The explosion of the superintelligence economy

What does this mean for techno-intelligence capabilities in 2030 or 2032? Many fear an apocalyptic scenario in which AI terrorists build biological weapons capable of killing billions of people, or that a "rogue" AI will spiral out of control and overtake humanity. It's easy to understand why such extreme risks garner so much attention. However, they distract from the immediate, likely, and predictable (and equally surprising) effects of a non-apocalyptic AI.

Before 1700, the global economy grew, on average, by 8% per century. Anyone who predicted what happened next would have been called a lunatic. For the next 300 years, as the Industrial Revolution progressed, growth averaged 350% per century. This led to lower mortality and higher fertility. Larger populations generated more ideas, which led to even more rapid expansion. Because of the need to incorporate human talent, the cycle was slow. Over time, greater wealth led people to have fewer children. This boosted the standard of living, which grew at a steady rate of about 2% per year.

AI doesn't face that demographic constraint. Technologists promise it will rapidly accelerate the pace of discovery. Sam Altman, CEO of OpenAI, expects AI to be capable of generating "novel insights" next year. AI already helps program better techno-intelligent models. Some say it will oversee its own improvement by 2028.

Hence the possibility of a second burst of economic growth. If computing power brings about technological advances without human intervention, and if a sufficient portion of the profits are reinvested in the creation of even more capable machines, wealth could accumulate at an unprecedented rate. Economists have long been aware of the relentless mathematical logic of automating idea discovery. According to a recent projection by Epoch AI, an optimistic think tank, when AI can perform 30% of tasks, annual growth will exceed 20%.

Convinced techno-believers, including Elon Musk, conclude that self-improving AI will create superintelligence. Humanity will then have access to every conceivable idea, including those needed to build the best robots, rockets, and jets. Access to energy and human lifespan will no longer be limited. The only constraints on the economy will be the laws of physics.

You don't have to go to that extreme to imagine the astonishing effects of AI. Consider hypothetically the simple incremental step toward human-equivalent intelligence. In labor markets, the cost of using computing power for a task will limit the wage to perform it: why pay a worker more than digital competence? In turn, the dwindling number of superstars whose skills are not automatable and can directly complement AI would enjoy enormous benefits. Most likely, the only winners will be the owners of AI-relevant capital, who will capture an ever-increasing share of economic output.

Everyone else will have to adapt to gaps in tech-savvy capabilities and the spending of the nouveau riche. Where there is a bottleneck in automation and labor supply, wages will rise rapidly. These effects, known as "cost sickness," could be so strong as to limit the explosion in measured GDP, even if the economy were to completely change.

The new patterns of abundance and scarcity will be reflected in prices. Anything AI helps produce—for example, goods from fully automated factories or digital entertainment—will see its value plummet. If anyone fears losing their job because of AI, at least they can look forward to many such things. Where humans are still needed, cost sickness could set in. Knowledge workers who shift to manual labor may eventually find they can't afford as many childcare centers or restaurant outings as they can today. And humans may eventually compete with AI for land and energy.

Read also The race for AI drives up the salaries of elite computer programmers The Economist

This economic disruption will be reflected in financial markets. Large swings in company stocks are likely as the uncertainty of which companies are winning and which are losing in the fight for market share becomes clearer. There will be a voracious desire to invest, both to generate more tech-smart capabilities and to ensure infrastructure and factories can keep pace with economic growth. At the same time, the desire to save for the future may disappear, as people anticipate much higher incomes (especially the wealthy, who save the most).

Therefore, persuading people to give up capital to invest will require much higher interest rates—perhaps so high that they will drive down long-term asset prices despite explosive growth. Experts disagree, but in some models, interest rates rise at the same pace as growth, or even more. In an explosive scenario, that would mean having to refinance debt at 20–30%. Even debtors whose incomes were rising rapidly could be affected; those whose incomes were not tied to runaway growth would be severely hurt. Countries unable or unwilling to take advantage of the AI boom will face capital flight. There is also the potential for macroeconomic instabilities elsewhere, as inflation soars as people spend their expected gains and central banks fail to raise interest rates quickly enough.

There will be a voracity to invest, but less tendency towards traditional savings.

Considering this scenario, even as a thought experiment, is dizzying. Will humanity be able to cope with such a situation? Growth has accelerated before, but during the Industrial Revolution, mass democracy didn't exist; the Luddites, history's most famous technophobes, didn't have the right to vote. Even if average wages were to rise, greater inequality could lead to demands for redistribution. The state, for its part, would also have more powerful tools to control and manipulate the population. Politics would therefore be highly unstable. Governments would have to rethink everything from the tax base to education and the protection of civil rights.

Despite this, the rise of superintelligence should provoke astonishment. Dario Amodei, chief executive of Anthropic, told The Economist last week that he believes AI will help treat previously incurable diseases. The way to view another acceleration, if it ever happens, is as the continuation of a long miracle, possible only because people have embraced disruption. Humanity may be outsmarted. It will still need wisdom.

© 2025 The Economist Newspaper Limited. All rights reserved

Translation: Juan Gabriel López Guix