THE ESSENTIALS IN 30 SECONDS

  • Menlo Ventures estimates enterprise spending on deploying AI reached $176.8 billion as of September 2026, up from $33 billion a year earlier.
  • Coding tools are the largest application category at $34 billion, and Menlo puts Anthropic's share of enterprise model API spending at 55%.
  • The figures come from Menlo's own model and a survey of nearly 400 US decision-makers; Menlo lists Anthropic and OpenRouter, whose data it uses, as portfolio companies.

Venture capital firm Menlo Ventures published its annual State of Enterprise AI report on November 3. Its headline estimate is that companies spent $176.8 billion deploying AI in 2026, measured as an annualised run rate in September, compared with $33 billion a year earlier. That is a 5.4-fold rise, after growth of 3.9 times in 2025. The report, written by Tim Tully, Joff Redfern, Deedy Das and Derek Xiao, attributes much of the increase to AI agents.

Where the money goes

Menlo splits the deployment total into $99.5 billion for infrastructure and $77.3 billion for applications, the latter up from $19 billion. Coding is the largest application category at $34 billion, of which the firm attributes $24 billion to the labs' own tools, Claude Code and Codex. Workplace chatbots and assistants account for $25 billion, industry-specific software for $10.1 billion (led by healthcare at $3.7 billion), and other horizontal tools such as customer support, sales and HR for $8.8 billion. Separately, Menlo estimates $102.5 billion was spent building models, mostly on training compute.

The report says Anthropic, OpenAI and Google together capture 69% of deployment spending. For enterprise model APIs, Menlo puts Anthropic at 55%, OpenAI at 31%, Google at 9% and open-source models at 4% of spending, although open models account for about 19% of tokens.

The role of agents

The report cites OpenRouter data indicating that agent-driven requests now produce about 80% of tokens on that platform, and that agent traffic has grown 37-fold since overtaking human use in February. Among builders surveyed, 74% use agents to run workflows. The authors acknowledge that OpenRouter's users lean towards individual developers and startups rather than large enterprises.

How to read the numbers

These are Menlo's estimates, not audited market data. They come from a bottom-up model informed by public filings, third-party reports and a Morning Consult survey of nearly 400 US enterprise decision-makers fielded in September. Menlo also restated earlier years to fit a new taxonomy, so comparisons with past editions may not line up.

Menlo is also an interested party. The report flags Anthropic, the market-share leader in its own ranking, as a portfolio company, and lists OpenRouter, Mercor and Fireworks among the companies it backs. The report offers no numerical forecasts; its authors expect spending to keep rising as agents handle longer tasks.