Menlo Ventures puts enterprise spending on deploying AI at an annualized $176.8 billion as of September 2026. The rounded headline figure of $177 billion expresses the pace of spending on a yearly basis. To understand what happened to an individual company’s budget, look at a different set of numbers in the report.
What goes into the $177 billion?
Menlo allocates $99.5 billion to infrastructure for running models and agents and $77.3 billion to applications enterprises use. Running a model for company workloads belongs in the first group. An AI application employees use belongs in the second. The report counts model development separately. On its revised classification, Menlo puts the 2025 deployment figure at $33 billion. That gives it an estimated 5.4-fold increase in the annualized market figure.
Why did the average rise more than the median?
Survey respondents reported spending an average of 4.7 times their 2025 level, against a plan for 1.9 times. At the median, the middle value when you rank the results, spending reached 2.2 times the prior level against a plan for 1.8 times. Both exceeded plan. Menlo attributes the wide gap between the two measures to a small group of particularly heavy spenders. For a company drawing up its own budget, the average helps show how far the market can move. The median gives a better sense of where the middle respondent landed.
How did Menlo arrive at its estimate?
Morning Consult surveyed nearly 400 US enterprise decision-makers from September 8 to 26, 2026. Menlo built its market estimate from survey responses, public filings and third-party reports. It also reclassified earlier years to fit the new spending categories. Menlo invests in Anthropic and other AI companies. Its position as an investor and the US survey sample both matter when you read its interpretation of future demand.
What changes when agents do the work?
Menlo argues that agents can keep working after a person moves on and run jobs in parallel, so usage can grow faster than headcount. I’d budget for per-user fees and workload-based charges separately. An overnight workflow that processes records can cost more when the volume rises even if the number of people with access stays the same. The report gives you a reason to track that workload. Your own volume and usage data are what let you price it.
