AI Usage Climbs but Spending Stalls: A Chilling Signal Ahead of Anthropic's IPO

Introduction: The Divergence Between AI Spending and Usage

New data from Ramp, the corporate spend management platform, reveals a curious contradiction: AI usage continues to climb, but corporate spending on AI has stalled. This trend arrives at a sensitive moment as Anthropic approaches its IPO, potentially shaking investors' assumptions about the industry's "hypergrowth."

According to Ramp's analysis, although the frequency of AI tool calls and the number of active users keep growing, total corporate spending on AI services has not risen in tandem. This "more usage, less spending" phenomenon is driven primarily by two forces: a sharp drop in model prices, and increasingly refined cost control among top enterprise customers.

Drivers of the Spending Stagnation

Price Wars and Falling Model Costs

Fierce competition in the AI model market has directly driven down the cost per unit of usage. Vendors are slashing prices to win enterprise customers, so the same usage volume now translates into a smaller bill. For enterprises, this is undoubtedly a boon—but for AI companies that rely on API revenue growth, it means a shift in the growth logic: from "usage-driven" to "value-driven."

Unit economics is a key framework for assessing the quality of growth at AI companies, measuring revenue and cost at the granularity of a single customer, a single API call, or a single token. When model prices fall rapidly, usage growth may fail to translate into revenue growth, because the margin per unit of usage gets compressed. Investors who look only at total call volume can easily overestimate growth; by breaking down unit economics, they can spot the real gap between revenue growth and cost decline. For API-centric companies like Anthropic, per-token pricing, inference costs, and customer retention together determine long-term profitability. The "more usage, less spending" pattern revealed by Ramp's data is precisely a signal of deteriorating unit economics: usage rises while revenue per unit falls, meaning the company must rely on ever-larger volumes to sustain the same revenue—and the quality of growth weakens accordingly.

Cost Control Among Large Customers

Ramp's data specifically notes that the largest enterprise customers are actively implementing smarter cost-control measures. These customers are no longer blindly expanding their AI budgets; instead, they are cutting expenses by optimizing call strategies and choosing more cost-effective model combinations. This shift in behavior reflects enterprise AI procurement moving from an early "experimental investment" phase into a phase of "refined operations."

Potential Impact on Anthropic's IPO

The timing of Anthropic's IPO window, coinciding with the release of this data, has prompted the market to re-examine its valuation logic. Investors typically treat growth in AI usage as a leading indicator of industry health, but Ramp's data shows that the positive correlation between usage and spending is weakening.

If this trend continues, AI companies' revenue growth may fall short of market expectations—especially for those whose core revenue comes from API call volume. As a leading model provider, Anthropic's IPO pricing will inevitably be affected by investors' worries about "peak AI spending."

The "growth narrative" in IPO pricing typically relies on verifiable leading indicators such as active users, call volume, or revenue growth. When usage decouples from spending, underwriters and institutional investors will demand a higher risk premium or lower valuation multiples. For Anthropic, an API-revenue-centric model provider, the revenue structure, customer concentration, and gross margins disclosed in its prospectus will directly determine how the market prices concerns about "peak AI spending." If large customers continue to optimize costs, revenue growth may fall short of roadshow expectations, in turn affecting the offering price range. Moreover, market sentiment during the IPO window has an anchoring effect on valuations of comparable AI companies; Anthropic's pricing performance could become the reference benchmark for subsequent AI listings.

Investors Should Recalibrate Expectations

Ramp's analysis points to a key question: does the narrative of AI hypergrowth still hold? Usage climbing while spending stalls means the industry's monetization efficiency is declining. For investors, tracking usage metrics alone is no longer sufficient to judge AI companies' growth prospects; more attention must be paid to unit economics and the quality of customer retention.

This data does not negate AI's long-term value; rather, it signals to the market that the industry's growth curve may be flatter than expected, and that price competition and cost optimization will become the norm. Anthropic's IPO performance may become the first litmus test of how well the market accepts a new paradigm for AI growth.

Monetization efficiency measures an AI company's ability to extract revenue from usage, with common metrics including average revenue per active user, revenue per million tokens, and gross margin trends. Usage climbing while spending stalls means monetization efficiency is declining: the same call volume generates less revenue, and companies must rely on greater scale to sustain growth. For investors, this means tracking usage alone is no longer sufficient to judge growth prospects; assessments must combine customer retention, net revenue retention (NRR), and per-unit inference costs. As price competition and cost optimization become the norm, the AI industry's growth curve may be flatter than early expectations, and valuation frameworks will need to shift from "usage multiples" to "profit multiples."

One-line Summary

Ramp data shows AI usage climbing while enterprise spending stalls, potentially undermining the valuation logic behind Anthropic's IPO.

Paragraph Summary

New data from Ramp reveals a notable contradiction: AI usage continues to climb, but total enterprise spending on AI has stalled. This divergence is driven mainly by two factors: price wars among models that sharply lower the cost per unit of usage, and top enterprise customers actively implementing refined cost control. The article notes that this trend arrives at a sensitive moment as Anthropic approaches its IPO, potentially shaking investors' assumptions about the industry's "hypergrowth." If the positive correlation between usage and spending keeps weakening, revenue growth at AI companies whose core revenue comes from API call volume may fall short of expectations. Investors need to shift from merely tracking usage to focusing on unit economics and the quality of customer retention—and Anthropic's IPO performance will be an important litmus test of how well the market accepts a new paradigm for AI growth.