AI Usage Is Rising, but Enterprise Spending Is Stalling

Companies are expanding their use of AI, but their bills are not rising at the same pace. New data from Ramp shows that enterprise AI token spending grew only 1.5% across August and September, a sharp contrast with the 42% month-over-month increase recorded in July. Spending picked up again in October, rising 8.9% from September 23 to October 23, but the rebound has not erased concerns about a broader slowdown.

Key takeaways

• Usage and revenue are separating. Token volumes continue to rise, yet lower prices and more efficient consumption are limiting the amount companies pay.
• The heaviest users have paused. Ramp’s top 1% of AI-intensive companies dipped slightly below their summer peak of nearly $10,000 per employee per month. Because this group accounts for a large share of spending, even a modest pullback can weigh on total revenue growth.
• Customers have more bargaining power. OpenAI and Anthropic have cut prices and offered enterprise discounts, while Meta, Nvidia, and other companies are bringing lower-cost or open models to market.
• AI buyers are becoming more sophisticated. Companies are setting budgets, restricting expensive models for long-running tasks, and routing simple requests to cheaper systems.

Why it matters

Ramp’s figures do not show that AI demand is disappearing. They show that the cost of satisfying that demand is falling. Median companies spend roughly $12 to $15 per employee per month on AI, while the top 10% spend about $600 to $1,000. That gap suggests the most aggressive spending seen in software and finance may not be representative of doctors’ offices, car dealerships, or other less AI-intensive sectors.

The distinction matters for model companies. Investors have often treated token growth as a close proxy for revenue growth. But when prices are falling, more calls do not automatically produce more sales. According to TechCrunch, Ramp lead economist Ara Kharazian attributes the pressure to intense competition and customers learning how to control their usage, rather than to a collapse in demand.

The timing is particularly important for Anthropic as it approaches a potential IPO. The report says the company’s implied valuation expectations fell from about $2 trillion in early October to roughly $1.2 trillion to $1.3 trillion, while OpenAI also lowered its proposed valuation and fundraising target. Those changes cannot be attributed solely to Ramp’s data; financing conditions and broader concerns about profitability also play a role. Still, slowing monetization is likely to be part of investors’ analysis.

The industry’s next opportunity may be a new category capable of generating spending on the scale of AI coding tools. Personal agents are frequently presented as the candidate, but whether they will produce comparable economic value remains uncertain. The clearer trend today is that businesses can use more AI while spending less per task. That is good news for customers, but it means model companies will need new use cases, stronger retention, or better ways to capture value if they want to sustain hypergrowth.

Source: TechCrunch AI