Are enterprises actually spending more on AI that works?

Andreessen Horowitz's 2025 survey of enterprise leaders (a follow-up to its prior-year report on "16 changes to the way enterprises approached building and buying gen AI") finds that gen AI spending has moved out of experimental budget lines and into the core IT and business-unit budget. In the firm's words, enterprise AI budgets "graduated from pilot programs and innovation funds to recurring line-items in core IT and business unit budgets." A year earlier, innovation budgets still accounted for a quarter of LLM spending; a16z reports this "has now dropped to just 7%." Leaders surveyed expect roughly 75% further budget growth over the next year, with one CIO quoted saying "what I spent in 2023 I now spend in a week."

a16z frames this as evidence that gen AI has stopped being treated as an experiment. The report attributes the shift to "the growing sentiment that gen AI is no longer experimental but essential to business operations," and quotes a CTO explaining the trajectory from the inside: "more of our products are adding AI enablement, so our spending growth will rise across all of these products." The reasoning is that as AI capability spreads into the product line rather than sitting in a standalone innovation fund, its budget inherits the growth pattern of the products themselves.

This sits in tension with Why do most enterprise AI pilots fail to deliver returns?: a16z's CIOs describe spending conviction and budget normalization at the same moment NANDA's case analysis finds the large majority of pilots produce no measurable P&L return. It parallels Why do individual AI gains not translate to business results?, where a similar gap shows up between individual-level enthusiasm and organization-level proof, and it corroborates the direction (if not the exact framing) of Are companies actually generating real value from AI investments?, which also finds enterprise AI maturing from novelty into mainstream spend over the same window.

The excerpt does not establish that this budget growth tracks any measured business return — only that the accounting category changed, from discretionary experiment to recurring line item. It is also worth naming the source directly: this is a venture firm's own survey of enterprise buyers, published by an investor with a commercial interest in signaling that the enterprise AI market has matured into a buying category startups can sell into. Taken at the strength the excerpt supports, the finding says enterprises have stopped budgeting for gen AI as a side bet, not that the bet is paying off.