The Signal · Column 018

The €114K Number Everyone Will Misread

Andrei van Noordt · 5 August 2026 · 1 min read

Qubit Labs' new figures put average data science pay in Western Europe at $114,000 for 2026. In a year when every second headline announces AI replacing analysts, that number should stop people mid-scroll.

Most readers will file this under general tech inflation and move on. The more useful reading is narrower. Salaries for data science specifically, not AI engineering broadly, are climbing at a moment when generic model-building has become commoditised. That gap tells you where the scarcity actually sits: not in people who can call an API, but in people who can frame the right problem, validate the data, and know when a model's output is quietly wrong.

A rising average salary is not a market getting generous. It is a market admitting it has run out of alternatives.

Companies spent the last two years assuming that cheaper tooling would mean cheaper headcount. What is happening instead is a split. The commodity layer of the work is getting automated and cheapened, while the judgment layer, the part that decides what to build and whether to trust it, is getting paid a premium that keeps rising even as the tools around it get better and cheaper.

For hiring leaders, the practical read is simple. Stop benchmarking data science comp against last year's market and stop assuming AI tooling lets you downgrade the seniority of the hire. Budget now for the premium this data confirms, because the candidates who clear that judgment bar are the ones your AI systems will depend on, and they know their price is only going one direction.