AI coding agents could soon cost more than the developers using them

[usonian]

https://www.theregister.com/ai-and-ml/2026/06/24/ai-coding-agents-could-soon-cost-more-than-the-developers-using-them/5260864

As a result of the lack of cost optimization tools among vendors and consumption-reduction strategies among users, AI development is in a situation where a developer's coding agents may cost more than they earn, at least in some parts of the world.

Gartner predicts that by 2028, AI coding costs will overtake the average developer's salary due to rising LLM token consumption and the shift to consumption-based licensing models.

Since the main AI coding agent vendors have shifted from seat-based licensing to consumption-based pricing, developer teams now face highly variable cost structures.


Developer teams face an emerging problem of escalating costs, said Nitish Tyagi, senior principal analyst at Gartner. AI coding bills were leaping from $20 or $100 to $2,000 to $5,000 per developer per month, while in extreme cases, the bill might hit $20,000 in token charges. .

About that "free lunch" ......

[Klarkashton]

1. And the code is shit too.

[exboyfil]

2. I vaguely understand this

My SIL is a software engineer, and he was discussing it with me. He said his productivity has improved dramatically since adopting this approach (I was "retired" two years ago from the same company). What I previously had done was more ad hoc depending upon the specific circumstances was becoming standardized. I really couldn't keep up in hindsight. My current job is a jack of all trades that I don't see being standardized and brought under AI anytime soon (or hopefully not in the next five years).

I do like telling AI to write a specific set of Excel commands and it does it quickly for me.

[usonian]

4. A problem as I see it, is that when you offer and suport code, it has to be very well documented.

AI tends to reinvent the wheel every time you ask it something, and currently, there are no solid libraries of thoroughly tested code that you can call. I recall some math packages that are incredibly well tested.

So, a customer calls, desperately asking for a fix, and you have zero clue as to what the code is, except by reverse engineering.

And software is one of the few areas with useful results so far.

Useful, to a point.

[SWBTATTReg]

3. I have to laugh at this! This is so funny, that it's pathetic and goes to show that this new 'tech', vouched as a

'wonder drug' or the likes of, has turned out to be a white elephant, especially when so many aren't really sure what to apply AI-wise, if even possible.

[orthoclad]

5. The true costs are externalized and hidden

The cost of stealing all that human labor and creativity for "training"
The cost of using up rivers of cooling water. This winds up on municipal water bills.
The cost of burning gigaWatts of power. This winds up in utility customers' electricity bills.

The are other more subtle costs, like degrading human minds. Or surveillance in fief to profit and political power.

When you disperse and diffuse costs, it's hard or impossible to invoice them.