Nearly nine in ten organisations surveyed by Cloudera say token costs have forced them to scale back or redesign AI plans this year. The findings suggest that for many businesses, the question is no longer just what AI can do, but whether its value justifies the bill.
Watch Desk analysis
What happened
Cloudera’s State of Enterprise AI 2026 report, covered by ITPro, found that 89% of respondents had scaled back or redesigned AI plans because of token costs. Nearly half, 45%, said they had moved AI workloads as a result. Token prices fell from more than $2 per million in June to $0.95 on 25 October, according to the LLM Token Expenditure Index cited in the report.
The spending is not yet translating into clear wins for everyone. More than half of respondents said their AI investment had increased, while 23% said AI generated more value than it cost. Another 27% reported little or no measurable impact, and 41% had encountered higher-than-expected costs.
Key findings
- Plans are being revised
89% said token costs had led them to scale back or redesign AI plans. - Workloads are moving
45% said they had shifted AI workloads because of cost volatility. - Clear returns remain uncommon
23% said AI generated more value than it cost; 27% reported little or no measurable impact.
Why it matters
A lower price per token does not automatically mean a cheaper AI operation. Companies are still weighing usage, workload choices and returns, and the survey suggests cost volatility has already changed where some work runs. For teams planning deployments, tracking the bill and measuring the result are part of the technical decision, not paperwork to sort out afterwards.
Our read
The headline is not that AI has become unaffordable; it is that plenty of organisations are changing course while trying to make the economics add up. That is a more useful reality check than another grand forecast. Treat these survey figures as a snapshot of respondents’ experience, not a universal verdict on every AI project.
What to watch
- Whether the next round of reporting shows fewer plans being cut or redesigned.
- Whether workload moves lead to measurable savings or simply shift costs elsewhere.
- Whether more organisations can show clear returns as they improve cost tracking.
Discussion spark: If an AI project’s costs are volatile and its returns are hard to measure, should businesses pause it until the numbers are clear, or keep investing while they learn?
Sources and evidence
- Source update (28 October 2026, 08:25 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.