Gartner says AI coding agents could cost more than software developers

AI coding agents are driving variable costs through token-based pricing

AI coding agents are becoming a standard part of software development. They help developers write code faster, review changes, generate tests, and work through complex technical problems. That creates real value. The challenge is that the business model behind these tools is changing much faster than many organizations expected.

Many leading AI coding platforms, including Claude Code, Cursor, and OpenAI Codex, are moving away from predictable seat-based subscriptions toward consumption-based pricing. Instead of paying a fixed monthly fee for each developer, companies increasingly pay based on the number of tokens consumed. Tokens are the units AI models use to process requests and generate responses. The more developers rely on AI, and the more complex their work becomes, the more tokens they consume.

This changes the economics of software development. A fixed software expense becomes an operational cost that can rise or fall every month. During periods of intensive development, legacy modernization, or large-scale code generation, spending can increase rapidly. That is a very different financial model from traditional software licensing.

For executives, this is not a reason to slow AI adoption. It is a reason to improve financial visibility. AI should be treated as a strategic resource with measurable business outcomes. Organizations that monitor token consumption alongside engineering productivity will be in a much stronger position than those that simply approve AI subscriptions and hope costs remain stable.

According to Gartner Peer Insights, 23% of technology leaders currently spend between $200 and $500 per developer each month on tokens for AI coding agents. While that level of investment may be justified by productivity improvements, it also shows that AI spending is becoming a meaningful component of software engineering budgets rather than a small software expense.

The organizations that benefit most will be those that understand where AI creates the highest value. Different engineering activities produce different returns. Routine coding assistance may require lower-cost AI models, while complex architecture, security analysis, or modernization projects may justify premium models with higher token consumption. Managing this balance is becoming a core leadership responsibility.

Limited transparency in token consumption complicates cost forecasting and management

The biggest challenge is not necessarily the price of AI. It is understanding what drives the price.

Gartner warns that many AI coding providers do not provide sufficient transparency into how token usage is calculated or billed. Without detailed reporting, organizations can see the invoice but struggle to understand why costs increased or which development activities generated the largest expenses.

That creates a management problem. Finance teams need predictable budgets. Engineering leaders need flexibility to innovate. Without clear usage data, both groups lose visibility. Budget forecasts become less reliable, and it becomes difficult to determine whether higher AI spending is producing proportional business value.

This also affects governance. Executives increasingly expect technology investments to demonstrate measurable returns. If an organization cannot connect token consumption to productivity improvements, software quality, faster delivery, or reduced technical debt, it becomes difficult to defend continued investment as AI adoption expands.

The solution is creating better operational discipline around it. Organizations should establish reporting that tracks token consumption by developer, team, project, and business outcome. Engineering managers should regularly review which AI models are being used, how frequently they are accessed, and whether premium models are necessary for every task.

This level of visibility also improves procurement decisions. As competition among AI providers increases, organizations with detailed usage data will be in a stronger position to negotiate contracts, compare platforms, and select pricing models that align with their engineering workloads.

For C-suite leaders, AI cost management should become part of broader technology governance rather than a standalone engineering issue. Companies that combine financial oversight with engineering metrics will be better positioned to scale AI adoption without allowing costs to grow faster than the business value those tools create.

Forecasts predict that by 2028, AI coding costs could surpass the average developer salary

AI is changing software development at a remarkable pace, but business leaders should pay close attention to how the cost model is evolving. Productivity gains are important, but they do not automatically guarantee lower operating costs. In many cases, higher productivity can encourage greater AI usage, which in turn increases consumption and spending.

Gartner forecasts that by 2028, the cost of AI coding could exceed the average salary of a software developer. The forecast is based on two major trends. First, developers are expected to rely more heavily on large language models throughout the software development lifecycle. Second, AI vendors are increasingly adopting consumption-based pricing, where costs rise with usage rather than remaining fixed.

This projection should not be interpreted as a prediction that developers will become less valuable. Human expertise remains essential for designing systems, making architectural decisions, validating AI-generated code, managing security, and aligning software with business objectives. AI expands what developers can accomplish, but it also introduces a new operating expense that organizations must actively manage.

For executives, this changes how software engineering budgets should be evaluated. Instead of comparing AI costs with software licensing budgets alone, organizations may need to compare them with labor costs, outsourcing expenses, and overall engineering productivity. AI is becoming part of the core cost structure of software development.

Business leaders should also recognize that higher AI spending is not necessarily a negative outcome. If AI enables faster product delivery, shorter development cycles, reduced maintenance effort, or improved software quality, higher operating costs may still produce stronger business results. The key question is whether AI creates enough measurable value to justify continued increases in spending.

According to Gartner, by 2028 AI coding costs are expected to overtake the average developer’s salary because of increasing large language model token consumption and the continued shift toward consumption-based licensing models.

High AI token costs require clear business justification

As organizations increase their reliance on AI coding tools, the discussion should move beyond cost reduction. The focus should be on value creation. Every significant AI expense should be connected to a business outcome that executives can measure and understand.

Nitish Tyagi, Senior Principal Analyst at Gartner, makes this distinction clear. He supports continued AI adoption despite rising costs. As he stated, “I’m a big believer that AI is bringing gains. You should not move away from AI because the total costs are increasing. But I believe that token costs will certainly increase.”

He also warns that organizations should expect substantial price increases as vendors expand consumption-based pricing. Monthly AI costs could rise from around $20 per developer to $200, or even $2,000 in some situations. Gartner reports that 6% of organizations already spend more than $2,000 per developer each month on AI token usage.

These figures can appear alarming without context. The more important question is why those costs exist. High AI spending may be entirely justified when developers are working on projects that generate significant strategic value. Examples include modernizing legacy systems, improving security, reducing technical debt, or accelerating the delivery of products that directly support business growth.

Tyagi highlighted an example from an Indian IT organization where one developer generated approximately $20,000 in AI token costs. Rather than immediately reducing usage, the leadership team investigated the situation. They found that the developer was working on a legacy modernization project where the AI investment produced sufficient value to justify the unusually high level of consumption.

This demonstrates an important leadership principle. Cost alone is not an effective performance metric. Executives should evaluate AI investments using business outcomes such as delivery speed, engineering efficiency, software quality, customer impact, and long-term operational improvements. Organizations that apply this level of discipline will be better positioned to scale AI responsibly while maintaining confidence in their technology investments.

Evaluating software engineering maturity is essential for optimizing AI tool usage

AI coding agents can deliver significant productivity gains, but simply making them available to developers is not enough. Organizations achieve the best results when AI adoption follows a structured plan rather than expanding without clear direction. Before increasing investment, leaders should understand whether their software engineering practices are ready to support more autonomous AI capabilities.

Gartner recommends that software engineering leaders begin by assessing the maturity of their existing development processes. This means evaluating how software is planned, built, tested, reviewed, deployed, and maintained. AI produces the strongest results when these underlying processes are well defined. If development practices are inconsistent or poorly documented, AI may increase activity without delivering proportional improvements in quality or efficiency.

The next step is to understand how developers are already using AI coding agents. Different teams often use AI in different ways. Some rely on it for writing code, while others use it for debugging, documentation, testing, code reviews, or understanding large and complex codebases. Without visibility into these usage patterns, organizations cannot determine where AI delivers the highest return or where spending can be optimized.

This assessment should also distinguish between work that genuinely requires advanced AI models and tasks that can be completed effectively with less expensive alternatives. Premium models often provide stronger reasoning capabilities and better performance on complex software engineering problems, but they are not necessary for every development activity. Matching the AI model to the complexity of the task helps control token consumption without reducing productivity.

For C-suite leaders, this is ultimately a governance issue rather than simply a technology decision. AI investments should be aligned with broader business objectives, supported by clear performance metrics, and reviewed regularly as usage evolves. Organizations should establish policies for AI adoption, monitor costs alongside engineering outcomes, and encourage teams to share successful use cases that can be scaled across the business.

This disciplined approach also supports long-term competitiveness. Companies that understand where AI creates measurable business value will be able to expand adoption with greater confidence while maintaining financial control. Rather than viewing AI as a universal solution, they can deploy it where it has the greatest strategic impact and continuously refine their approach as both the technology and pricing models evolve.

Gartner outlines these recommendations in its report, “How to Optimize Token Consumption for AI Coding,” which advises software engineering leaders to assess organizational readiness, identify current AI usage across development teams, and distinguish between use cases that require autonomous development or premium AI models and those where lower-cost options are sufficient.

Key takeaways for decision-makers

• Treat AI coding as a variable operating cost: AI coding agents are shifting to token-based pricing, making software development costs less predictable. Leaders should monitor AI usage alongside productivity to ensure spending scales with business value rather than adoption alone.
• Build visibility into AI consumption: Limited transparency around token usage makes budgeting and ROI difficult. Executives should require detailed usage reporting and governance to forecast costs accurately and identify where AI delivers the strongest returns.
• Prepare for AI to reshape engineering budgets: Gartner forecasts that AI coding costs could exceed the average developer salary by 2028. Organizations should evaluate AI spending as part of their overall engineering investment, balancing higher operating costs against measurable gains in speed, quality, and innovation.
• Justify premium AI spending with measurable outcomes: Rising token costs are not inherently a problem if they support high-value work. Leaders should assess expensive AI usage against clear business metrics such as faster delivery, legacy modernization, improved software quality, and long-term cost reduction.
• Align AI adoption with engineering maturity: Expanding AI without disciplined processes can increase costs without improving outcomes. Leaders should assess development maturity, match AI models to the right use cases, and continuously optimize usage to maximize value while maintaining financial control.