Menlo estimates enterprise AI spending hit $176.8 billion as agents drove demand

Enterprise AI’s spending surge is coming more from heavier use than from a rush of new buyers. Menlo Ventures’ report, published November 3, estimates deployment spending reached $176.8 billion by September 2026, up from $33 billion in 2025. The share of enterprises using AI rose much less sharply, from 55% to 65%.

The 2026 State of Enterprise AI report puts spending at 5.4 times last year’s level. Menlo attributes the acceleration to agents: AI systems that can carry out multi-step tasks with limited human oversight. These systems, the firm argues, let companies send more work to AI—not merely expand the number of employees using it.

The report draws on a survey of roughly 400 US enterprise decision-makers, according to Yahoo Finance. Its spending figures are Menlo market estimates, not an audited tally of company purchases. Menlo is also an investor in Anthropic, one of the report’s biggest estimated beneficiaries.

Heavy users pull spending beyond the plan

Companies planned to spend an average of 1.9 times their 2025 AI budgets. Their reported spending instead reached 4.7 times last year’s level—roughly 2.5 times the plan. The median company showed a smaller gap, spending 2.2 times its prior-year amount against an expected 1.8 times.

That difference matters: the average does not describe a typical buyer. Menlo says a small group of power users accounts for a disproportionate share of spending and grew much faster than most enterprises. The firm’s explanation is that agents can work independently, run in parallel and continue after people step away. Consumption can therefore grow without a matching rise in headcount.

Menlo points to OpenRouter, a service connecting developers with hundreds of models, as evidence of the shift. Its analysis puts agents at 80% of the platform’s tokens, the units used to measure model input and output. That is a view of one distribution channel, not a measurement of all enterprise AI activity.

New money, not just redirected software budgets

New budgets funded 55% of incremental AI spending, rather than money shifted from existing software or labor budgets, Menlo reports. Another 22% came from software budgets, most of it staying with incumbent vendors converting conventional software licenses into AI offerings.

Menlo found limited evidence that enterprises were replacing core customer-management, human-resources or financial systems with AI. It describes AI as enlarging the software market while modestly slowing traditional software growth, a conclusion about budget flows and purchasing patterns—not proof that every AI purchase earns a return.

Three labs take a larger share

Anthropic, OpenAI and Google

50%→69%

% of enterprise AI deployment spending

Menlo estimates the three labs’ combined share rose from 50% in 2025 to 69% in 2026, equivalent to $122.7 billion.

Capability captures dollars; cheaper models capture usage

Within enterprise spending on model APIs—the connections developers use to access models—Menlo estimates Anthropic’s share rose from 40% to 55%. OpenAI increased from 27% to 31%, while Google fell from 21% to 9%. Menlo interprets those shifts as buyers favoring capability over price.

Usage tells a different story from revenue. Open-source models’ token share rose from 11% to 19%, while their spending share fell from 6% to 4%. Menlo attributes the divergence to a 3.3-fold drop in average open-source token prices. Growing use, in other words, does not necessarily translate into a larger share of dollars.

Application spending is also concentrated. Two categories account for 76% of Menlo’s estimated $77.3 billion enterprise AI application market:

• Coding products: $34 billion.
• Chatbots and workplace assistants: $25 billion.