AI Spending Stalls as Usage Rises: What Ramp Data Means for Anthropic’s IPO
TL;DR
- Ramp data indicates AI usage continues to spread even as aggregate spending growth has flattened.
- Falling model prices, procurement scrutiny, and tighter budgets may be separating AI adoption from revenue growth.
- For Anthropic, the trend could complicate the investor case for uninterrupted hypergrowth ahead of any potential IPO.
WHAT RAMP’S DATA APPEARS TO SHOW
AI adoption is still expanding across businesses, but corporate spending on AI services is no longer accelerating at the same pace.
That is the central tension in recent Ramp data: more companies appear to be using AI, and existing customers may be using it more frequently, yet the amount they spend has stabilized. The divergence suggests that AI demand remains real, while the economics of supplying that demand are changing quickly.
The pattern is important because early artificial-intelligence investment narratives often treated usage growth, revenue growth, and spending growth as essentially interchangeable. They are not. A company can process more prompts, deploy more AI-powered features, or add more employees to an AI tool while spending less because model prices have fallen or because customers have negotiated more aggressively.
That distinction could become increasingly significant for the large AI companies whose valuations depend on years of rapid expansion.
USAGE IS RISING, BUT THE BILL MAY NOT BE
Several forces can push AI usage higher without producing equivalent growth in customer spending.
First, model prices have declined sharply in many parts of the market. Providers have introduced cheaper models, reduced inference costs, and offered lower-priced access for high-volume workloads. Customers can also route simpler tasks to less expensive systems and reserve premium models for complex reasoning or sensitive work.
Second, AI software is becoming embedded in products that employees already use. A business may increase its use of AI features without signing a large new contract or materially expanding its software budget. In some cases, AI is becoming a standard capability rather than a separately approved purchase.
Third, companies are becoming more disciplined about experimentation. During the first wave of generative AI adoption, businesses often approved pilot programs with limited scrutiny. As those pilots mature, finance teams are asking which tools deliver measurable productivity gains, which overlap with existing software, and which should be consolidated.
The result is a market in which activity can rise while the average price paid for each unit of activity falls.
PRICE CUTS ARE CHANGING THE GROWTH EQUATION
Lower prices are generally good for customers and can accelerate adoption. They are more complicated for vendors.
A cheaper model can attract more users, increase the number of tasks performed, and make AI viable in applications that were previously too expensive. But providers must generate enough additional volume to offset the reduction in revenue per request. If usage rises more slowly than prices fall, topline growth can weaken even while the product becomes more popular.
This is especially relevant for foundation-model companies. Their financial performance depends not only on how many customers they have, but also on the mix of models those customers use, the complexity of their workloads, the amount of inference each application generates, and the share of revenue retained after computing costs.
A company may therefore report impressive usage metrics while facing pressure on average revenue per customer. That does not mean demand is weak. It means the market is moving from a scarcity model, in which access to advanced models commanded a premium, toward a more competitive market where customers have greater leverage.
WHY COST CONTROLS MATTER
Corporate buyers are also becoming more sophisticated.
Many enterprises now compare AI providers, negotiate volume discounts, impose usage limits, and require internal approval for new tools. Some are building systems that use several models at once, allowing them to shift workloads according to cost, speed, reliability, or data-security requirements.
Those practices can reduce spending growth even when AI becomes more deeply integrated into daily operations.
The same dynamic is visible in broader technology procurement. Businesses may continue investing in AI while cutting redundant software, delaying hiring, or requiring new projects to demonstrate a measurable return. AI is not necessarily being removed from budgets; it is being forced to compete more directly for budget space.
That distinction matters for interpreting Ramp’s findings. A spending plateau should not automatically be read as a collapse in demand. It may instead indicate that AI is moving from an experimental category into a managed operating expense, where buyers expect predictable pricing and clear business outcomes.
THE IMPLICATIONS FOR ANTHROPIC
Anthropic has built its profile around advanced models, enterprise adoption, and partnerships that give customers access to its Claude family of systems. Those strengths could position the company well as businesses move from experimentation to production deployments.
But an IPO would expose Anthropic to a higher level of scrutiny than it faces as a private company. Public-market investors would likely examine revenue growth, customer concentration, gross margins, cash consumption, infrastructure commitments, and the relationship between usage and monetization.
Ramp’s data raises a difficult question for that potential offering: Is rising AI usage translating into durable economic value for model providers, or is competition forcing more of that value toward customers?
Anthropic could argue that lower prices are an investment in market expansion. Cheaper access may encourage developers to build more applications, increase enterprise adoption, and create larger long-term workloads. If those workloads become deeply embedded in business processes, volume could eventually compensate for lower prices.
Investors, however, may want evidence that this conversion is happening at scale. They may ask whether customers are expanding contracts, whether usage is concentrated among a small number of large accounts, and whether the company can improve margins as model efficiency advances.
A potential IPO would therefore need to present more than raw growth. It would need to explain the quality and durability of that growth.
THE HYPERGROWTH NARRATIVE FACES A TEST
The AI market has been valued partly on the expectation that demand will continue compounding with few interruptions. Ramp’s data does not disprove that thesis, but it introduces a more nuanced scenario.
Usage can continue climbing while revenue growth becomes less explosive. Customers can adopt AI broadly while reducing the amount they pay per task. Model providers can gain market share while absorbing substantial costs for chips, data centers, energy, and research.
That combination could produce a market with strong technological progress but less predictable financial performance.
For investors, the key issue is not simply whether companies are using AI. It is whether AI vendors can capture enough of the economic value generated by that usage. If productivity gains accrue mainly to customers while model prices decline, enterprise adoption may increase without supporting the revenue multiples that investors once expected.
Anthropic could face particular pressure because the company operates in an expensive part of the AI stack. Training and serving advanced models require substantial capital, and the race to improve performance can demand continued spending even when prices are falling.
A successful public-market story would likely require confidence that scale, efficiency improvements, premium products, and enterprise contracts can eventually produce attractive margins.
WHAT INVESTORS MAY WATCH NEXT
Several indicators could help determine whether the current spending slowdown is temporary or structural.
One is the pace of enterprise contract expansion. If customers are using AI more but not increasing commitments, providers may be seeing volume growth without equivalent revenue growth. If contracts expand as deployments move into production, the current plateau could prove to be a transition period.
Another is the mix of workloads. Routine summarization and basic content generation are more vulnerable to price competition than specialized coding, research, compliance, and workflow automation. High-value applications may support stronger pricing even as commodity inference becomes cheaper.
Gross margins will also matter. Lower computing costs could help providers retain more revenue over time, but those gains may be offset by larger models, higher usage, and continuing investment in infrastructure.
Finally, investors will examine customer concentration. Large strategic partnerships can accelerate revenue, but reliance on a small number of major customers can make growth more volatile and increase buyers’ negotiating power.
A MORE COMPLICATED AI MARKET
The emerging picture is not one of AI losing momentum. It is a market becoming more economically mature.
Early adoption was driven by curiosity, urgency, and the fear of falling behind. The next stage will be shaped by procurement teams, measurable returns, model substitution, and pricing pressure. Companies will continue using AI, but they will expect more output for every dollar.
That transition could benefit the strongest providers by expanding the overall market and eliminating weaker competitors. It could also expose business models that depend on premium pricing or assumptions of uninterrupted spending growth.
For Anthropic, the potential IPO question is therefore larger than whether Claude usage is rising. The company will need to demonstrate that growing usage can become durable, diversified, and profitable revenue in an environment where customers have more choices and models cost less to access.
Ramp’s data offers an early warning for investors: AI adoption may remain one of the strongest technology trends in the market, while the financial returns from that adoption become harder to predict.
The companies that succeed will not necessarily be those with the fastest usage growth alone. They will be the ones that can turn expanding activity into lasting customer value, pricing power, and sustainable margins.
Get All The Latest Updates Delivered Straight To Your Inbox For Free!