Artificial intelligence: revolution is guaranteed, but not returns.

This series of articles aims to answer investors’ questions about how to invest their capital in the financial markets. These reflections take a long-term perspective and seek to provide general answers to the most difficult decisions. Our analysis is based on historical data and evidence spanning time. At the heart of each article lies a simple question: which investment option should I prioritize?

In 1964, US President Lyndon Johnson created a federal commission dedicated to the future of employment. The president of International Business Machines (IBM) and the inventor of the Polaroid camera were among its members. The commission’s mandate seemed urgent: automation was transforming factories and offices, threatening America’s skilled workforce. When the committee published its report in 1966, the unemployment rate in the United States had fallen from 5.1% to 3.8%. The committee had taken fifteen months to diagnose a problem that prosperity had solved.

One of the commission’s conclusions deserves to be displayed in every investor’s office: ” Technology eliminates jobs, but it doesn’t destroy work. ” Sixty years later, nearly 60% of American jobs are in professions that didn’t exist in 1940. Not only does technology not destroy work, but it creates new categories of jobs that no one could have anticipated.

This pattern has repeated itself for four centuries. During the first Industrial Revolution (from approximately 1760 to 1840), the Luddites who began destroying English looms in 1811 were not simply unskilled laborers, but the highest-paid artisans in their field. They opposed manufacturers using new machines to produce inferior goods with cheaper labor. Yet, this textile industry became Britain’s largest employer, as cheaper fabrics created new markets, new consumers, and new industries.

The lesson of industrial revolutions is this: upheavals are temporary, not permanent. Old and new technologies can coexist for years, even decades, creating space for both policy measures and investments. The Second Industrial Revolution (from approximately 1870 to 1914) produced a counterintuitive economic outcome. The number of horses—a technology that the railroad was supposed to eliminate—exploded. In England, the population of draft horses peaked in 1901, seven decades after the opening of the first line connecting Liverpool to Manchester. In the United States, the number of horses and mules increased sixfold during this period of railway expansion. This phenomenon can be explained by the fact that the railroad created a demand for last-mile logistics, a task that only horses could perform. The technology meant to make horses redundant gave them a new function.

Technology eliminates jobs, but it doesn’t destroy work.

At the same time, widespread electrification brought household appliances into American homes. In 1900, only 5% of American wives worked outside the home; by 2000, that figure had reached nearly 60%. Household appliances played a significant role in this shift . Far from eliminating drudgery, the washing machine, for example, freed working women trapped in unpaid domestic labor. Yet, in the 1920s, no one saw the washing machine as the key to the future of the American job market.

The third industrial revolution, which began around 1950 with the rise of mass electronics, was supposed to generate a “jobless recovery.” However, the number of jobs worldwide increased by more than a billion between 1990 and 2020. The automated teller machine (ATM) is perhaps the best example of this. When more than 400,000 ATMs were installed, massive unemployment among tellers was anticipated. Instead, because ATMs reduced costs per branch, banks expanded their networks and tellers were redeployed.

The same phenomenon has been observed with looms, horses, and washing machines: the automation of one category of jobs increased demand in the adjacent category. The Third Industrial Revolution effectively led to a polarization of employment, with the erosion of routine middle-level tasks accompanied by a simultaneous expansion of highly skilled jobs and low-wage services, which reshaped income distribution and laid the groundwork for the ongoing Fourth Industrial Revolution.

Technological disruption and capital allocation

The Austrian economist Joseph Schumpeter explained the chaotic nature of capitalism: it operates through creative destruction, meaning that it does not develop by progressively improving what already exists, but by replacing it. The railway rendered the stagecoach obsolete; the automobile made the horse-drawn carriage a novelty. Each wave creates considerable wealth, but distributes it unevenly and destroys certain forms of subsistence inherited from previous structures.

This is crucial for investors. Schumpeter argued that transformative technologies ultimately reshape economies and generate extraordinary value. What he failed to mention is the pattern revealed by historical analysis: the initial phases of disruption systematically lead to a misallocation of capital and speculative excesses. It is precisely in this gap that Schumpeter passes the baton to Hyman Minsky.

Stability itself is destabilizing.

Minsky examined this gap between the certainty that a technology will transform the world and the uncertainty about who will benefit. He found that stability itself is destabilizing. When an economy performs well for an extended period, caution gives way to confidence, confidence to optimism, and optimism—without safeguards—to recklessness. For success can erode the very caution that made it possible.

Minsky formalized this progression into three stages: first, covered financing. Borrowers service the debt, both interest and principal, using their cash flow. Today’s artificial intelligence (AI) giants—Microsoft, Alphabet, Amazon, and Meta—are in this phase. Their capital expenditures are colossal in absolute terms but are financed by the free cash flow generated by their core businesses. If AI monetization proves disappointing, they absorb the loss. Second, speculative financing. Borrowers cover the interest but rely on refinancing to repay the principal, anticipating a potential revenue recovery. Some large-scale AI strategies now operate at this stage: coherent in their design but dependent on demand materializing at the scale and pace required by their commitments. The margin for error is smaller. Third, survival depends on rising asset prices. This is the final, most dangerous stage, which Minsky called “Ponzi” finance. The parallel isn’t systemic in AI, of course, but there are startups valued at billions with little revenue, sustained by their narrative rather than their cash flow. When the narrative takes precedence over the numbers, Minsky would identify a familiar pattern.

For AI, the question isn’t whether the technology is transformative, but whether the financial architecture surrounding it can withstand a delay in the timing of returns. Schumpeter would argue that the technology will endure regardless. And Minsky would add that many of the investors funding this transformation won’t survive it.

Few initiators, many followers

Revolutions create extraordinary wealth and destroy it with remarkable efficiency. A technology can be transformative without protecting the investor from total loss. The railway frenzy of the 1840s illustrates what happens when private capital loses all sense of proportion. The first lines paid dividends, and investors took notice. By 1845, the Bank of England had lowered its rates, government bonds offered unattractive yields, and it was possible to buy shares in railway companies with a 10% down payment. In 1846 alone, Parliament authorized the creation of 263 new companies. By 1848, railway companies accounted for 71% of British stock market capitalization. Then, basic arithmetic took over. When interest rates rose, share prices halved, and a third of the planned lines were never built. Ironically, the skeptics were wrong about the technology, and the enthusiasts who financed 90% of the current British rail network were wrong about the investment. The rail revolution did happen, but many of the investors who backed it ended up ruined.

The question today regarding AI is whether consumers will once again favor integrated and practical solutions over self-service coding.

The early days of the American auto boom saw the emergence of over 1,900 car manufacturers. Only three survived. On January 30, 2000, seventeen internet companies each spent approximately $2.2 million on television advertising during the Super Bowl. Less than a year later, most had gone bankrupt or been sold off at bargain prices. The internet was transformative, but most of the companies that embodied it were not.

Demand, too, can put technologies to the test. The Concorde crossed the Atlantic in three and a half hours, while the market was looking for cheaper, not faster, travel. Three decades after its emergence, virtual reality remains a niche market, despite colossal investments. Bitcoin proved to be a solution in search of a problem. Conceived as a peer-to-peer electronic currency, it found its niche as a speculative asset, but its initial use as a means of payment failed. For investors, this failure of demand is crucial because it cannot be diagnosed by simply analyzing the technology. The question now facing AI is whether consumers will once again favor integrated and practical solutions over self-service coding.

The problem of circularity is perhaps most instructive in the current context. In Japan, the “keiretsu” business network system of the early 1990s created a self-reinforcing dynamic: banks financed companies within the network, which in turn sourced their supplies from suppliers within that network, who then deposited their funds with those same banks. Each company’s revenue corresponded to another’s expenses. For an investor analyzing the current AI supply chain, where chip manufacturers sell to cloud providers that host startups whose valuation depends on their cloud spending, this circularity may seem familiar. The collapse of the “keiretsu” system raises a simple question: where does the customer enter this circle?

The Fourth Revolution

If automation destroyed jobs, the most automated country in the world should be an economic wasteland. South Korea deploys more than 100 industrial robots for every 1,000 manufacturing workers , yet its unemployment rate is below 3%. While the digital revolution has certainly eliminated jobs, it has also created new ecosystems in fields such as “app developer” or “YouTuber,” which would have been unnecessary a generation ago. Even in sectors directly exposed to digital platforms—such as advertising distribution and travel bookings—employment has remained broadly stable despite post-pandemic volatility, and the average US hourly wage has followed the national trend. Technology appears to be reshaping tasks within jobs more than leading to job losses.

Technology seems to be reshaping tasks within jobs more than leading to job losses.

Will AI create more jobs than it destroys over the next decade? Two caveats are in order. First, AI is encroaching on non-routine cognitive tasks that previous waves of AI have spared. Second, even if total employment rises, the bias of technological progress toward capital exacerbates inequality and generates political risk—both relevant to portfolio performance.

The potentially most important distinction today is what’s known as the ” Turing trap “: using AI to replace human labor rather than as a tool to augment its capabilities. A hospital that uses AI to eliminate a radiologist position saves a salary. A hospital that provides its radiologists with AI tools to interpret more scans with greater accuracy achieves better outcomes, higher throughput, and diagnostic services that were previously unsustainable. History shows that the most substantial gains go to those who rethink tasks with humans in mind, not those who remove them from the loop. Microsoft’s Excel spreadsheet didn’t replace the accountant; it simply made data analysis so accessible that every business adopted the software.

History is repeating itself. In early February of this year, new AI-powered productivity tools triggered one of the sharpest sector corrections in recent memory. Software stocks, whose revenue is tied to individual licenses, plunged more than 30% in a matter of weeks amid fears about AI’s impact on white-collar jobs. Despite AI’s ability to take over some human tasks, its impact on unemployment is expected to be limited and gradual. AI has a greater impact on skilled workers in the middle and upper-middle classes, but these individuals typically possess skills that allow them to adapt as some tasks are automated or to find alternative employment as the economy restructures.

The inevitability of a technology tells us little about its attractiveness as an investment.

That said, history also reminds us that total employment can grow even as job quality deteriorates – polarization, wage stagnation and worsening inequality are not side effects of failed transitions, but features of successful transitions that public policies have failed to support.

Rather than wondering whether jobs will disappear, we should be asking where the next wave of demand will come from. Part of the answer may lie in demographics. In the United States, baby boomers and the Silent Generation hold some $110 trillion in household wealth, nearly two-thirds of the national total. The health and social care sector in the US will create 2.3 million jobs by 2033. Globally, the World Health Organization projects a shortage of 10 million health professionals by 2030. These are precisely the roles that AI will struggle to replace, and the ones an aging population will be willing to pay for.

Inevitability and investment potential

AI is undeniably revolutionary. But where does its real risk lie? Four centuries of technological evolution suggest that AI will transform employment, but not destroy it. Thus, however uncomfortable this observation may be, the inevitability of a technology tells us little about its attractiveness as an investment.

Revolutions reward the patient, diversified investor who invests against the grain—almost never those who arrive during the boom. At the heart of this upheaval, new opportunities emerge. Rather than trying to predict how AI will reshape the economy, the smart allocator recognizes that major transformations typically yield their greatest gains where no one has thought to look.