Enterprise AI spending is shifting from licenses to API consumption.

ChainCatcher report: FundaAI’s enterprise AI applications research indicates that enterprise AI budgets are still expanding, but trajectories are diverging in the second half of 2026 and into 2027. AI spending guidance from large U.S. telecom operator A shows an increase from a January baseline of 100 to approximately 190 by December, with an expected additional 40%–50% year-over-year growth in 2027. In contrast, large European automaker A has seen only a 10%–15% increase year-to-date, with next year’s guidance largely flat. Incremental spending is shifting from licensed seats toward API/token consumption and production workflows. For the aforementioned telecom operator, the split between subscriptions and APIs has changed from roughly 50%/50% to 40%/60%, and may evolve further toward 35%/65%; large mid-sized biopharma company A has adjusted its ratio from 80%/20% to approximately 70%/30%. Open-source adoption is uneven, with active use cases accounting for 30%–40% of usage but contributing a smaller share of spending due to lower unit costs; experts estimate open-source inference is approximately 40%–70% cheaper than proprietary frontier models, narrowing to a 20%–40% gap when accounting for total costs, with further potential savings of 20%–30% in API expenditures through model routing, caching, and context compression. Production-side AI budgets are increasingly being built bottom-up based on workflow ROI; this telecom operator’s typical production ROI ranges from 1.5x to 2x, with payback periods of 6–18 months, rising to 3x–5x for mature use cases. The next wave of spending will relate to agents, software modernization, network operations, commoditized workflows, and longer-cycle business processes—but engineering capacity, process reengineering, governance, and data readiness are becoming tighter constraints than funding.