Forrester Warns AI Usage Charges Will Lift Software Budgets
The Register cited Forrester research saying software and AI vendors are adding price increases and usage charges, while 80 percent of surveyed decision-makers expect higher data and software budgets.
Eighty percent of surveyed technology decision-makers expect data and software budgets to rise as vendors pass AI costs to customers through higher prices and usage charges.
The Register cited Forrester research covering more than 2,600 respondents and services from Anthropic, OpenAI, GitHub and Microsoft.
Forrester identified services from Anthropic, OpenAI and GitHub that moved away from flat-rate subscriptions during the previous six months.
The research firm also listed Microsoft after a recent premium licence launch.
Bain & Company estimated last year that AI data-centre build costs would reach $2 trillion by 2030.
In Forrester's survey, 80 percent of decision-makers expected data and software budgets to increase as AI costs moved through vendor pricing.
Staffing Budgets Have Not Fallen With AI Rollouts
Forrester also found that personnel costs have not yet fallen despite layoffs in the technology sector.
The report said IT staffing spend has not declined in recent years, even as Oracle, Microsoft and Meta announced significant layoffs.
The Forrester figures cited by The Register put staffing at 35 percent of IT budgets in 2025.
For 2027, the research firm said 68 percent of data technology decision-makers expected staffing budgets to rise.
FinOps Teams Face Token-Based AI Spending
FinOps practices should be adapted for unpredictable AI costs, according to Forrester.
Its report described traditional FinOps as unbuilt for token-based and usage-driven AI costs, but argued that those teams were best placed to build new controls.
The research firm named model routing, semantic caching and usage guardrails as runtime cost controls that could limit runaway spending.
KPMG research found in July that nearly a third of corporate leaders had difficulty understanding and controlling operating costs when implementing business AI at scale.
The cited material did not include measured results showing how much those controls reduced token-based AI spending.
