Companies are on track to spend $176.8 billion putting AI to work in 2026, up from $33 billion in 2025. That is the finding of Menlo Ventures' fourth annual report on AI in business. The surprise is the source. Only 22% of the new AI money came out of software budgets. More than half was brand-new budget.
What changed
Menlo surveyed about 400 U.S. enterprise decision-makers who together control more than $4 billion in AI spend. It asked where their extra AI dollars came from. 55% was net-new budget. 22% came from software budgets. The rest came from staffing and other lines.
AI spending now equals about 30% of public cloud revenue, up from about 6% a year ago. Menlo also estimates AI trimmed growth in business software from about 13% to about 11% on a market above $600 billion. That is roughly $12 billion of growth that did not happen.
Where the money went
- Coding tools: $34 billion, up 7.7 times in a year.
- Workplace assistants and chatbots: $25 billion. With coding, that is 76% of AI app spending.
- Healthcare AI: $3.7 billion. Legal AI: $1.8 billion.
- Sales AI: $1.4 billion, about 4% of the coding total.
Three companies took most of the spending. Anthropic, OpenAI and Google captured $122.7 billion, or 69%, up from 50% a year ago. Claude Code and Codex alone make up $24 billion, or 70% of coding spend.
Who it affects
Software sellers face a split. Where a company already owns the workflow and the data, such as ServiceNow for IT tickets or Workday for employee records, it captures the AI spending. Customer support went to newer AI firms like Sierra and Decagon, in the fastest-growing department at 4.3 times. Menlo found little sign that companies are replacing core systems like CRM, HR or ERP with AI.
For buyers, the budget is often people. The SaaStr author says its AI agents were paid for from headcount it did not backfill. It now runs with 3 humans and more than 20 agents, with higher revenue than before.
What to do
When judging an AI tool, ask what work it takes over, not how many seats it replaces. The author's rule for investing is that a product priced at 2 times its pre-AI rivals, and still bought, is likely tapping budgets beyond software.
Read with care
Menlo is an investor in many companies it covers, including Anthropic, Lovable and Factory. The budget and department figures come from its own survey. The agent-versus-human usage figures come from OpenRouter, whose users lean toward developers and startups.
The takeawayAI money is mostly coming from new budgets and staffing, so judge any AI tool by the work it takes over, not by the software line it replaces.