Gartner warns AI coding costs may top developer pay by 2028

Gartner said AI coding costs will exceed the average developer's salary by 2028, driven by rising token use and consumption-based pricing.

The forecast suggests a shift in how companies account for software development spending as AI coding agents move from limited trials into wider use. Many organisations are underestimating the budget impact of token consumption, particularly when suppliers charge by usage rather than by seat.

Tokens are the units of data processed by generative AI models and sit at the centre of the cost model for many AI coding tools. As developers rely more heavily on these systems to generate code, review software and handle development tasks, spending can rise quickly if usage is not closely managed.

That is creating a new budgeting problem for engineering leaders. Under older seat-based licensing models, companies could usually forecast spending with more certainty. Consumption-based charging introduces more variation because costs depend on the volume and type of work processed.

"Organisations are rapidly moving from experimentation to scaled deployment of AI coding agents, but many are underestimating the financial impact of rising token consumption," said Nitish Tyagi, Senior Principal Analyst, Gartner.

"Token discipline will not emerge through developer choice alone, as developers tend to optimise for speed and convenience over cost efficiency. Without a governed engineering operating model, costs can escalate faster than the productivity gains these tools are designed to deliver," Tyagi said.

Pricing shift

The market is moving away from simple per-user licensing towards models that tie charges more directly to the underlying computing consumed. That shift is making software engineering costs harder to predict because enterprises often have limited visibility into how vendors calculate token usage and which tasks drive spend.

In practice, a team may adopt an AI coding agent expecting productivity gains, only to find monthly costs vary sharply depending on prompt complexity, the amount of code and documentation fed into the model, and how often it is used across the wider engineering organisation. The result can be early budget depletion and a weaker link between spending and measurable business outcomes.

This lack of transparency is becoming a concern for software engineering leaders trying to justify AI expenditure. If organisations cannot link token consumption to development outcomes, it becomes harder to determine where AI coding tools add value and where they create avoidable costs.

"Most organisations still lack the maturity and frameworks to effectively measure cost versus business impact," Tyagi said.

"Software engineering leaders are increasingly concerned as token-driven AI spend becomes harder to justify, with budgets often being depleted earlier than expected," he added.

Governance gaps

Beyond the pricing model itself, internal usage patterns are adding to cost pressure. Gartner pointed to several common issues, including broad autonomy for agent-led workflows without clear controls, oversized context windows that increase token consumption, and a lack of structured feedback processes to refine how developers use the tools.

Those gaps matter because the economics of AI coding do not depend only on who has access to the software. They also depend on how tasks are designed, which model is selected, how much information is included in each prompt, and whether teams regularly review high-cost usage patterns.

Vendors have yet to provide mature built-in tools for cost optimisation in AI coding products, leaving customers to create their own controls, monitoring systems, and operating rules for their engineering teams.

"AI coding costs will continue to rise as infrastructure investment and profitability challenges push model pricing higher," Tyagi said.

"At the same time, as more developers adopt AI tools, light users are expected to rapidly become mainstream users as familiarity and reliance increase, driving further growth in token consumption and overall spend," he said.

Managing spend

To control costs, software engineering leaders should define clearer rules for when to use AI coding agents and how much autonomy to give them for different tasks. Development work should be classified into execution models that distinguish between tasks led by developers, tasks shared with agents and tasks handled largely by agents.

Organisations should also match model choice to task complexity, using smaller models for simpler, more frequent work and reserving larger models for more demanding development tasks. That approach can reduce unnecessary token use without removing AI support from engineering workflows.

Another recommendation is to train developers in context engineering so they provide only relevant information to AI systems, summarise content where possible and remove unnecessary data from prompts. Gartner also called for token thresholds, escalation policies and automated monitoring, alongside regular reviews of high-token workflows as part of normal development cycles.

AI coding is becoming a material line item in engineering budgets rather than a marginal software tool expense. Whether that spend remains manageable will depend less on enthusiasm for AI tools than on how tightly organisations govern their use and measure costs against business impact.