AI Is Getting Expensive, and the Executives Paying for It Are Starting to Notice
A major survey finds nearly a third of senior business leaders cannot keep track of what AI is costing them. A new forecast says AI coding tools could soon cost more per developer than the developer's own salary.
Key points
- A KPMG survey of more than 2,000 senior executives across 20 countries found that 29 percent struggled to understand the running costs of their AI deployments.
- Nearly half of those executives said they were looking to slow down or reshape their AI projects when costs outweigh the benefits.
- Gartner analyst Nitish Tyagi forecasts that by 2028, the per-developer cost of AI coding assistants will exceed the global average developer salary.
- In countries with lower wages, such as India, that crossover is already happening today.
- Major AI providers including Anthropic, OpenAI, and GitHub have shifted from flat monthly fees to usage-based billing, where companies pay per "token," meaning each small chunk of text the AI reads or writes.
Businesses rushed to adopt AI tools over the past two years. Now the invoices are arriving, and the numbers are causing genuine alarm in boardrooms.
A new survey by KPMG, the global consulting and professional services firm, captured the mood. The firm polled more than 2,000 senior executives across 20 countries and found that 29 percent of them could not get a clear handle on what their AI deployments actually cost to run as they grew. Reported first by The Register AI, the findings paint a picture of companies that jumped in fast and are now scrambling to understand what they signed up for.
The confusion has a simple cause. For a long time, AI tools came with flat monthly subscriptions. You paid a fixed fee, your team used the tools as much as they wanted, and the bill was predictable. That model is disappearing. Anthropic, OpenAI, and GitHub have all moved to usage-based billing built around tokens. A token is a small unit of text, roughly three quarters of a word, and every time an AI model reads a prompt or writes a reply, it burns through tokens. The more your team uses the tools, the higher the bill, and costs can spike without warning.
Nearly half of the executives surveyed said they were considering "re-phasing" their deployments, which means slowing down, cutting scope, and looking for cheaper or smaller models rather than always reaching for the most capable (and most expensive) option.
What does this mean for workers whose companies use AI tools?
It may actually protect some jobs. Gartner's Nitish Tyagi found that in parts of the world where developer salaries are lower, the cost of running an AI coding assistant already exceeds what a human developer earns. Globally, Tyagi projects that crossover will arrive by 2028. When a tool costs more than the person it was meant to replace, the financial case for replacing that person collapses.
Gartner's research also found something counterintuitive about how these tools work. Using more tokens does not automatically produce better code. Developers who were disciplined and precise in how they wrote their prompts got higher-quality results and spent less money than those who used the AI freely and often.
The wider financial picture is hard to ignore. One major investment analysis found that the AI industry is on track to spend 1.5 trillion dollars on data centre infrastructure between now and 2030. That spending has to be paid back somehow, which is exactly why pricing is shifting toward usage-based models that can grow with each customer.
For companies already dependent on these tools, switching is not simple. Developers who have used AI assistants daily for a year or two can lose fluency with the underlying skills. That dependency gives AI providers real pricing power. Charge too much, though, and enterprise customers will look elsewhere: cheaper open-source models, smaller specialist tools, or simply hiring back the staff they let go.
