AI Coding is Creating a New Enterprise Cost Challenge

AI coding assistants have become one of the fastest-growing applications of generative AI, promising faster software delivery, improved developer productivity, and shorter development cycles.

A new Gartner prediction suggests the economics of those tools are becoming an important consideration for enterprise technology leaders.

The analyst firm forecasts that by 2028, the cost of AI coding tools could exceed the average software developer’s salary, driven by rising consumption of large language model (LLM) tokens and a shift towards usage-based pricing.

AI Coding is Moving Beyond Software Licensing

For many organisations, AI coding has progressed well beyond experimentation. Development teams are using AI assistants to generate code, explain applications, write documentation, review changes and support testing.

As adoption expands across engineering teams, the commercial model is beginning to change. According to Gartner, AI coding is moving away from predictable seat-based licensing towards consumption-based pricing, where costs increasingly reflect how much AI is used rather than how many developers have access to it.

That introduces a different way of thinking about AI investment.

Instead of behaving like a conventional software subscription, AI coding begins to resemble cloud infrastructure, where consumption directly influences operational cost.

Token Consumption is Becoming an Operational Cost

At the centre of Gartner’s forecast is token consumption.

Tokens are the units of text that large language models process when reading prompts and generating responses. Every coding request consumes tokens, and more complex tasks generally require more of them.

As developers work with larger context windows, generate greater volumes of code and make increasing use of AI agents capable of completing multi-step tasks, token consumption rises accordingly.

According to Gartner, many organisations currently have only limited visibility into how that usage translates into operational cost.

Nitish Tyagi, Senior Principal Analyst at Gartner, said:

“Organisations are rapidly moving from experimentation to scaled deployment of AI coding agents, but many are underestimating the financial impact of rising token consumption.”

He added:

“Without a governed engineering operating model, costs can escalate faster than the productivity gains these tools are designed to deliver.”

The discussion, therefore, extends beyond software procurement. Understanding how AI is consumed becomes part of managing enterprise technology costs.

Productivity and Consumption Are Not Always Aligned

One of the more interesting aspects of Gartner’s research concerns developer behaviour. Software engineers naturally optimise for solving problems quickly.

They are far less likely to consider token efficiency, model selection or consumption costs while writing code. That creates a practical management challenge.

Without clear operational guidance, organisations may see AI usage increase while finding it more difficult to understand which activities create value and which simply generate additional consumption.

The issue is less about limiting adoption than understanding how AI resources are used across engineering teams.

AI FinOps is Beginning to Emerge

Gartner argues that organisations should address this through operational management rather than restricting AI adoption.

The firm identifies several common drivers of unnecessary expenditure, including ungoverned agent autonomy, oversized context windows, limited monitoring, weak feedback mechanisms and the absence of intelligent model routing.

Its recommendations focus on introducing greater visibility into AI consumption through practices such as model selection, token thresholds, continuous monitoring and regular reviews of high-cost workflows.

Taken together, these approaches resemble the operational disciplines that emerged around cloud computing as organisations sought to manage increasingly variable infrastructure costs.

As AI becomes embedded within software engineering, similar approaches are beginning to appear around AI consumption.

Enterprise AI is Becoming an Economics Discussion

The findings reinforce a broader development across enterprise AI.

The first phase of adoption focused on capability.

The next phase increasingly centres on economics.

As AI becomes part of everyday software development, organisations are balancing productivity gains alongside infrastructure costs, operational visibility, and long-term financial sustainability.

The Gartner forecast suggests that managing AI consumption may become as important as selecting the right models.

For engineering leaders, the discussion is extending beyond how AI improves software development to how those improvements are measured, monitored, and sustained as usage continues to grow.