🤖 PLC Meets Model

DeepL 🇩🇪 now translates for half the Fortune 500. That is a translation department no one will ever hire. Nobody files it under AI triumph. It is the quiet shape of the largest market forming in Europe: software that absorbs the work a shrinking workforce will no longer do.

The engine is demographic. Germany 🇩🇪 could be short 768,000 skilled workers by 2028, and the working-age population falls in 22 of the EU’s 27 members by 2050. Italy, Spain and France sit on the same curve. These posts will not go through redundancy. They will stay empty for want of applicants, a demand that deepens on its own for twenty-five years.

DeepL 🇩🇪, Celonis 🇩🇪, SAP 🇩🇪, Siemens 🇩🇪 factory automation: nothing connects these firms except that each makes shrinking headcounts produce more. Celonis finds, across more than a quarter of the Fortune 500, the work no employee will ever have to redo. At Amberg, Siemens cut defects by 75% with AI on a line that ships six million modules a year. The top bricks already exist.

The best software in the world is worth nothing if the Mittelstand will not buy it. A Horváth study from January 2026 puts German SMEs’ AI spending at 0.35% of revenue, down, against a market average of 0.5%. Barely a quarter of German firms with ten or more staff use AI at all. The country ageing fastest is braking exactly when its age pyramid tells it to accelerate. And a shortfall need not be met by software: migration and offshoring have closed European gaps before, and a market that refuses to buy for a decade may not be much of a market. The bullish case assumes the shortage stays a software problem for twenty-five years, and that is an assumption dressed as a forecast.

Capital has read the same page. European AI startups raised more than €20bn in 2025, and in the first quarter of 2026 AI took more than half the continent’s venture capital for the first time. All four of the quarter’s largest rounds went to AI, the biggest, €1.8bn, to Nscale 🇬🇧, a data-centre operator. That layer is legible, it benchmarks, it resells, and the sovereignty story pays. The vertical software that actually serves the labour gap moves slower, one product per trade, long sales into mid-sized firms with no AI function, and draws less money for its trouble. The gap is structural rather than a phase the cycle will later close, because the slower and less liquid layer will always take the thinner cheque.

So the largest European software companies of the decade may well be built where the arms are missing. On the condition, rarely priced in, that someone has the patience to sell to factories that have never once bought AI.