What should a country do when a technology reshapes its economy but most of the money it generates is booked somewhere else? That question sits behind a proposal in Germany to tax AI revenue, reported by Golem: the politician Wildberger argues that turnover generated by artificial-intelligence services in Germany should be captured by the tax system.
The logic is familiar from the digital-services-tax fights of the last decade. AI models are trained, hosted and sold largely by a small number of mostly US firms, with Chinese labs increasingly competitive. Customers in Germany pay for the subscriptions, the API calls and the inference, but the profit lands in another jurisdiction, where it is taxed at another rate. Germany's own AI sector is smaller and mostly consumes foreign models rather than selling its own, which means a levy would, in the short run, mostly hit overseas providers and their local customers.
Economists have two standard objections. The first is incidence: a tax on a company's revenue is always a tax on something — customers, workers or investment — and in markets with little competition it is often passed to the buyer. Taxing turnover rather than profit can also punish thin-margin resellers and startups harder than it punishes the giants the measure is aimed at. The second is legal: a country cannot simply tax a foreign company's revenue without running into double-taxation treaties, and unilateral digital levies have already drawn retaliation threats against France, the UK and Austria. The OECD's attempt to reallocate taxing rights globally — its Pillar One work — has stalled, which is precisely why national proposals keep resurfacing.
For now the proposal is at least as much a political signal as a finished policy design. Whether it becomes a bill, is folded into an EU-level instrument, or fades back into commentary is the thing to watch. What is no longer in question is the tension underneath it: Europe buys AI far more than it sells it, and every budget season makes that imbalance more visible.
