem ^[4].

Why AI cost management became its own discipline

The FinOps Foundation's sixth State of FinOps survey, released in February 2026, is based on 1,192 respondents representing more than $83 billion in annual cloud spend ^[5]. The findings:

A quotable statistic: 98% of FinOps practitioners now manage AI spend, up from 31% two years earlier ^[5]. The report names FinOps for AI the top forward-looking priority and AI cost management the number one skill teams want to add ^[5]. Many organizations are being asked to self-fund AI investments through optimization savings ^[5].

What drives token costs: 4 hidden factors

A FinOps Foundation working group paper on token pricing explains why the advertised per-token price can be misleading ^[7]:

The paper recommends techniques such as prompt caching and batch processing, and a unit-economics view (cost per use case) instead of simply picking the cheapest model ^[7].

The picture in Türkiye: AI adoption doubled

According to TÜİK data released in October 2026, the share of Turkish enterprises using AI rose from 7.5% to 14% ^[8]. Citing the same release, Hürriyet Daily News reports that companies not using AI most often cite a lack of internal expertise (72.3%), unclear liability (66.4%) and privacy risks (65.4%) ^[9].

For most companies in Türkiye, AI cost management is still early. But as usage doubles, building cost discipline from day one is easier than dealing with a bill shock later.

Which approach fits whom?

Whatever you choose, knowing where AI agents connect matters for security as much as for cost. See our articles on employee-built AI agents and Microsoft Execution Containers. For small teams just starting out, our free AI tools guide is a useful entry point.

Checklist: How to manage AI token costs

• Do you have an inventory of which team uses which model and provider?
• Is every API key tagged to a team and a use case?
• Are monthly budgets and alert thresholds set per team and use case?
• Are input and output tokens reported separately?
• Do long conversations use context summarization or caching?
• Have you tested a smaller model for simple tasks?
• Do you track cost per use case (per summary, per ticket)?
• Is the AI bill part of finance's monthly cloud report?

Frequently asked questions

What is AI cost management?

Making spend on AI models, APIs and infrastructure visible, attributing it to teams, budgeting and optimizing it. According to State of FinOps 2026, 98% of FinOps teams now manage AI spend ^[5].

What are token costs?

The fees paid for the chunks of text (tokens) a language model processes. According to the FinOps Foundation, output tokens often cost three to five times more than input tokens, and modalities such as images and audio can cost more ^[7].

What is FinOps?

In the FinOps Foundation's definition, an operational framework and cultural practice that maximizes the business value of technology, enables timely data-driven decisions and creates financial accountability through collaboration between engineering, finance and business teams ^[10]. In 2026 the Foundation widened its mission from the value of cloud to the value of technology ^[5].

What does North do?

It is a U.S.-based cloud and AI cost management platform. According to Webrazzi, it pulls AWS, Azure and Google Cloud spend into one place and tracks AI model usage and cost with its TokenFlow product ^[1].

Where do AI bills grow the most?

According to the FinOps Foundation, the biggest hidden cost is context window creep, where conversation history is resent with every request in multi-turn chats ^[7].