When AI replaces workers, who pays the taxes?

How can governments avoid going bankrupt if they no longer have sufficient tax revenue? This question is becoming an especially pressing one with the impending advent of the Universal Basic Income (UBI) era. With less of an incentive to work, large numbers of people will likely become “workless” — they will not pay any taxes, but will still draw a monthly basic living allowance from the state. 

Once artificial intelligence (AI) upgrades from a tool to infrastructure and genuinely starts to penetrate production and service systems, the first problem that arises will not be rampant unemployment, but the collapse of the tax base. In the industrial era, fiscal systems were built on a stable chain: people worked for an income, governments collected income tax and relevant social contributions; people consumed goods and services, and governments collected consumption taxes. 

In the AI era, however, the traditional link between human production and economic output is beginning to unravel. Output is able to keep growing steadily, but the need for people to participate in production keeps declining. The result might be a stronger economy but also a poorer government, leading to a “growth-driven fiscal deficit”.

Most AI debates focus on job displacement. But in the long run, the more existential question for governments is: who will they tax?

As firms replace labour with algorithms and automation, total wages and headcount are declining, shrinking income tax and social contributions. At the same time, profits brought about by AI tend to be concentrated in the hands of a few platforms and capital holders. As the middle class diminishes and consumption structures weaken, consumption taxes will also become fragile. 

Public finances may therefore fall into a trifold spiral: rising unemployment and falling tax revenues; rising welfare spending and increased fiscal pressure; a shrinking middle class and lesser consumption tax collected. Hence, a situation in which “large numbers of people rely on state support” is not merely a conspiracy theory, but rather the structural outcome of markets that no longer need the same scale of human labour.

The root of the question is: where would the money come from? The conclusion is clear: the tax base must shift from human labour to the income generated through equipment and systems that replace human labour. The so-called “equipment labour” refers to systems that can substitute human work and continuously generate returns on investment, such as algorithms, data, computing power, energy, self-driving vehicles, robotic arms, humanoid robots, drones, platform networks and energy storage. 

In the industrial era, governments mainly taxed humans; in the AI era, governments must learn to tax systems. This is not merely about raising corporate tax rates, as multinationals can shift profits across jurisdictions and reduce their tax burden through accounting strategies.

The more sustainable source of corporate tax revenue rests on real productive capacity and data utilisation, whereas taxation of individuals may increasingly focus on the income generated by the devices and systems associated with them.

The first new category of tax is an AI tax or an automated dividend tax. The logic is: companies gain efficiency dividends by using AI to replace labour, but society as a whole bears the costs of unemployment, retraining and welfare spending. Part of that dividend, therefore, must flow back into the public system.

Surcharges may be levied based on the rate of automated substitution, on AI-driven value added, or through progressive tax rates on additional profits generated by AI. What is crucial is not the form, but the principle: the gains from replacing labour cannot only benefit capital, or society would inevitably fracture.

The second new category is a data-dividend tax. AI capabilities come from huge amounts of data, and the source of that data is often public behaviour, content and the essence of public knowledge. In future, data will increasingly resemble a public resource that firms are allowed to use, but must pay for or share returns on.

Governments can recoup part of the gains through data-use licensing fees, value-added tax on data commercialisation, taxes on cross-border data flows and so on. In essence, this is “collecting rent on behalf of the people”, preventing data returns from long-term monopolisation by a few platforms that use it for free.

The third new category is a computing power tax and energy tax. AI does not consume wages, but it consumes electricity and computing power. The more widespread training and inference are, the stronger the reliance on chips, data centres and power infrastructure. These links are measurable and hard to hide.

Governments can bind fiscal revenues to the physical foundations of AI through a combination of import duties on key chips and GPUs, surcharges on data-centre electricity use, as well as carbon taxes on AI training and so on — the more computing power one uses, the more public costs one bears. This is also one of the most practical approaches.

But taxation alone might not suffice. A deeper trend indicates that the state’s role will evolve from that of a regulator and tax collector, to a provider of infrastructure and an agent of profit distribution.

Historically, key infrastructure such as power grids, ports and airports have gone through nationalisation or semi-nationalisation. In the AI era, key infrastructure would refer to computing power and data networks. Building state-owned or semi-state-owned computing platforms, having sovereign wealth funds hold equity in core AI companies, and taking strategic stakes in critical supply chains would all become realistic options.

The reason is simple: when wealth is primarily generated by intelligent capital, public finances will never be able to keep up with the growth in welfare spending if the state does not participate in the ownership structure.

At the societal level, UBI will no longer be the subject of ideological debates and disputes; it will become a governance tool. UBI will also function as a mechanism to sustain demand, in addition to providing welfare.

This will be achieved by, firstly, stabilising purchasing power through the provision of a basic cash flow. Secondly, consumption will then generate value-added tax and sales-tax revenue. This enables companies to maintain stable operations and pay profit taxes, and the public finances from accumulated taxes will in turn fund the next round of redistribution. Policy priorities will gradually shift from job creation to the maintenance of consumption capacity and social stability, and the importance of consumption taxes will rise accordingly.

The AI era will not push all countries towards the same fate. Factors that will enable countries to transition smoothly include: an ecosystem that supports broad participation in work and income, a computing power and energy base, digital sovereignty and data governance capacity, sovereign funds and fiscal instruments as well as a flexible tax structure coupled with effective, uncorrupted public finance management. By contrast, countries that rely heavily on income tax and that are lacking in technological and energy sovereignty will be more prone to structural deficits, ultimately relying on inflation and debt expansion to survive.

The real challenge is not AI itself, but old-fashioned logic and thinking rooted in the traditional fiscal mindset. AI and job substitution are irreversible, but fiscal systems can be restructured. From an endgame perspective, countries should stop pinning their hopes on fully restoring universal employment. Instead, they should move quickly to shift their tax base and upgrade their institutions, redirecting taxation from human labour to the income generated by the substituting equipment labour and intelligent capital. The role of the state should also be elevated from mere tax collector to infrastructure provider and equity participant, and transform welfare from a moral battlefield into a tool for stabilising demand. 

The core proposition of future public finance can be summed up in one sentence: it is not the people that fund the government, but instead equipment labour and intelligent capital that support the state, which in turn leverages institutions to maintain social stability.

This article was first published in Lianhe Zaobao as “UBI时代国家财政如何重构才不破产”.

Related: Socialism and universal basic income: Creating happy societies in the age of the knowledge economy | Can China turn the trade war into a post-capitalist revolution?