AI, Autonomous Labor, and the Reconfiguration of the Global Economy

What happens to taxation, ownership and economic power when productivity no longer depends on employment in the same proportion?

Ronen Moshe, Adv.

August 2026

---

The familiar debate asks whether artificial intelligence will replace jobs. This article asks a different question: what happens to an economy when human labor is no longer the main link between productivity, income, taxation and participation? The answer may reshape not only employment, but also ownership, sovereign finance, energy policy, legal responsibility and geopolitical power.

---

Beyond another technological revolution

Most discussions about artificial intelligence begin with the same question: which jobs will disappear?

I believe this is the wrong place to start. Employment is certainly part of the story, and probably an important part of it. But the more consequential question is what happens to an economic system when human labor gradually ceases to be the principal mechanism through which productive capacity is created, income is distributed, taxes are collected and economic participation is organized.

For more than a century, modern economies have operated around a relatively stable relationship. People work, businesses pay wages, households consume, governments tax income and consumption, and economic growth allows both public and private debt to be serviced over time. That relationship is not the whole economy, but it is one of its central organizing mechanisms.

What first drew my attention to this issue was not the usual prediction that AI might eliminate particular professions. It was the tension between two trends that are unfolding at the same time: the possibility that production becomes progressively less dependent on labor, and a global fiscal system that remains heavily dependent, directly and indirectly, on income and consumption generated by human participation in the economy. Global gross public debt stood at 93.9 percent of GDP in 2025 and, on the IMF’s April 2026 projections, is set to reach 100 percent by 2029.[1]

That does not mean AI will trigger a sovereign-debt crisis. It means that the institutional assumptions behind taxation and debt deserve closer examination if the source and location of economic value begin to change.

Nor does any of this require a science-fiction world in which human work disappears altogether. It is enough for a growing share of output to be generated without a proportional increase in human labor. Once productivity and employment begin to separate on a meaningful scale, the consequences extend far beyond the labor market.

AI as a cognitive multiplier of capital

Earlier industrial revolutions primarily multiplied physical capability. Machines allowed one worker to lift more, manufacture more, travel farther and produce at scales that previously required much larger amounts of labor. Artificial intelligence introduces a different form of leverage: it begins to multiply cognitive capability.

A company can increasingly perform analysis, scheduling, monitoring, translation, programming, optimization and parts of professional reasoning without increasing its workforce at the same rate. That does not make human judgment irrelevant. In many fields it remains decisive, and the International Labour Organization’s most recent global assessment emphasizes transformation of jobs rather than wholesale replacement. At the same time, the scale of exposure is already significant: the IMF has estimated that almost 40 percent of global employment is exposed to AI, rising to about 60 percent in advanced economies, while the ILO’s 2025 index finds that roughly one in four jobs worldwide has some degree of exposure to generative AI.[2]

The economic point is not that every exposed job disappears. It is that additional units of analysis and coordination can increasingly be produced at very low marginal cost. That changes the relationship between labor and capital. A business with access to powerful models, proprietary data and computing infrastructure may be able to produce output that once required a much larger organization.

This is why I see AI less as a simple substitute for individual employees and more as a multiplier of capital. The question is not only who can use AI, but who owns the systems, data and infrastructure through which AI becomes economically productive.

The ownership problem

There is a tendency to speak about automation as though society automatically receives the benefits created by machines. It does not. The distributional effect of autonomous production depends heavily on ownership.

Imagine two economies with identical technological capabilities. In the first, ownership of productive assets is broadly distributed through pensions, investment funds, public participation and private savings. In the second, the same autonomous productive capacity is controlled by a very small number of corporations and individuals. The technologies may be identical. The societies produced by them could be radically different.

This concern is not merely theoretical. The OECD has noted that the global labor share of income declined by around six percentage points between 1980 and 2022, and that AI-driven automation could continue the shift toward capital income, with stronger inequality effects where capital ownership is concentrated.[3]

The central economic question of the autonomous age may therefore be not who works, but who owns the systems that work.

I do not regard that as an argument against private ownership. Private investment, risk-taking and the expectation of profit are major engines of technological development. It is an argument that ownership structure will matter more, not less, if machines become capable of producing a larger share of society’s economic value.

For that reason, I am not persuaded that a simple “robot tax” captures the real issue. Taxing a machine because it replaces a worker focuses on the visible event - the lost job - rather than the deeper transformation. The more important question is how a society preserves broad participation in an economy whose most productive assets require progressively less human labor.

Autonomous labor and the decline of labor scarcity

AI on its own remains largely confined to information. Robotics changes the equation because it connects machine intelligence to the physical economy. A sufficiently capable autonomous system can work continuously, operate in environments unsuitable for humans and reproduce the same task across multiple machines without the demographic constraints that apply to a human workforce.

The transition is already visible in industrial automation. According to the International Federation of Robotics, 542,000 industrial robots were installed worldwide in 2024, more than double the number installed ten years earlier, and the global operational stock reached about 4.66 million units. These figures do not imply that humanoid robots are about to replace the global workforce. They do show, however, that physical automation is no longer a marginal feature of manufacturing.[4]

I would be cautious about predicting an end to labor scarcity. Human capability is too diverse, and economies are too complex, for such a claim. But autonomous systems create a credible path toward making labor scarcity materially less important as a constraint on production.

That could alter the logic of global manufacturing. Countries with large populations or inexpensive labor have long enjoyed important cost advantages. If a growing number of tasks can be performed by machines at comparable cost across different regions, geography does not disappear, but the reasons for choosing one location over another begin to shift. Energy, logistics, regulation, capital, access to chips and infrastructure may matter more; the local wage bill may matter less.

The demand problem: production without wages

There is another part of the debate that, in my view, receives too little attention. An economy does not function merely because it can produce goods efficiently. Someone must also be able to buy them.

The modern economic cycle links production and consumption through wages. Employees are not only a cost of production; collectively, they are also consumers. Employment generates income, income creates demand, and demand creates corporate revenue. Large-scale automation can weaken this circular relationship if productivity rises while labor income represents a declining share of output.

This creates a paradox. A highly automated economy could become extraordinarily efficient at producing goods and services while simultaneously weakening the mechanism through which people obtain the purchasing power required to consume them.

That is why I do not think technological unemployment can be analyzed only as a welfare question. It is potentially a macroeconomic question. If labor ceases to be the primary mechanism through which income reaches households, another mechanism will eventually have to perform part of that function - through ownership, transfers, dividends, public investment, tax policy or arrangements that have yet to emerge.

The precise solution is open to debate. The structural problem is harder to avoid.

Fiscal policy after wage-centered taxation

Modern fiscal systems do not rely exclusively on personal income tax. Governments collect corporate taxes, consumption taxes, property taxes, capital taxes and many other forms of revenue. It would therefore be wrong to suggest that a fall in wage taxation automatically deprives the state of revenue.

But labor income is embedded in the fiscal system more deeply than the tax label suggests. Wages generate income tax and social contributions directly, while also financing household consumption that generates VAT, sales taxes and business income. The relevant risk is therefore not that economic activity disappears, but that the location of taxable value changes faster than fiscal institutions adapt.

A country could become more productive while finding that a growing share of that productivity is concentrated in mobile capital, intellectual property, cloud infrastructure and multinational platforms that are easier to structure across borders than a domestic payroll.

The fiscal question then becomes unavoidable. Should autonomous output be taxed differently? Should governments rely more heavily on profits, capital income, consumption or energy use? Should sovereign wealth funds or public pension systems hold larger stakes in the infrastructure of the AI economy? I do not think there is yet a single convincing answer. But continuing to assume that employment and economic growth will remain tightly correlated seems increasingly risky.

Energy, compute and the new scarcity

Economic systems are always organized around scarcity. Industrial economies were constrained by human labor, physical capital, raw materials and energy. If intelligence becomes cheaper to reproduce and some forms of labor become easier to automate, the physical constraints do not disappear. They move.

AI requires computing infrastructure. Computing infrastructure requires semiconductors, data centers, cooling systems and large quantities of reliable electricity. The International Energy Agency estimates that data centers consumed about 415 TWh of electricity in 2024 and projects roughly 945 TWh by 2030 in its base case, with AI the most important driver of the increase.[5]

Compute itself is also highly concentrated. The World Bank reports that, as of mid-2025, high-income countries accounted for about 77 percent of global co-location data-center capacity and 97 percent of the capacity represented by the world’s top 500 high-performance computing systems.[6]

This leads me to a broader conclusion. As labor becomes less binding, economic power may depend increasingly on the ability to combine energy, compute and physical infrastructure. In the twentieth century, control over oil strongly influenced industrial and geopolitical power. In the AI economy, the strategic resource may be less a single commodity than an integrated system: abundant energy, advanced semiconductors, computing capacity and the ability to convert all three into productive intelligence.

Semiconductor supply chains illustrate the problem. OECD research describes the sector as highly concentrated and globally interdependent, with critical inputs and production stages clustered in a limited number of regions and economies. A country with excellent AI models but inadequate energy or chip access may therefore face very different constraints from a country that controls those physical foundations.[7]

Law, responsibility and autonomous decision-making

As a lawyer, I find the legal side of this transition especially revealing because law is built around actors. Someone signs a contract. Someone drives a vehicle. Someone makes a professional decision. Someone breaches a duty. Autonomous systems complicate that structure.

Consider an AI system that receives data from one company, runs on a model created by another, is configured by a third party and then instructs a physical machine to perform an action that causes damage. Who made the relevant decision? The developer? The model provider? The owner? The operator? The person who supplied the data? Traditional legal systems can allocate responsibility among several actors, but autonomous systems can make those chains longer, faster and harder to reconstruct.

European regulation already points toward the kinds of institutional tools that will become increasingly important. The EU AI Act requires high-risk systems to support logging and traceability, provide sufficient transparency for deployers and incorporate effective human oversight. These are regulatory obligations, but they also reveal something broader: in an autonomous economy, accountability will depend less on reconstructing a single human decision and more on documenting how a system was designed, supervised and allowed to act.[8]

Liability law is moving in the same direction. The EU’s revised Product Liability Directive expressly treats software, including AI systems, as products for the purposes of no-fault product liability and addresses defects linked to software updates and machine-learning behavior under a manufacturer’s control.[9]

The legal transition, in my view, will therefore be from a system focused primarily on individual human error toward one increasingly concerned with system design, auditability, supervision, foreseeable failure and allocation of responsibility across technological chains.

Trust will become an economic asset. Businesses and consumers will not adopt autonomous systems simply because they are efficient. They will also need reasonable confidence that those systems behave predictably, can be audited, and can fail in ways that the legal system knows how to address. Legal certainty will not merely regulate the AI economy from outside; it may become one of the conditions that allows it to scale.

The geopolitical reallocation of power

For much of history, population was closely connected to national power. Large populations meant larger armies, larger workforces, larger internal markets and greater productive potential. Autonomous systems may weaken some of those relationships without eliminating them.

A relatively small country with abundant energy, advanced computing infrastructure, capital and technological capability may eventually operate productive systems that would once have required a much larger workforce. Population will still matter - for markets, creativity, legitimacy and many forms of military power - but it may become a less reliable proxy for productive capacity.

At the same time, control over platforms and infrastructure becomes more important. Semiconductors, cloud systems, foundational models, autonomous manufacturing and energy networks are not merely commercial assets. They are increasingly part of industrial policy, strategic autonomy and national security.

This may also complicate the traditional boundary between states and corporations. Some technology companies already operate infrastructure at a scale once associated almost exclusively with governments. If autonomous systems become central to production, the entities controlling those systems could acquire forms of economic and strategic influence that do not fit neatly within the old distinction between private corporate power and geopolitical power.

Money after labor scarcity

This brings me to what I consider the deepest question in the argument: what is money actually doing in an economy? Money is a medium of exchange and a store of value, but economically it is also a mechanism for allocating access to scarce resources.

In the present system, employment is one of the main channels through which individuals receive money and therefore obtain access to goods and services. If human labor becomes less scarce while energy, compute, land, infrastructure and ownership remain scarce, money does not become irrelevant. Its role changes.

I am therefore less persuaded by predictions that advanced automation will make capitalism or money disappear. The opposite may initially occur. Ownership of scarce productive systems could become even more valuable because those systems would embody not only machinery, but productive intelligence itself.

My concern is not that capitalism necessarily collapses under AI. It is that the mechanism by which people participate in capitalism may change faster than the institutions that distribute income, ownership and political legitimacy.

Conclusion

The most important consequence of artificial intelligence may not be technological unemployment. It may be the gradual separation of productivity from human labor.

Once that separation becomes economically significant, a series of assumptions that currently appear unrelated begin to move together. Tax systems designed around wages become less reliable. The connection between employment and consumption weakens. Ownership of autonomous productive systems becomes more important. Energy and compute become increasingly strategic. Legal systems must allocate responsibility for decisions that no single human being directly made. And geopolitical power depends more heavily on control of technological platforms and the infrastructure that supports them.

None of this means that we are about to enter a world without work. Nor does it mean that technology inevitably produces either prosperity or inequality. Those outcomes will depend on institutions, policy and ownership. History gives us good reason to be skeptical of confident forecasts in either direction.

But I believe the direction of the transition is visible enough to justify asking the question now rather than after the economic structure has already changed.

For the industrial economy, the central problem was how to combine labor, capital and energy to increase production. For the autonomous economy, the central problem may be how to distribute income, ownership and economic power when production no longer requires human participation in the same proportion as before.

That is not simply a question about artificial intelligence. It is a question about the economic architecture that comes after it.

[1] International Monetary Fund, Fiscal Monitor: Fiscal Policy under Pressure—High Debt, Rising Risks (Apr. 2026), ch. 1 (global gross government debt: 93.9% of GDP in 2025; projected 100.0% in 2029).

[2] International Monetary Fund, Gen-AI: Artificial Intelligence and the Future of Work, Staff Discussion Note 2024/001 (Jan. 2024); International Labour Organization, Generative AI and Jobs: A Refined Global Index of Occupational Exposure, Working Paper 140 (May 2025), DOI: 10.54394/HETP0387.

[3] OECD, The Impact of Artificial Intelligence on Productivity, Distribution and Growth, OECD Artificial Intelligence Papers No. 15 (2024), DOI: 10.1787/8d900037-en.

[4] International Federation of Robotics, World Robotics 2025: Industrial Robots (2025) (542,000 installations in 2024; approximately 4.664 million industrial robots in operational use worldwide).

[5] International Energy Agency, Energy and AI (2025), ‘Energy demand from AI’ (about 415 TWh of data-center electricity consumption in 2024; approximately 945 TWh projected for 2030 in the base case).

[6] World Bank, Digital Progress and Trends Report 2025: Strengthening AI Foundations (2025), Compute chapter (77% of global co-location data-center capacity and 97% of top-500 HPC capacity located in high-income countries as of June 2025).

[7] OECD, Mapping the Semiconductor Value Chain: Working Towards Identifying Dependencies and Vulnerabilities, OECD Science, Technology and Industry Policy Papers No. 182 (June 2025), DOI: 10.1787/4154cdbf-en.

[8] Regulation (EU) 2024/1689 (Artificial Intelligence Act), arts. 12–14 (record-keeping, transparency and human oversight requirements for high-risk AI systems).

[9] Directive (EU) 2024/2853 on liability for defective products, recitals 13, 32 and 50 and arts. 4 and 12 (software, including AI systems, within the product-liability framework; software updates, learning behavior and multiple economic operators).