Cloudera Report Finds 89% of Enterprises Cut AI After Costs Peaked

Token prices fell to $0.95 per million by late October 2026, but the timing helped few. Cloudera’s “State of Enterprise AI 2026” report, published October 28, shows nearly nine in ten enterprises had already cut or redesigned AI programmes after costs peaked above $2 per million in June, decisions that do not reverse easily once headcount and roadmaps adjust.

The ROI Gap

A separate IDC analysis from October 2026 found that nearly 70% of organisations acknowledge a gap between AI spending and measurable financial outcomes, with only 15% achieving an ROI above 100%.

Gartner’s October 19, 2026 “2027 CIO and Technology Executive Agenda” survey of 2,002 CIOs found only 13% reported significant or transformative value from AI. Alicia Mullery, VP Analyst at Gartner, pointed to the need to define use-case value before deployment and set clear criteria for exiting underperforming projects, saying volume is not value. Gartner links better outcomes to CIOs who apply stopping rules at both pilot and production stages, a practice only 31% have in place at the pilot stage and 17% at production. Speed is not the problem: 81% of companies in the survey report faster decision-making. Converting that speed into measurable business change is where programmes stall.

Where the Hidden Costs Live

AI cost management firm Mavvrik found in its 2026 “State of AI Cost Governance Report” that 62% of organisations experienced an unexpected AI cost that materially altered a business decision in the past year. Of those, 40% required board-level escalation, 33% triggered emergency spending freezes and 25% delayed or cancelled an initiative outright. Token costs were the top driver, cited by 43% of respondents.

Shadow AI compounds the problem. A 2025 TELUS Digital Experience survey found that 68% of employees accessed generative AI assistants through personal accounts, with 57% admitting to entering confidential information into public tools. That behaviour creates two cost lines: underutilised enterprise licences and elevated breach exposure. IBM’s Cost of a Data Breach Report 2025 identified shadow AI as a top-three breach cost driver for the first time. Beyond token spend, the majority of engineering effort in production AI programmes goes to retrieval pipelines, evaluation, guardrails, access control and audit logging, foundational layers rarely costed in the prototypes that secure initial funding. Accenture reported cumulative advanced-AI bookings of $11.5 billion through Q1 FY2026 across more than 1,300 clients; a significant portion of that work likely involves rebuilding exactly those layers.

Unit Economics Over Spend Tracking

Gartner’s prescription is a shift from tracking aggregate AI costs to “AI unit economics”, linking token consumption directly to measurable outputs and business outcomes, then rightsizing spend rather than simply cutting it. The visibility gap makes that difficult. Deloitte’s October 2026 report found that while 71% of companies track total inference and API spend, only 51% track token volume and 44% track per-unit inference cost. Without per-unit cost data, accurate ROI assessment at scale is close to impossible.

Preparing business processes for AI agents and scaling multi-agent adoption remains a separate challenge, one that leaders expect to drive widespread process redesign within four years, but have not yet solved at the governance layer.

Governance as a Cost Control

Full compliance requirements for high-risk AI systems under the EU AI Act phase in through 2027 and 2028, carrying potential fines of up to 7% of worldwide annual turnover or €35 million, whichever is higher, for prohibited practices. Those liabilities are pulling procurement and legal teams into AI budget conversations they previously sat out.

Sundeep Goel, CEO of Mavvrik, attributed the situation to a lack of spending discipline, describing it as a significant problem of wasted money.