Software vendors shift AI costs to customers via usage charges

Forrester, a major technology research firm, surveyed over 2,600 business and technology decision-makers and found that software and AI vendors are reshaping how they charge customers to offset rising infrastructure costs. In the last six months, Anthropic, OpenAI, GitHub, and Microsoft have shifted from flat-rate subscription models to usage-based billing, where customers pay based on consumption—typically measured in tokens processed by AI models. Microsoft's move includes launching the premium E7 license, which bundles M365 Copilot, Agent 365, and security tools with the base E5 offering, forcing customers to pay more for AI capabilities they may not have budgeted for.

Forrester's findings underscore the magnitude of the underlying financial pressure. Bain & Company estimated last year that the build cost for AI datacenters alone will hit $2 trillion（約320兆円） by 2030. Rather than absorb these costs, vendors are passing them to customers. Forrester found that 80 percent of business and technology decision-makers expect data and software budgets to rise in 2027 specifically because vendors are increasing prices or adding usage charges. Sharyn Leaver, Forrester's chief research officer, cautioned that "the organizations that outperform in 2027 won't be those that spend the most on AI. They'll be the ones that invest in the foundations that make AI effective: trusted data, strong governance, organizational readiness, and the ability to continuously adapt as technology and customer behavior evolve."

The shift to usage-based pricing has exposed a critical gap in organizational readiness. Traditional financial operations (FinOps) practices were designed for cloud infrastructure with predictable, recurring costs, not for token-based, consumption-driven AI expenses. Forrester recommended that organizations fund runtime cost controls—including model routing (directing queries to cheaper models when appropriate), semantic caching (reusing results for similar queries), and usage guardrails (hard limits on spending)—to prevent costs from spiraling. A separate study by KPMG in July found that nearly a third of corporate leaders reported difficulty understanding and controlling operating costs when implementing business AI at scale, suggesting that most organizations have not yet built these capabilities.

Personnel costs add another dimension of budget pressure. Despite recent high-profile layoffs at Oracle, Microsoft, and Meta, Forrester found that IT staffing spend has not declined in recent years. Staffing accounts for 35 percent of IT budgets in 2025. Looking ahead to 2027, 67 percent of tech decision-makers expect to increase their staffing budget, while 23 percent said it would stay flat and 10 percent expected it to decline. Forrester noted that staffing for data and analytics-specific roles is expected to rise particularly sharply, with 68 percent of data technology decision-makers expecting this budget to increase. The report concluded that the "AI washing of layoffs"—where vendors announce job cuts for financial restructuring while continuing to hire in other areas—will obscure the reality that AI implementation is driving, not reducing, headcount needs.

The shift to usage-based billing represents a fundamental change in how AI infrastructure costs are distributed. Bain & Company estimated last year that the build cost for AI datacenters would hit $2 trillion（約320兆円） by 2030, creating massive financial pressure on vendors. Rather than absorb these costs themselves, major players like Anthropic, OpenAI, GitHub, and Microsoft have moved to pass them directly to customers through token-based usage charges. This transition occurred over the last six months and has begun triggering cost concerns among users.

Forrester's survey reveals that organizations are unprepared for this shift. Eighty percent of business and technology decision-makers expect software and data budgets to rise in 2027, but most lack the tools and processes to forecast and control usage-based spending. A parallel KPMG study found that nearly a third of corporate leaders already struggle to understand and control operating costs when implementing business AI at scale. The disconnect suggests that many organizations will face sticker shock as bills arrive—and will be forced to invest in FinOps capabilities (financial operations for cloud/AI spending) that traditional IT budget management cannot provide.

Staffing budgets present an additional layer of cost pressure. Despite high-profile layoffs at Oracle, Microsoft, and Meta, Forrester found that IT staffing spend has not declined in recent years and accounts for 35 percent of IT budgets in 2025. For 2027, 67 percent of tech decision-makers expect to increase their staffing budget, suggesting that the "AI washing of layoffs"—cost-cutting announcements that mask other hiring—is masking continued or rising headcount expenses.