The Social Democratic Party's parliamentary group has floated a plan to tax the extra revenue companies generate by adopting artificial intelligence, aiming to shore up Germany's strained social welfare system as automation threatens to erode payroll tax collections. The proposal, which has no specific tax rate yet, targets efficiency gains from AI, digitalization, and automation. The initiative puts the SPD at odds with Chancellor Friedrich Merz, who advocates for economic growth before redistribution, and has drawn sharp criticism from industry groups warning it could drive investment away from Europe's largest economy.
The SPD's concept, described by German outlet BlackoutNews, envisions a new revenue stream the party calls an "AI dividend" meant to cushion the blow to social security contributions and payroll taxes if artificial intelligence increasingly takes over jobs currently done by people. Party leaders argue the welfare state needs new funding sources to remain solvent in an era of rapid technological change, though they have not specified how high the tax might be set.
The proposal directly conflicts with Merz's stated economic priority: boosting competitiveness and encouraging investment before discussing how to redistribute wealth. Merz, who leads Germany's Christian Democratic Union, has publicly committed to a "growth first, then redistribution" strategy, viewing tax cuts and business incentives as essential to reviving the German economy.
Industry representatives have attacked the proposal as economically self-defeating. Oliver Süme, chair of the eco Association of the Internet Industry, said in a statement that such a tax would give the impression that "using AI could become more expensive and burdensome for companies" at a time when Berlin should be encouraging technological adoption. Rather than impose new levies or regulation, he argued, Germany "should promote this transformation," calling AI essential to industry's ability to innovate and create opportunities for workers.
Süme cited an IW Consult study commissioned by eco showing that AI-driven products and services generated more than €120 billion in German turnover annually — a figure he says underscores what's at stake if Berlin discourages adoption rather than encouraging it. He warned that imposing new costs risks putting Germany at a competitive disadvantage against other nations racing to develop and deploy AI.
Opponents note that German corporate tax rates already hover around 30.1% for 2025, among the highest in the developed world. While the federal government has talked about reducing that burden to attract investment, an AI-specific tax would move policy in the opposite direction.
Beyond economic concerns, critics argue the tax would be nearly impossible to implement fairly. Isolating productivity gains attributable specifically to AI, as opposed to routine software upgrades, new equipment, smarter management, or ordinary market swings, would require a level of accounting precision regulators have never had to apply before.
Small and medium enterprises would likely bear disproportionate compliance costs compared to large corporations, since tracking AI-specific gains would demand new accounting systems and oversight that bigger firms can more easily absorb.
Germany isn't alone in eyeing AI's profits as a piggy bank. In Washington, Rep. Greg Casar (D-Texas) has introduced a federal excise tax on major AI firms to fund a government jobs program, while Sen. Bernie Sanders (I-Vt.) has proposed a one-time 50% tax on AI companies' stock.
The instinct on both sides of the Atlantic is the same: treat efficiency as something to be confiscated rather than rewarded. But businesses that adopt AI are the ones taking on the financial risk in terms of the infrastructure costs, the retraining, and the uncertainty of whether it pays off. Taxing them only when they succeed doesn't protect workers. It just guarantees fewer companies bother trying, and fewer jobs get created in the industries that would have grown around them.
Sources for this article include: