Unexpected AI costs altered business decisions at 62% of surveyed enterprises over the past year, while 40% required board-level escalation, according to a new report from Mavvrik and Benchmarkit.
Unexpected AI costs reach the boardroom
The 2026 State of AI Cost Governance Report found a growing disconnect between AI adoption and financial governance.
The report is based on feedback from 396 enterprise organizations across IT/SaaS, financial services, retail, manufacturing, and other industries between April and May 2026. It found that while organizations know what AI produces, they’re missing the complete cost.
62% of organizations said a business decision was altered by an unexpected AI cost this past year, with 40% requiring board-level escalation. Meanwhile, 33% implemented emergency spending freezes and 25% delayed or canceled an AI initiative outright.
Forecast accuracy continues to decline, with only 11% of organizations forecasting AI costs within ±10% — down from 15% in 2025.
Developer AI tools create additional cost blind spots
Developer tools, data platforms, agentic workloads, and private infrastructure generate expenses that don’t show up in the finance teams’ numbers.
”AI is fundamentally changing how infrastructure is consumed and how costs accumulate,” said Sundeep Goel, CEO of Mavvrik.
”What used to be predictable IT spending is now dynamic, distributed, and increasingly difficult to attribute. The organizations succeeding with AI are not necessarily spending less, but they’re implementing the governance systems necessary to understand where costs originate, how they scale, and how they impact margins.”
2026 state of AI cost governance report: key findings
Board escalations, emergency spending freezes, and products repriced due to incorrect unit economics are just some of the consequences that appear in the report.
Here are other key findings:
- 89% missed AI forecasts by more than 10%.
- Self-assessed maturity scores showed little correlation with actual forecasting performance.
- 68% of organizations now operate hybrid AI environments spanning cloud and on-prem infrastructure, up from 61% in 2025.
- 44% include on-premises AI infrastructure in cost reporting, up from 35% in 2025.
- 15% of organizations running agentic workloads cannot attribute agent-related costs at any level.
The costs of developer AI
The report also found that developer AI spending is becoming increasingly unmanageable, with 98% of organizations surveyed using AI coding assistants. Organizations also used an average of 2.4 AI coding tools simultaneously.
According to 39% of respondents, coding tool costs exceeded expected license or usage costs. Meanwhile, only 42% of organizations included developer AI tools in AI cost reporting.
Even AI-native companies are not immune from these trends, with AI-native application companies reporting the worst forecasting performance of any segment.
Only 3% of organizations forecasted AI costs within ±10% — despite over half of respondents describing themselves as ”advanced” in AI governance.
Data platforms drive unexpected AI spending
Organizations cited data platform overages as the most frequently cited source of unexpected AI spending.
Meanwhile, costs associated with data movement, storage, and processing often exceed expectations, with 33% of organizations implementing emergency spending freezes.
25% of organizations delayed or canceled AI initiatives due to unexpected costs, while 49% were forced to reprice AI-powered products due to unforeseen cost increases.
”Everyone is talking about AI to ROI, but ROI is math,” said Ray Rike, CEO of Benchmarkit. ”You can’t accurately calculate ROI if you don’t know your costs. Most organizations are still operating with incomplete cost visibility across their AI stack, which means many of the ROI conversations happening today are based on assumptions rather than financial reality.”