Fivetran has published research putting a number on the data clean-up behind AI projects. Its report, The AI tax, found that 39 percent of companies spend at least a quarter of their AI budget on fixing data quality issues, rebuilding integrations and reprocessing data.
The survey covered 500 senior IT, data and AI/ML leaders across the United States, Europe and Asia-Pacific, questioned between 24 August and 7 September 2026. Fivetran defines the AI tax as the self-reported share of AI spending consumed by avoidable work over the previous 12 months.
Delay tracks cost. Companies losing at least 25 percent of AI spending to avoidable work are five times as likely to have at least a quarter of their AI projects delayed or stalled. At the other end, 62 percent of companies losing less than 10 percent say AI returns beat expectations, against 40 percent of those losing half or more.
Regional differences are marked. In the US, 45 percent say at least a quarter of AI spend goes on avoidable work, compared with 34 percent in EMEA and 37 percent in Asia-Pacific. Organisations relying mainly on custom integrations fare worse than those on standardised platforms: 53 percent against 40 percent lose that share, and they are twice as likely to see data-related delays of more than a month.
Spending is not the constraint. Some 83 percent of respondents put at least $1 million into AI over the past year and 54 percent put in $5 million or more, yet only 37 percent have multiple AI initiatives running at scale. Asked where extra AI budget would go, 31 percent chose their data foundation, ahead of talent (17 percent) and AI models (16 percent).
“Companies are spending millions on AI, but too much of that investment is going toward fixing the same data problems over and over again,” said George Fraser, CEO and co-founder of Fivetran.
To stay across the latest in cloud, AI and enterprise tech analysis from Compare the Cloud, subscribe to our weekly newsletter at https://www.comparethecloud.net/newsletter
