AI spending now dominates FinOps agendas as companies struggle to measure returns
Financial operations teams are racing to manage artificial intelligence costs, with 98% of FinOps practitioners now responsible for AI spending-up from 31% two years ago, according to the FinOps Foundation's 2026 State of FinOps report.
The shift reflects how quickly generative AI has moved from experimental projects into operational budgets. Most organizations remain in proof-of-concept phases, still figuring out how to control expenses for services like ChatGPT and Google's Gemini.
Tokens are becoming the standard cost unit
The AI industry has standardized on tokens as the primary billing metric. A token represents a fundamental unit of data processed by the AI model, and optimizing queries to use fewer tokens is emerging as the most direct way to control costs.
Some companies are now treating tokens like corporate currency. They assign developers monthly token budgets-similar to how teams manage compute resources-forcing engineers to think about efficiency before writing code. This approach creates accountability earlier in the development process, before workloads reach production.
FinOps teams are building pricing calculators and offering pre-deployment guidance to platform engineering and enterprise architecture teams, shifting cost optimization left in the software development lifecycle.
Building AI in-house carries hidden expenses
Off-the-shelf AI services offer simplicity. Homegrown AI solutions do not. Building custom models requires securing graphics processing units (GPUs) in data centers or cloud environments, plus the electricity to power them continuously.
Nearly half of FinOps teams now actively manage physical data center costs to capture the full footprint of AI computing. In the Asia-Pacific region, new climate laws are forcing companies to measure and reduce carbon emissions, tying cost management directly to environmental impact.
FinOps teams are increasingly collaborating with ESG and sustainability teams, recognizing that optimizing cloud usage simultaneously lowers bills and carbon footprints.
Most companies lack clear ROI targets
Despite substantial investments, many organizations cannot articulate what AI is delivering financially. Only 7.5% of enterprises have integrated FinOps into AI projects, according to IDC.
Practitioners are pushing companies to calculate unit economics. A bank processing home loans, for example, could establish a baseline cost per loan and measure how AI changes that metric. The goal is quantifiable improvement: more loans processed, lower per-unit costs, or both.
The Technology Business Management (TBM) model provides a framework for this work. It combines traditional IT financial management with FinOps, creating a single view of costs across different AI services and deployment models-whether SaaS, on-premise, or cloud-based.
Culture, not technology, is the biggest obstacle
Organizations across mature cloud markets and technology hubs face the same challenge: getting people to adopt cost-conscious practices. Executives may not fully support the effort. Engineers may resist scrutiny of their resource usage.
Ironically, managing AI costs may require more AI. Anomaly detection algorithms could flag misconfigured instances before they generate bill shocks. Natural language chatbots could replace dashboards, letting executives query spending data conversationally.
But tools cannot overcome organizational resistance. Achieving buy-in across all levels-where executives, engineers, and finance teams align on cost discipline-remains the hardest part.
For management teams overseeing technology spending, understanding generative AI and LLM cost structures is now essential. Finance leaders may benefit from exploring an AI learning path for CFOs to better manage these emerging expenses.
Your membership also unlocks:
