AI coding tokens could soon cost as much as salaries
What's the story
The cost of artificial intelligence (AI) coding tokens could soon match or even surpass the monthly salary of a software engineer. This prediction comes from research firm Gartner, which says that within two years, enterprises will be spending as much on AI token usage as they do on their developers' salaries. The trend is driven by the growing adoption of generative AI and agentic tools among developers.
Licensing changes
Shift toward consumption-based licensing models
The shift toward consumption-based licensing models is also a major factor in the rising costs of AI token usage. Vendors are now balancing their infrastructure investments with profitability, leading to a change from the traditional flat per-seat Software as a Service (SaaS) model. Now, enterprises are paying for developer token use as well, further adding to their costs.
Cost implications
Global average salary of $2,000 per month
Gartner's prediction is based on a global average salary of $2,000 per month. This doesn't mean that AI token usage will exceed all salaries, especially in countries like the US, where annual pay rates can reach six digits or more. However, Gartner senior principal analyst Nitish Tyagi says he has heard of cases where businesses have incurred huge expenses due to high AI token consumption by developers and business users alike.
Cost underestimation
Cost optimization capabilities not provided yet by AI coding vendors
Many enterprises are moving from experimentation to scaled deployment of AI coding agents, but they still underestimate the token costs. This is mainly because software engineering workloads have "highly variable" cost structures, and there isn't much transparency into how token consumption is calculated and billed. Tyagi also pointed out that AI coding vendors haven't provided mature cost optimization capabilities yet, leading to rising prices as vendors build their models while trying to remain profitable.
Cost control challenges
Lack of frameworks to determine ROI from AI technologies
Enterprises are finding it difficult to forecast and control costs due to the rapid growth of AI. Many organizations lack the "maturity and frameworks" needed to determine ROI from these technologies. Tyagi noted that agent-driven workflows are hard to govern, context windows get bloated, budgets get wiped out earlier than expected, and token spend becomes difficult to justify.
Increased consumption
Non-developers becoming more familiar with AI tools
As non-developers become more familiar with and reliant on AI tools, their usage will increase. This will further drive up token consumption and spending. Despite the high costs of token consumption, Tyagi emphasized that there is no "direct relationship" between the number of tokens developers consume and their productivity gains. Instead, applying context engineering principles to optimize or reduce token consumption can improve quality.
Cost optimization
How to optimize token consumption?
Tyagi stressed that organizations shouldn't abandon AI coding agents, but rather optimize token consumption to avoid overspending while maintaining the quality and value brought by AI. Gartner also advises establishing strong governance and cost controls, such as introducing token thresholds, automating usage monitoring, and creating explicit escalation policies. Embedding these controls into engineering workflows ensures consistency and prevents uncontrolled cost growth.