Enterprise spending on artificial intelligence has reached a scale that is hard to ignore. Menlo Ventures' 2026 State of Enterprise AI report, published November 3, puts enterprise AI spending at $176.8 billion by September 2026, up about 5.4 times from a year earlier. The same report finds that three frontier labs now capture most of that money, and that AI agents are reshaping how usage grows.
A Fivefold Jump in One Year
The report's headline growth figure stands out against the 2025 baseline of roughly $33 billion. A jump of that size suggests that companies are moving AI from pilot projects into production budgets. Growth appears to come mainly from organizations that were already buying AI rather than from a flood of new customers.
That distinction matters. The share of enterprises paying for AI rose only modestly, from 55 percent to 65 percent. In other words, the larger headline number reflects existing adopters spending more, not a broad wave of first-time buyers. For vendors, that means the revenue opportunity lies in deepening relationships with customers who already run AI in their operations.
Frontier Labs Capture Most of the Spend
Menlo says Anthropic, OpenAI and Google together captured 69 percent of deployment spending in 2026, up from 50 percent in 2025. The shift is concentrated in the enterprise API market, where the company's share estimates show a clear reordering.
- Anthropic: rose from 40 percent to 55 percent of enterprise LLM API spending.
- OpenAI: rose from 27 percent to 31 percent.
- Google: fell from 21 percent to 9 percent.
Those numbers suggest that buyers are consolidating around a small number of providers, and that the companies with the strongest coding and agent products are gaining the most ground. Taken together, the three labs' share of deployment spending rose from half to roughly seven of every ten dollars in a single year, which leaves smaller model vendors and cloud resellers competing for a shrinking slice of the budget. Google's decline in this measure is the most striking, given its scale, and it will be a useful test of whether that slide continues in future reports.
Agents Drive the Usage Curve
The report also highlights how agents change consumption. According to figures cited alongside the report, agents accounted for about 80 percent of tokens on the OpenRouter platform, roughly seven times human-driven usage. Since agent traffic overtook human traffic in February, agent token consumption has grown 37 times, compared with five times for human-driven use.
The contrast is sharp. Agent token consumption has grown 37 times since agent traffic overtook human traffic in February, while human-driven usage grew five times. The gap helps explain why spending is rising faster than user counts. An agent can run many model calls to finish one task, so each deployed workflow consumes far more tokens than a single chat message. Budget planners who model AI costs on seats or prompts may find their forecasts underestimate usage once agents are running in production.
Why It Matters
The figures paint a picture of an enterprise AI market that has moved from experimentation to scaled operations, with money concentrating among a few suppliers. For chief financial officers and procurement teams, the message is that AI spending is becoming a core line item rather than a discretionary experiment. That brings a familiar set of questions about return on investment, vendor concentration and cost control.
There is also a caveat that readers should keep in mind. Menlo Ventures invests in both Anthropic and OpenRouter, so its data on those two companies should be read with that relationship in mind. The firm's methodology and sample are central to the numbers, and independent verification of the share estimates would strengthen them. Several secondary outlets restate the report's figures differently, so readers should check the original document before citing a specific breakdown.
Even with those caveats, the broad trend is hard to dismiss. Enterprises are paying more for AI each year, the biggest model providers are taking a larger share of that spending, and agents are turning usage into a volume business. The next question is whether return on investment keeps pace with the bill.
What to Watch
Three indicators will show whether the trend holds. The first is whether the share of enterprises paying for AI moves meaningfully past 65 percent, which would signal broader adoption. The second is whether Google's API share stabilizes or continues to fall. The third is how companies handle agent costs, since usage-based pricing can turn a successful pilot into an unexpectedly large invoice. Budget owners who track those indicators will be better placed to judge whether the current spending wave is a durable shift or a bubble still inflating.
