Software used to be simple to budget. You counted your staff, multiplied by a seat price, and knew what you owed. AI has ended that logic, and the shift is happening right now, not in some distant future.
Anthropic, OpenAI, and GitHub have all moved parts of their services away from flat-rate subscriptions toward usage-based billing. Microsoft joined them with its new premium E7 license, which bundles AI tools and charges accordingly. These are not isolated experiments: they reflect a coordinated commercial reality. Running large AI systems is expensive, and vendors are no longer willing to absorb that cost on behalf of their customers.
Forrester's 2027 Budget Planning Survey, drawn from more than 2,600 business and technology decision-makers globally, found that 80 percent expect data and software budgets to rise. The reason cited is direct: vendors are raising prices or adding usage charges to pass their AI infrastructure costs to customers. After a year of cautious spending, Forrester also notes that more than 80 percent of leaders expect their overall budgets to increase, with up to one in four expecting growth of 10 percent or more.
The scale of what is being financed is worth understanding. Consultants at Bain estimated that building out AI infrastructure globally would cost $2 trillion by 2030. Amazon Web Services alone is planning capital expenditure of around $200 billion this year, largely for AI capacity, up roughly 50 percent on the year before. Microsoft's total capital spend is expected to reach $190 billion, up 61 percent. Those numbers have to come from somewhere, and the answer is increasingly: from you, via your software bills.
The shift to usage-based billing creates a specific problem for organizations that have no experience managing this kind of variable cost. A KPMG survey of 2,145 senior leaders across 20 countries found that 29 percent struggle to understand their operating costs as they scale AI. A third also identified limited understanding of AI costs and economics as a barrier to deploying AI agents. Uber's technology team reportedly burned through its entire 2026 AI budget in four months. These are not isolated incidents: multiple companies are discovering that AI costs scale with use in ways that traditional software costs never did.
Anthropics's pricing restructure illustrates how the change works in practice. The company has moved enterprise customers from fixed fees to usage-based charges, structured so that every interaction with the AI is metered and billed. Heavy use under the old flat-rate model was effectively subsidized by the vendor: companies processing millions of AI requests per month paid the same as those running a handful. That subsidy is gone. As one analysis put it, flat rates were a user acquisition strategy, and metered billing is the mature commercial model.
Staffing costs are not falling to compensate, either. Forrester found that personnel costs account for 35 percent of IT budgets, and 67 percent of technology decision-makers expect to increase their staffing budget for 2027. The idea that AI spending would be offset by headcount reductions has not materialized at the organizational level. Roles in data and analytics are growing, not shrinking.
Forrester's practical recommendation is to treat AI costs the way cloud computing costs are managed: with real-time monitoring, usage caps, and someone in the organization who is specifically accountable for controlling spend. The organizations that figure out cost discipline early will have a structural advantage. Those that leave AI usage ungoverned will face unpleasant surprises at the end of every billing cycle.
Forrester's chief research officer put it plainly: the organizations that outperform in 2027 will not be those that spend the most on AI. They will be the ones that build the foundations that make AI effective, including good data, governance, and the ability to adapt. Spending more without those foundations does not accelerate results. It accelerates waste.