The AI economy's petrostate problem

Artificial intelligence is widely expected to become one of the most important drivers of economic growth in the coming decade. The World Economic Forum estimated in April 2026 that AI infrastructure investment could exceed US$5 trillion by 2030, as governments and companies race to build the computing capacity needed to power the technology.

Periods of extraordinary wealth creation, however, have rarely been defined solely by the scale of the economic gains they produce. They have also been defined by how that wealth is distributed. Oil transformed global markets and generated immense prosperity, yet in many countries it also concentrated economic and political power among relatively few actors.

Artificial intelligence is not a natural resource, but it is prompting a similar question. If the infrastructure underpinning the world's future economy becomes controlled by a handful of firms, the defining challenge of the AI age may not be developing more powerful models. It may be preventing unprecedented technological wealth from becoming unprecedented economic concentration.

When wealth no longer depends on workers

Modern economies have historically relied on a broad fiscal bargain. Businesses hired workers, workers earned wages and governments taxed both labor and corporate profits to finance public services. As artificial intelligence enables businesses to generate more output with fewer workers, that relationship could begin to change.

A March 2026 Brookings Institution paper argues that governments whose tax systems rely heavily on wages could face growing fiscal pressure if AI reduces labor's share of national income. The IMF reached a similar conclusion in its June 2026 report, warning that AI could erode labor tax bases while increasing demand for social spending, requiring governments to rethink how public revenues are generated.

The concern echoes one of the central lessons of the political economy of oil. Governments financed primarily through resource rents often face different political incentives than those whose revenues depend on broad-based taxation.

Artificial intelligence is unlikely to eliminate the role of labor altogether. Even optimistic forecasts anticipate decades during which people continue to work alongside AI systems. Yet if a growing share of economic value flows toward the owners of compute, data and frontier models rather than toward workers themselves, governments may eventually confront many of the same fiscal questions that resource-dependent economies have faced for decades.

The new oil wells are data centers

The comparison between artificial intelligence and oil is not based on the technology itself. It rests on the economics of ownership. Oil became strategically valuable because those controlling production captured a disproportionate share of global wealth, while those without it became consumers rather than producers.

Artificial intelligence depends on a different resource, but one that is becoming no less strategic. Training frontier AI models requires vast computing power, advanced semiconductors, enormous quantities of electricity, sophisticated cloud infrastructure and increasingly large data centers. Building and operating that infrastructure demands investments measured in tens of billions of dollars, creating barriers to entry that few companies or governments can overcome.

A June 2026 Carnegie Endowment report argues that compute is emerging as a strategic resource comparable to energy or industrial capacity. Countries able to secure computing infrastructure, reliable electricity and advanced semiconductor supply chains will be better positioned to capture AI's long-term economic benefits, while others risk becoming consumers rather than producers of the technology.

Ownership therefore becomes central to the debate.

The companies building frontier AI systems are not simply developing software. They are constructing the infrastructure on which future AI applications may depend. As businesses integrate AI into sectors ranging from manufacturing and finance to healthcare and logistics, an expanding share of the global economy could come to rely on computing infrastructure controlled by a relatively small number of firms.

The politics of concentration

Unlike previous digital technologies, frontier AI systems exhibit unusually strong economies of scale. Training the most advanced models requires specialized chips, vast computing infrastructure, proprietary datasets and enormous financial resources, allowing relatively few firms to dominate frontier development.

The Yale Law & Policy Review argues that many concerns surrounding artificial intelligence stem less from the technology than from concentration of the infrastructure underpinning it. The IMF similarly warns that these dynamics could reinforce "winner-take-most" markets in which a small number of firms capture a disproportionate share of AI-generated profits, leaving countries with weaker digital infrastructure increasingly dependent on technologies developed elsewhere.

That distinction is particularly important for countries outside today's leading AI hubs. During the industrial revolution, nations could industrialize by building factories. During the digital revolution, they could develop software industries with comparatively modest capital requirements. Competing at the frontier of artificial intelligence, however, increasingly depends on access to computing infrastructure whose cost places it beyond the reach of most governments and companies.

Those dynamics do not make a digital petrostate inevitable. They do, however, make public policy more important than at any previous stage of the digital economy.

Avoiding the digital resource curse

Competition policy will be central to determining whether AI remains broadly accessible or becomes concentrated among a handful of firms. Governments are also likely to face growing pressure to invest in digital infrastructure that allows businesses of all sizes to adopt advanced AI systems rather than leaving access confined to those able to finance enormous computing resources.

Fiscal policy will also require adaptation. As AI changes the balance between labor and capital, governments will need tax systems capable of capturing AI-generated economic activity while preserving the revenues needed to finance public services. Because AI markets operate across borders, international coordination on competition and taxation is also likely to become increasingly important.

Oil demonstrates that extraordinary wealth can either broaden prosperity or concentrate power. Artificial intelligence presents governments with the same choice. Whether the AI economy resembles a diversified knowledge economy or the political economy of a modern petrostate will depend less on the technology itself than on the institutions governing who owns it, who profits from it and how its gains are shared.