Companies spent $177 billion on AI this year, over five times last year

Businesses spent about $177 billion on AI this year, up from $33 billion last year. That is more than five times as much, according to a new report from the venture capital firm Menlo Ventures. The firm says the jump is driven by AI agents, which are tools that can carry out tasks on their own instead of just answering questions.

What changed

The report puts enterprise AI spending at $176.8 billion, up 5.4 times in a year. Companies expected to spend more and budgeted nearly twice what they spent in 2025. Menlo's survey found they actually spent 2.5 times their budgets. Menlo also says that for every dollar spent training AI models, companies spent $1.70 putting AI to work.

Why agents change the bill

Menlo's Derek Xiao says agents change how software is paid for. In the past, costs tracked the number of employees using a tool. Now, he says, spending is limited by how much work the AI can get done. More work done can mean a bigger bill, even if headcount stays the same.

Who is winning

The makers of the most advanced AI models are taking most of the money. Menlo says these labs capture nearly seven of every ten dollars spent, up from half last year. Anthropic grew its share of enterprise spending on API access, the paid connection that lets software use a model, from 40% to 55%. Google fell from 21% to 9%.

The picture differs by type of software. In coding tools, AI-native companies hold 96% of the market. In areas like IT and HR, established vendors still have the edge.

What to keep in mind

• Menlo is an investor in Anthropic and OpenRouter, so it has a stake in this market.
• The total is a Menlo estimate, built from a model, public filings and third-party reports.
• The survey covered nearly 400 U.S. enterprise decision-makers, mostly large companies.

What to do

Smaller businesses were not the focus of the survey, so the numbers may not match your costs. Still, the lesson is useful. AI spending can grow faster than planned when tools do more work. Before signing up, check what each tool does, how it is priced, and whether the price rises with usage.