AI Spend Benchmarks: How Much Are Companies Spending on AI in 2026?

How Much Are Companies Spending on AI in 2026?

AI spending has moved from experimental line items to one of the fastest-growing categories in enterprise software budgets. According to Gartner, global AI spending reached approximately $1.5 trillion in 2025 and is on track to top $2 trillion in 2026. IDC reports that enterprises alone invested $307 billion on AI solutions in 2025, with year-over-year growth of nearly 32%.

But those macro numbers obscure the question that actually matters to finance leaders at scaling companies: how much should we be spending on AI tools?

Gartner and IDC publish these global forecasts, but they are macro-level and largely paywalled. No one is translating them into practical, actionable benchmarks for mid-market companies. That gap is what this piece aims to fill.

To answer that question, we analysed anonymised spending data from companies on the Cledara platform (predominantly tech-forward businesses between 30 and 500 employees) and combined it with the latest industry research from Zylo, Gartner, IDC, and others. The result is a practical set of AI spend benchmarks designed for mid-market finance and IT leaders heading into budget season, board meetings, or renewal negotiations.

The headline number: 79% of companies on the Cledara platform are now paying for at least one AI tool, and the average AI adopter spends roughly $15,300 per year across an average of 3.7 AI subscriptions. That figure is growing fast, so understanding where your spend sits relative to peers has never been more important.

AI Spend Benchmarks by Company Size

AI spending varies significantly by company size, stage, and the degree to which teams have embedded AI into daily workflows. A 20-person startup experimenting with ChatGPT has a fundamentally different cost profile from a 300-person company running Cursor across its engineering team and Anthropic API calls in its product. Based on Cledara platform data and industry benchmarks, here is what mid-market companies typically spend on AI tools each month.

On the Cledara platform, the median AI adopter spends roughly $330 per month on AI tools, while the average is closer to $1,280 per month. That gap tells an important story: a small number of companies with heavy API usage or large engineering teams pull the average significantly higher. The 25th percentile sits at around $76 per month, while the 75th percentile reaches approximately $1,100 per month.

These ranges align with broader industry data. A 2026 report from Zylo found that organisations spent an average of $1.2 million on AI-native applications, though that figure includes larger enterprises. For mid-market companies in the 30 to 500 employee range, a more realistic annual AI budget falls between $6,000 and $150,000, with the median closer to $40,000 to $50,000.

Per-employee AI spend ranges from $25 to $100 per month depending on industry and the maturity of AI adoption. Companies in software development and technology tend to sit at the higher end because of AI coding tools like Cursor and Windsurf, which are now standard parts of the engineering toolkit. Professional services and marketing-heavy companies tend to skew lower, with spend concentrated on general-purpose AI assistants and content tools.

AI Spend by Category: Where the Money Goes

Not all AI spend is equal. Understanding which categories dominate your AI budget helps identify where costs are concentrated and where consolidation opportunities exist. Based on Cledara platform data, here is how AI spend breaks down across the five major categories.

General-Purpose AI Dominates the Budget

The most striking finding is that general-purpose AI platforms (OpenAI, Anthropic/Claude, Perplexity, Mistral) account for roughly 72% of total AI spend on the Cledara platform. This concentration makes sense: these tools serve every department, from engineering and product to marketing and customer success. They are also the tools most likely to have usage-based pricing, which means costs can fluctuate significantly from month to month.

On the Cledara platform, OpenAI is the single largest AI vendor by average spend per company (roughly $7,700 per year), followed closely by Cursor (approximately $5,900 per year) and Anthropic (about $6,200 per year including both API usage and Claude Pro/Team subscriptions). Together, OpenAI and Anthropic represent the vast majority of general-purpose AI costs for most organisations.

This concentration also creates negotiation leverage. If 72% of your AI budget flows to two or three vendors, enterprise agreements or committed-use discounts with those vendors can produce meaningful savings.

AI Coding Tools: The Fastest-Growing Category

AI coding tools represent approximately 18% of AI spend but are the fastest-growing category by adoption. Cursor alone is now used by a significant share of companies on the Cledara platform, with average annual spend per company exceeding $5,900. Windsurf (average spend of around $1,400 per year per company) and CodeRabbit (roughly $1,900 per year per company) are also growing rapidly as engineering teams embed AI assistants directly into their development workflows.

For companies with large engineering teams, AI coding tools can easily become the second-largest AI cost category after general-purpose platforms. A team of 20 developers using Cursor Pro at $20 per seat per month adds $4,800 per year before accounting for usage-based overages. Add CodeRabbit for automated code review and Windsurf for a second IDE option, and the coding tool line item can reach $10,000 or more annually.

The business case for these tools is typically strong (developer productivity gains of 20 to 40% are commonly reported), but the proliferation of options creates a governance challenge. Many engineering teams adopt tools organically, leading to duplicate subscriptions across teams.

The Long Tail of AI Content, Creative, and Vertical Tools

AI content and creative tools (Midjourney, Leonardo.ai, KLING AI) represent only about 1% of total AI spend, but they are among the most widely adopted in terms of the number of distinct tools purchased. On the Cledara platform, AI content and creative tools are used by roughly 25% of companies, despite their small share of the total budget. This creates a governance challenge: many small subscriptions across different teams can add up, and these tools are often purchased on individual credit cards outside of standard procurement processes.

Vertical AI tools (AI for sales, customer support, data analysis, and infrastructure) are also a growing category. Tools like AssemblyAI for transcription, Retell AI for voice agents, and Langsmith for LLM observability are becoming standard for companies building AI-powered products. Expect this category to grow significantly as companies move from using AI as a productivity tool to embedding it in their products and services.

AI Spend Trends to Watch in 2026

AI Spend Is Growing Dramatically Faster Than Overall SaaS

According to Zylo's 2026 SaaS Management Index, spending on AI-native applications grew 108% year-over-year, while the broader AI tool category saw 181% growth. Compare that to overall SaaS market growth of roughly 19% annually. AI is not just another line item in the software budget; it is reshaping the entire spend profile.

On the Cledara platform, AI tools now represent approximately 5.6% of total SaaS spend on average across all companies. For heavy adopters (75th percentile and above), that figure climbs to over 10% of total software costs. Finance leaders who are not tracking AI spend as a distinct category in their reporting risk losing visibility into one of their fastest-moving cost centres.

For context on the broader SaaS picture, see our analysis of average SaaS spend per company, which provides the baseline against which AI spending should be measured.

Usage-Based Pricing Makes AI Costs Less Predictable

Unlike traditional SaaS subscriptions with fixed per-seat pricing, many AI tools charge based on consumption: API calls, tokens processed, or compute hours used. A 2025 industry survey found that 78% of IT leaders reported unexpected charges due to consumption-based or AI pricing models, up from 65% the previous year. This unpredictability makes budgeting and forecasting significantly harder for finance teams accustomed to predictable monthly SaaS invoices.

OpenAI and Anthropic (API usage), in particular, can produce wildly variable monthly bills depending on how development and product teams use them. One month might cost $200; the next could hit $5,000 if a new feature ships that calls the API at scale. On the Cledara platform, the gap between the smallest and largest single transaction for OpenAI is enormous, ranging from under $1 to over $60,000 in a single payment.

For practical guidance on managing these negotiations, see our guide to negotiating AI and SaaS agreements.

Shadow AI Adds 20 to 30% to Your Known AI Costs

Shadow AI, where employees sign up for AI tools using personal accounts or unapproved purchasing channels, is one of the biggest blind spots in enterprise AI budgets. Research shows that 59% of employees use AI tools that have not been formally approved by their company, and 93% of executives and senior managers admit to using shadow AI at work.

The financial impact goes beyond subscription costs. Data breaches caused by shadow AI cost an average of $670,000 more than breaches from sanctioned AI tools, according to recent research published in Fortune. And with AI governance spending projected to reach $492 million globally in 2026 (Gartner), the cost of not governing AI is increasingly quantifiable.

For a deeper dive into how shadow AI creates hidden costs and security risks, see our guide to managing AI tool sprawl.

Vendor Consolidation Is Beginning

While the average AI adopter on the Cledara platform uses 3.7 AI tools, the most aggressive adopters use up to 16 distinct AI subscriptions. This level of tool sprawl is unsustainable. Expect 2026 to bring consolidation as companies standardise on fewer platforms, negotiate enterprise agreements, and retire overlapping tools.

Companies that consolidate from, say, five separate AI coding tools to one or two enterprise agreements can typically reduce per-seat costs by 20 to 30%. The same principle applies to general-purpose AI: rather than paying for OpenAI, Claude, Perplexity, and Mistral subscriptions across different teams, standardising on one or two platforms and negotiating volume pricing is a natural cost optimisation play.

Are You Overspending on AI? Key Warning Signs

AI spend is not inherently good or bad. The companies generating the highest ROI from AI often spend more than average, not less. What matters is whether the spend is intentional, governed, and delivering value. Here are the warning signs that your AI budget may be out of control.

• AI spend is growing faster than headcount. If your AI costs are climbing 50%+ year-over-year but your team size is flat, you may be paying for unused capacity or uncontrolled API usage. On the Cledara platform, the average AI adopter allocates roughly 8% of total SaaS spend to AI tools. If your figure is significantly above that without a clear strategic rationale, it is worth investigating.
• Multiple subscriptions to similar tools across teams. It is common to find engineering using Cursor, product using Windsurf, and marketing using a third AI coding assistant. Consolidating onto a single platform almost always reduces costs and simplifies governance.
• Low utilisation on paid plans. If fewer than 60% of licences on a paid AI tool are actively used each month, you are likely overpaying. This is especially common with tools purchased at the team level that individual employees never adopt.
• No approval process for new AI tool purchases. If anyone can expense an AI subscription on a corporate card without review, shadow AI spend is almost certainly inflating your costs by 20 to 30% beyond what finance can see.
• No visibility into usage-based charges. If you cannot see a daily or weekly breakdown of API costs from providers like OpenAI or Anthropic, you are flying blind on what could be your largest AI cost category.

For a framework on measuring whether your AI investment is paying off, see our CFO guide to measuring AI ROI.

How Cledara Benchmarks and Controls AI Spend

Tracking AI spend across dozens of tools, pricing models, and teams is a challenge that spreadsheets cannot solve at scale. Cledara provides purpose-built capabilities for managing AI costs alongside the rest of your SaaS portfolio.

Benchmarks: Know If You Are Overpaying

Cledara's Benchmarks feature (part of the Spend Optimisation module) shows how your price for every AI tool compares to the 25th and 75th percentile across Cledara's customer base. If you are paying $30 per seat for an AI coding tool while the median company pays $20, that is an immediate negotiation lever for your next renewal. Benchmarks turn pricing opacity into a competitive advantage.

AI Dashboard: Daily Spend Visibility

Cledara's AI Dashboard connects to AI providers (OpenAI, Anthropic, Cursor) via API keys and visualises daily spend by provider. Instead of waiting for a surprise invoice at month-end, finance teams can see usage-based costs accumulate in real time and set alerts when spending exceeds budgeted thresholds. For companies with significant API-based AI costs, this visibility alone can prevent thousands of dollars in unexpected charges.

Engage: Discover Shadow AI

The Engage browser extension deploys across your organisation's browsers and detects AI tools that employees are using but that are not captured in finance systems. This is how companies discover the 20 to 30% of AI spend that does not show up on any corporate card or invoice. Engage is privacy-first: it only tracks SaaS provider URLs, never general browsing history.

Virtual Cards and Approval Flows: Control AI Purchasing

Every AI subscription purchased through Cledara gets its own virtual card with a configurable spend limit. Approval workflows ensure that new AI tool purchases are reviewed before they are activated, whether that requires a single manager sign-off or dual IT and finance approval above certain thresholds. And if you need to cancel an AI subscription, freezing the card is instant: no vendor runaround, no waiting for the billing cycle to end.

Forecasting: Project Future AI Costs

Cledara's Budgeted Spend feature projects future AI costs based on historical payment data, helping finance leaders build accurate budget forecasts for board presentations and annual planning. When AI spend is growing at 100%+ year-over-year, having a data-driven forecast is the difference between a credible budget and a guess.

Want to benchmark your AI spend against companies like yours? Cledara's Benchmarks show you exactly where you stand. Book a demo to see your AI spend data in context.

Methodology

The Cledara platform data referenced in this post is based on anonymised, aggregated spending data from companies using the Cledara platform. The dataset includes companies predominantly in the 30 to 500 employee range across technology, financial services, and professional services sectors. All figures are presented as per-company averages, medians, or percentages to protect individual company confidentiality. AI vendors were categorised into five groups (general-purpose AI, AI coding tools, AI writing and productivity, AI analytics and vertical tools, and AI content and creative) based on their primary function.

External data sources include Gartner (September 2025 AI spending forecast), IDC (2025 AI investment report), and Zylo (2026 SaaS Management Index). Industry survey data on shadow AI and consumption-based pricing is sourced from published reports by ISACA, the Journal of Accountancy, Fortune, and Torii. All external figures are cited inline and reflect the most recent available data as of March 2026.