Will The Cost of Dutch AI Save Expat Jobs?

For the past couple of years, there has been an anxious underlying tone in the Dutch workplace. Everyone wondered how much longer their role would remain necessary before AI quietly (or not so quietly) replaced them.

With policy and regulations around AI tightening, there is also another catch: Using AI on subscription costs money. Not much, but it does add up. Whenever AI drafts an email, creates a report, or writes some code, you burn through ‘tokens,’ and the more you do so, the more it adds to your bill. The question then becomes: how much do these tokens cost, will the cost of AI tokens prevent mass redundancies, and what can I do to prevent it?

The Token Crisis

To understand how AI can price itself out of replacing us, we must understand the real cost of AI. While expats like yourself get a paycheck, paid NLP models charge per “token.” This means that each time AI reads or writes a word, the company’s account is charged for a “micro-transaction.” While the costs for tasks such as spell-checking were negligible, AI requests have morphed into planning, writing, and executing complex plans. It’s this sort of thing that has escalated costs, as it performs “thinking and acting” loops. To figure out step 18 of a problem, AI must go through and re-read the entire chat history, passing back and forth through its artificial memory dozens of times a minute. This can lead companies to spend money quickly.

As the first half of 2026 progressed, the theoretical concern above became very real among corporations. Uber, with its EMEA headquarters in Amsterdam, rolled out an AI coding tool named Claude Code to its engineering teams. But mid-year, the CTO had to admit that Uber had exhausted its 2026 AI budget in just four months. At Microsoft, after finding that the monthly token costs required to run AI coding tools exceeded the baseline salaries of the engineers using them, the tech giant canceled thousands of direct licenses.

But token prices have been falling, which could benefit the Dutch AI sector. The costs of these AI models fall each year as the tech gets better and the competition and the market do their bit to make it cheaper. What’s expensive today may not be in 12 months. If cost is the only thing keeping a role for a person, then that person could be on borrowed time.

Does Dutch AI Being Expensive Actually Keep People Employed?

In a way, but there’s more to it than that. The real cost of AI is complex. If you think of a medium-sized firm implementing AI to cover every single query, this may end up costing more than a small team. Within the setup, overseeing operations and dealing with almost certain mistakes from time to time, it may be worth keeping that team after all. Even with cheaper token prices, there are still trust and quality issues. A tool that produces lots of mistakes and dents credibility is not a worthwhile cost. That’s why the “will AI take my job” question rarely has a clean yes-or-no answer.

Read Also: 14 Great AI Writing, Translation, and Coding Tools for Expats

The Cost of Computing: Data Centers And More….

The problem is greater than just software loops. What’s needed to run Dutch AI at enterprise scale is mind-boggling. We’re talking about huge data centers, massive water use for cooling systems, property taxes, utility bills, and more. Around 75 U.S. data center projects worth about $130 billion have been canceled or delayed due to component shortages and infrastructure limitations.

Since the need for AI compute is massively greater than the actual physical supply of server space, the cost of digital labor has gone through the roof. If you are a huge telecom company or a tech giant that processes 8 billion tokens every day, that small cost bump turns into tens or hundreds of millions of dollars in monthly charges. Where endless power grids and industrial-scale infrastructure are needed for artificial intelligence, humans really require workspace and a laptop.

The Dutch Reality

The labor market in the Netherlands is one of the most educated in Europe. Currently, AI is far more likely to replace more repetitive manual work. Roles that involve complex problem-solving (of which the Netherlands has plenty) are more difficult for AI to automate.

In a country where half the land is reclaimed from the sea, and water management is a national sport, the Dutch instinct isn’t panic. It’s “fine, how do we resolve this?” This is backed up by history. The Netherlands has reinvented its economy more than once, moving from agriculture and shipping to finance, technology, and high-value services. With each transition, some jobs were displaced and new ones created.

The Productivity Promise and Its ‘Catch’

AI has the power to increase productivity. No question. When a marketer or lawyer can automate certain parts of the ‘leg work’ in their role, it leaves them freer to focus on the more impactful parts of their role. This isn’t the destruction of jobs; it’s their enhancement. The catch lies in its distribution. This gain will benefit society only if its wins are shared equally through better pay, working conditions, or opportunities. The kind of trade-off that the Dutch polder model aims to achieve.

The issue for expats could be reskilling, and maybe moving into the roles that AI creates. The country is strong in vocational training, affordable university courses, and learning, and offers many avenues for expats to change careers, whatever stage of life they’re in.

The Crunch: Will AI Costs Protect Dutch Jobs?

Likely not. And if so, not for too long. AI automation may not be as imminent as many think, especially as businesses figure out the real cost of AI and the drawbacks involved. The country, with its strong education system and options to reskill, is in a good position, and the real safeguard is how a skilled, pragmatic, and famously stubborn country decides to adapt. And if the past is any guide, adapting is something the Dutch tend to be rather good at. If the country prepares properly, artificial intelligence will likely follow the same pattern: short-term discomfort, manageable over the long run.