Nearly nine-in-ten (89%) organizations have scaled back or been forced to redesign AI plans due to token costs this year, according to research from Cloudera.
Findings from the firm’s State of Enterprise AI 2026 report show enterprises are still struggling with AI-related spending, despite a notable decrease in token costs over the last three months.
Brian Martin, CTO of Applied AI at Cloudera, said the findings highlight a “new reality for enterprise AI”, adding that “what works today may not make economic or operational sense tomorrow”.
“Costs change, models evolve, regulations shift, and the right place to run a workload changes with them,” he commented.
“Leaders need an AI strategy built for that reality, one that gives them the freedom to adapt without giving up control of their data, governance, or spend.”
Cost volatility spooks IT leaders
The study specifically highlighted cost volatility as a key challenge for enterprises, with nearly half (45%) having moved AI workloads as a result.
According to Cloudera, this shows that “AI’s price tag is starting to shape its path through the enterprise” – and that could spell further bad news in terms of ROI and adoption rates.
Sign up today and you will receive a free copy of our Future Focus 2026 report - the leading resource for IT decision-maker insight on priorities and investment areas in AI, security and more.
Last year’s report, for example, found cost was the leading barrier to AI implementation, and that trend has intensified across 2026 amidst surging token costs.
Figures from the LLM Token Expenditure Index show prices hit over $2 per million tokens in June this year. As of 25 October, the index shows token prices stand at $0.95 per million tokens.
That price decrease will offer little solace for IT leaders, however, many of whom have been battling costs throughout the first six months of the year. Cloudera said the cost of the technology has placed huge pressure on enterprises to “demonstrate returns and manage the ongoing cost of AI at scale”.
More than half (55%) of respondents said their AI investment increased, yet despite greater spending that is yet to translate into “clear returns”. Indeed, 58% of respondents the value generated by AI initiatives is roughly equal to what they cost.
Fewer than one-quarter (23%) said AI generates more value than it costs.
From a business impact perspective, the situation is roughly the same, according to Cloudera, with IT leaders reporting mixed returns on the technology.
“More than a quarter (27%) report little or no measurable impact from their AI initiatives, while 41% have encountered higher-than-expected costs,” the report states.
Cloudera added that for those investing heavily in the technology, its findings “raise questions about how quickly that spending will translate into meaningful returns”.
A year of AI costs
The study from Cloudera is the latest in a string of reports questioning the costs associated with AI over the last year - and there have been a number of high profile incidents involving lavish overspending.
Uber, for example, revealed it blew through its entire annual AI budget in just four months, prompting the ride hailing firm to introduce spending caps for staff.
As reported in late June, Accenture also ordered staff to stop using AI for basic tasks such as converting PDFs to slides amidst rising costs.
It’s worth noting that a key factor behind these incidents lay in the tokenmaxxing trend, rather than general costs associated with AI. This is a practice of maximizing AI use within the enterprise by incentivizing staff to use the technology.
Some companies introduced leader boards to track usage rates, while others used it as a means to measure productivity.
In response to growing costs, some enterprises have begun reviving time-tested FinOps practices, as reported in late July, enabling them to better track costs and cut down on spending.
Cloudera’s report also highlighted a sharpened focus on this front. Nearly all enterprises believe they now have “some visibility” into AI spending, which marks a move in the right direction.
“Ninety percent say they can measure end-to-end AI costs at least somewhat accurately,” the report notes. Cloudera added that there’s still room for improvement here, however.
“Only 27% say they can do so very accurately, suggesting there is a considerable gap between having a general understanding of AI costs and a precise picture of where that spending goes.”