A small group of companies is spending heavily on artificial intelligence, offering a clearer look at how quickly AI costs can rise when businesses move from casual experimentation to aggressive daily use.
New data from Ramp’s AI Index shows that the top 1 percent of AI-spending firms are spending about $7,500 per employee each month on AI tools and services. Ramp describes these companies as “AI-pilled,” meaning they are not only testing AI, but building it deeply into their workflows, budgets, and operating habits.
The number is striking because it shows how uneven AI adoption has become. Some companies are spending at levels that rival major salary costs. Others are still paying almost nothing. According to the same data, the top 10 percent of firms spend about $611 per employee each month on AI, while the median company spends only about $11.38.
That gap tells a bigger story about the AI economy. The business world is not moving into AI at one uniform speed. Instead, a small group of aggressive adopters is racing ahead while most companies are still experimenting carefully, buying a few seats, or waiting to see whether the productivity gains are real.
The companies spending $7,500 per employee each month are likely not using AI the way ordinary firms use AI.
For many businesses, AI spending may mean a few ChatGPT, Claude, Gemini, Microsoft Copilot, or Perplexity seats. It may include writing help, meeting summaries, customer support experiments, code assistance, or document analysis. That kind of adoption is useful, but limited.
The highest-spending firms appear to be treating AI more like a core operating layer. They may be paying for multiple frontier models, developer tools, API usage, agent platforms, internal automation, AI search, coding tools, data infrastructure, and experimentation across teams.
That explains why the per-employee number can become so high. AI costs are not limited to monthly subscriptions. They can include token usage, model API calls, cloud inference, workflow automation, developer tooling, data processing, and custom AI systems built around company operations.
The spending figure also suggests that some companies are replacing or augmenting labor with AI at a serious scale. If a firm believes AI can reduce headcount growth, accelerate engineering, improve sales workflows, automate support, or generate more output from smaller teams, it may justify very high per-employee AI budgets.
Ramp’s data shows that AI spending among the heaviest adopters grew 14.1 percent per employee last month. That suggests the most committed companies are not pulling back yet, even as the broader market becomes more concerned about AI costs.
This matters because the AI industry is currently facing two competing narratives. One says companies are burning too much money on AI without clear returns. The other says businesses that adopt AI deeply may gain a major productivity advantage over slower competitors.
The Ramp data supports both sides. The median company spending only about $11.38 per employee suggests most businesses are still cautious. But the rapid growth among the top 1 percent suggests a small group has decided the opportunity is worth the cost.
That split may define the next phase of enterprise AI. Instead of a simple adoption curve where everyone gradually buys the same tools, the market may divide between AI-intensive firms and AI-light firms.
The winners will not necessarily be the biggest spenders. They will be the companies that can connect spending to measurable business gains.
The median monthly AI spend of $11.38 per employee is just as important as the $7,500 figure.
It shows that most companies are not yet living in the AI future described by vendors, investors, and consultants. Many businesses are still using AI lightly, if at all. Some may have bought a few subscriptions. Others may be testing tools in small departments. Many may still be blocked by security reviews, budget approvals, workflow uncertainty, or employee skepticism.
That low median also suggests there is still a large adoption runway. AI vendors may see it as proof that most businesses have not yet converted into serious paying customers. But it also shows that the market remains unproven at scale.
Companies do not spend heavily just because a technology is popular. They spend heavily when the tool becomes essential. For most firms, AI has not yet crossed that line.
That is why the gap between the median firm and the top 1 percent matters. The AI market is full of extreme early adopters, but mainstream business adoption is still much more modest.
The Ramp data arrives during a period of growing concern about AI spending.
Some companies are already putting limits on employee AI usage after bills rose faster than expected. Developers and business teams are also becoming more aware of token-based pricing, model costs, and the difference between using a cheap AI assistant and running heavy AI workflows all day.
The cost issue is especially important because AI is not like traditional software. A normal SaaS product may have high upfront development costs but relatively predictable per-user expenses. AI tools can have much higher variable costs because every model interaction consumes compute.
That makes spending harder to forecast. A team may start with simple subscriptions and then move into API-heavy workflows, coding agents, research tools, image generation, voice generation, customer support automation, or document processing. Each new use case can increase usage quickly.
For finance teams, this creates a new problem. AI budgets need to be monitored like cloud spending. Leaders need to know which teams are using which models, how much each workflow costs, and whether the output justifies the expense.
The $7,500 monthly figure sounds high, but the comparison becomes more complicated when set against employee costs.
TechCrunch notes that the figure is still below the roughly $16,000 monthly cost of an average software engineer. That comparison helps explain why some AI-heavy companies may tolerate high spending. If AI tools help a smaller team do the work of a larger one, the economics can make sense.
A company that spends heavily on coding agents, research automation, sales tools, or internal AI workflows may believe it is buying leverage. The question is whether that leverage is real.
AI can save time, but it can also create hidden costs. Employees may spend time checking outputs, fixing errors, rewriting drafts, debugging generated code, managing tool access, and maintaining AI workflows. If companies do not measure the full impact, they may mistake activity for productivity.
That is why the labor comparison is useful but incomplete. AI spend should not only be judged against salaries. It should be judged against business outcomes such as revenue growth, faster product cycles, lower support costs, higher sales conversion, reduced manual work, and better customer experience.
The highest-spending firms are not necessarily loyal to one AI provider. Ramp’s data suggests that top AI adopters often mix and match across multiple frontier models and platforms, including services that provide access to cheaper open-source models.
That behavior is becoming more common. Companies may use OpenAI for one set of tasks, Anthropic for another, Google for research or productivity workflows, open-source models for cheaper internal jobs, and specialized tools for coding, sales, customer support, or data analysis.
This reflects a more mature AI buying pattern. Instead of choosing one AI tool for the whole company, sophisticated buyers are matching tools to tasks. A high-end model may be used for complex reasoning or important customer-facing work, while cheaper models handle routine classification, drafting, extraction, or internal automation.
That kind of model routing could become central to AI cost control. The companies that spend the most may not be careless. Some may be building the internal systems needed to use AI efficiently across many workflows.
The spending gap creates both an opportunity and a challenge for AI vendors.
The opportunity is clear. The top 1 percent of firms are willing to spend heavily if they see value. That gives AI companies a path to large enterprise revenue, especially if they can serve power users with advanced models, integrations, security controls, and usage analytics.
The challenge is that the median company is still spending very little. That means vendors cannot assume every business will quickly become a high-value AI customer. Many firms may resist expensive plans, limit employee access, or choose cheaper models once they understand the costs.
This could push AI vendors toward more segmented pricing. Power users may pay for premium models, agent workflows, enterprise controls, and high-volume usage. Mainstream businesses may prefer lower-cost plans, bundled AI features, and strict limits. Developers may demand flexible APIs and model routing options.
The market may not settle into one dominant subscription model. It may become a layered market with different pricing for casual users, teams, developers, enterprises, and AI-native companies.
The biggest unresolved issue is whether heavy AI spending produces heavy returns.
The Ramp data shows how much companies are spending, but spending alone does not prove value. A firm spending $7,500 per employee each month may be building a major productivity advantage. It may also be wasting money on overlapping tools, uncontrolled experimentation, and inflated usage.
That is why measurement is becoming critical. Companies need to connect AI budgets to specific business outcomes. They should ask whether AI is reducing time-to-market, improving engineering velocity, lowering customer support costs, increasing sales efficiency, or enabling work that was previously impossible.
Without that measurement, AI spending can become another form of software sprawl. Teams sign up for tools, usage grows, invoices rise, and leaders struggle to identify what is actually working.
The companies that win with AI may not be the ones spending the most. They may be the ones that learn fastest which AI workflows deserve money and which should be cut.
Ramp’s data shows that enterprise AI adoption is becoming sharply unequal.
A small group of companies is spending at extreme levels and treating AI as a core part of operations. A larger group is spending meaningful but controlled amounts. The median company is still barely spending at all.
That divide will shape the next stage of the AI market. If AI-heavy firms show clear productivity gains, pressure will rise on the rest of the market to catch up. If their spending fails to produce measurable results, finance teams may become more skeptical and force a broader pullback.
For now, the data points to a market that is still early, uneven, and expensive at the top. AI is not yet a universal business operating system. But for the firms most committed to it, AI has already become a major monthly cost.
The next question is whether that cost becomes a competitive advantage or another line item companies eventually have to cut.
Share your thoughts about this article.
Be the first to post a comment!
