How much should you plan for your AI budget in 2027? Menlo Ventures' annual report, published on November 3, 2026, gives an answer that should alert every small business owner: surveyed US enterprises had planned to multiply their AI budget by 1.9x over the year, but ended up multiplying it by 4.7x on average. This overrun is not a simple fad effect. It comes from a change in how AI is used: autonomous agents, which run continuously without supervision, now consume most of the resources.
In brief
- Venture capital firm Menlo Ventures puts US enterprise AI spending at $176.8 billion in 2026, up from $33 billion in 2025, a 5.4x year-over-year increase.
- Surveyed enterprises had planned to multiply their budget by 1.9x, but multiplied it by 4.7x on average, a gap Menlo Ventures attributes largely to AI agents.
- Autonomous agents account for roughly 80% of tokens consumed on inference platforms like OpenRouter, up from a marginal share a year earlier.
- Three vendors, Anthropic, OpenAI and Google, capture 69% of enterprise AI deployment spending in 2026, up from 50% in 2025: Anthropic's share of enterprise API spend climbs from 40% to 55%, while Google's falls from 21% to 9%.
- For a small business, the lesson is not to avoid AI agents, but to budget for their usage differently from a standard software subscription, since their cost tracks the volume of work performed, not the number of employees.
What Menlo Ventures' 2026 report reveals
Menlo Ventures has published a benchmark study on enterprise AI adoption every year for the past four years. The 2026 edition relies on a survey conducted by Morning Consult between September 8 and 26, 2026, among nearly 400 decision-makers at US enterprises collectively responsible for more than $4 billion in AI spending. These responses are cross-checked against a market analysis built from public data and third-party reports.
The central finding: enterprise AI adoption didn't just grow, it changed scale. Adoption rose from 55% to 65% of surveyed enterprises, but average spend per enterprise nearly quintupled. In other words, it's not just that more enterprises use AI: those that already use it are consuming far more of it.
Why budgets overran: agents that never stop working
This is the most useful point for a small business preparing its next budget cycle. According to Menlo Ventures, the gap between planned and actual spend is largely explained by the rise of autonomous AI agents: systems that can execute multiple steps of a task without human intervention at each step, and run in parallel across many cases at once.
A standard software subscription costs roughly the same every month, regardless of how much employees use it. An AI agent, by contrast, typically bills by task or by volume of tokens processed: the more it works, the more it costs, and it can work around the clock without fatigue or a shift limit. The report notes that agent-generated traffic overtook human traffic on some inference platforms as early as February 2026, and then grew 37 times faster than human usage over the same period.
Key takeaway
An AI agent that works continuously cannot be budgeted like a fixed-fee piece of software. Its cost tracks the volume of work it completes, not the number of employees using it. Before deploying an agent into production, simulate several volume scenarios rather than simply rolling over last year's budget line.
Three vendors share most of the market
The report also documents a growing concentration around a handful of model vendors, a useful signal for any small business assessing its dependence on a single provider.
| Vendor | Share of API spend in 2025 | Share of API spend in 2026 |
|---|---|---|
| Anthropic | 40% | 55% |
| OpenAI | 27% | 31% |
| 21% | 9% |
Together, Anthropic, OpenAI and Google now capture 69% of enterprise AI deployment spending, up from 50% a year earlier. Open source models, meanwhile, account for only 4% of spending but 19% of tokens consumed, a gap explained by their generally lower inference cost.
This concentration is neither good nor bad in itself: it reflects a natural consolidation around the most reliable models in production. But it's a reminder that a small business building critical processes around a single model vendor is exposed to dependency risk, price increases, or changes to usage terms.
Where the money goes: code and conversational assistants dominate
Of the $176.8 billion, part corresponds to concrete applications rather than direct model access. The enterprise AI application market reaches $77.3 billion in 2026, up from $19 billion in 2025. Two use cases concentrate most of this spend:
- AI-assisted coding: $34 billion, up 7.7x, with a large share captured directly by the labs' own tools (Claude Code, Codex).
- Enterprise chatbots and workplace assistants: $25 billion, up 3x.
Together, these two use cases account for 76% of application spending, far ahead of categories like sales, marketing or HR, each still under $1 billion. For a small business wondering where to start, this ranking confirms what most field experience already shows: coding assistance and general-purpose conversational assistants remain the most mature and immediately profitable use cases.
What this means for a small business setting its 2027 AI budget
A small business obviously doesn't operate at the volumes of a large US enterprise. But the overrun mechanism applies to any organization moving from occasional AI use to an agent running in production.
Simulate rather than extrapolate
Separate the test budget from the production budget
Set consumption caps
Review the budget every quarter
FAQ
Why did enterprise AI budgets overshoot forecasts so much in 2026?
According to the Menlo Ventures report, the gap is mainly explained by the rise of autonomous AI agents, which bill by volume of work performed and can run continuously, unlike a fixed-price software subscription. Enterprises had planned to multiply their budget by 1.9x but multiplied it by 4.7x on average.
Should a small business be wary of AI agents because of this budget risk?
No, but it should budget for them differently. An AI agent doesn't have the same cost profile as traditional software: simulate its cost against a realistic task volume and set consumption caps before deploying it into production.
Which AI vendors dominate enterprise spending in 2026?
According to Menlo Ventures, Anthropic, OpenAI and Google together capture 69% of enterprise AI deployment spending in 2026, up from 50% in 2025. Anthropic's share of API spend climbs from 40% to 55%, while Google's falls from 21% to 9%.
Which AI use cases concentrate the most application spending?
Coding assistance ($34 billion) and enterprise chatbots or workplace assistants ($25 billion) together account for 76% of the $77.3 billion spent on AI applications in 2026, according to the Menlo Ventures report.
Conclusion
Menlo Ventures' 2026 report confirms a change in kind, not just in scale: enterprise AI is shifting from a predictable software cost to an execution cost that tracks agent workload. For a small business, planning a 2027 AI budget means simulating real usage scenarios, capping consumption, and reviewing the trajectory every quarter, rather than rolling over a fixed budget line. To go further on the real cost of AI agents, read our guide to the real cost of AI agents for small businesses and our build vs buy comparison for AI agents.
