Everything About AI
NextFin News - Two hundred and fifty years after Adam Smith opened The Wealth of Nations with a pin factory, investors are gathering in the rooms where he lived and died to ask whether the father of capitalism has anything useful to say about artificial intelligence. The answer emerging from the 250th-anniversary events at Edinburgh's Panmure House is not a buy list for chip stocks, and it is not a warning against them. It is a reminder that the market's favourite stories - concentration, disruption, and the conviction that this time is different - are all older than the republic itself.
Smith published his inquiry on March 9, 1776, the same year as the Declaration of Independence. In 2026, the anniversary was marked at Panmure House - Smith's final residence for the last 12 years of his life, now part of Heriot-Watt University's Edinburgh Business School - with discussions of what the Wealth of Nations teaches about AI, government intervention, and investing. A special roundtable recorded there for a financial-news podcast brought together a senior financial reporter, Schroders' head of strategic research, and the chairman of UK natural-medicines producer Nelsons to ask what Smith would make of the forces reshaping the world economy. The question underneath the ceremony is sharper than the occasion suggests: is the AI boom a structural break in how wealth is created, or the latest episode of a very old cycle of concentration, exuberance, and dispersion?
The Pin Factory, Updated
Smith began the Wealth of Nations not with money or trade, but with a "very trifling manufacture." In a pin factory, "one man draws out the wire; another straights it; a third cuts it; a fourth points it; a fifth grinds it at the top for receiving the head." The whole business of making a pin is, in this way, "divided into about eighteen distinct operations."
"One man draws out the wire; another straights it; a third cuts it; a fourth points it; a fifth grinds it at the top for receiving the head."
Specialization, Smith argued, lifts productivity through three channels: greater dexterity, time saved by not switching tasks, and the invention of machines that abridge labour.
AI is, in this reading, the division of labour applied to cognition rather than muscle. The same three channels reappear: models execute narrow tasks with superhuman dexterity, they eliminate the switching cost between research and drafting and coding, and they are themselves the machinery that abridges mental labour. The investor's first question is not whether the technology works - Smith settled that in 1776 - but who captures the surplus. In the pin factory, the gain accrued to the master who owned the capital and organized the work. In the AI economy, the question of ownership has simply moved up the stack, from the factory floor to the model weights and the data centres that house them.
That framing matters because it separates the real productivity story from the equity-concentration story. A technology can transform output and still be a mediocre investment for anyone who buys it at the top of the hype cycle. Smith understood this distinction without ever using the word "valuation." He knew that productivity gains diffuse through competition, and that the first mover's excess profit is a rent that gets competed away.
Concentration: What Smith Would Recognize
The modern version of the pin-factory question is the Magnificent Seven. By the end of 2024, the group - Alphabet, Amazon, Apple, Meta Platforms, Microsoft, Nvidia, and Tesla - accounted for roughly one-third of the S&P 500's total market capitalization, and in 2025 they were responsible for more than 40% of the index's total return. As of August 2026 the seven companies had a combined market capitalization of about $23.7 trillion, or roughly 34% of the index. That is a concentration level with few historical peers, and it is the fact that makes the anniversary discussions feel urgent rather than ceremonial.
But 2026 has already begun to test the concentration thesis. Twelve Nasdaq-100 stocks doubled in the year, and none of them were members of the Magnificent Seven. SanDisk rose more than 400% year to date; Micron and Intel also more than doubled. Ed Yardeni put the rotation plainly in January:
"The Impressive-493 has outperformed the Magnificent-7 since last November. We expect this will continue in 2026, as last year's LargeCap laggards catch up."
Wall Street strategists at Morgan Stanley and JPMorgan have disagreed publicly over how far the semiconductor rally has run, with Morgan Stanley's chief investment officer calling the trade "meaningfully overbought" - a signal that the consensus has swung from "only the giants can win" to "the giants are cheap again."
Smith would not have been surprised by either leg of this sequence. The Wealth of Nations is built on a single observation about capital: it flows to the highest visible return, it overcrowds that return, and then it moves on. The pin factory made pins cheaper until every town had one. The AI factory is doing the same to intelligence. The investor's edge is not in identifying the technology - everyone can see it - but in identifying which claims about it are already priced in and which are not.
The Second-Order Question: Who Owns the Surplus?
The first-order effect of AI is obvious: it lowers the cost of producing cognitive work. The second-order effect is what matters for investors, and it is the one the market debates least cleanly. When the cost of a capability falls toward zero, the profit pool does not stay with the capability. It migrates to the scarce complement.
In the 1990s, the scarce complement to cheap computing was the network - the platform that sat between users and producers. In the 2020s, the scarce complements to cheap intelligence are power, silicon, and distribution. This is why the market has rewarded chip suppliers and memory makers even as some of the best-known AI platforms have lagged. It is also why a Smithian investor looks past the model to the bottleneck: who owns the thing that cannot be replicated quickly?
The scale of that build-out is itself a source of risk. Goldman Sachs Research estimates annual AI-related capital spending rising from about $765 billion in 2026 to roughly $1.6 trillion by 2031, with cumulative spending of about $7.6 trillion over the period; the bank's separate estimate of hyperscaler investment puts combined spending at $5.3 trillion from 2025 through 2030. Those figures are not a bear case - they are the price of admission for the structural thesis. But every infrastructure super-cycle in history has produced the same sequence: a genuine breakthrough, a rush of capex, a period of overbuilding, a shakeout, and then a mature industry with fewer, larger players. The railroads, fibre optics, and cloud data centres all followed it. There is no reason to assume the AI build-out is exempt from the arithmetic of overcapacity.
There is a trap here, and it is the trap Smith spent his career warning against. The belief that a new technology creates a permanent monopoly is as old as the great chartered companies he criticized in the Wealth of Nations. Smith understood security arguments for temporary protection, but he insisted such measures be kept temporary - a discipline that applies with equal force to technological advantage. Moats that look permanent in year two are often competitive in year ten. The investor who assumes otherwise is not investing; they are extrapolating.
Cyclical or Structural? The Call the Market Avoids
This is the judgment the anniversary panels circle but rarely state outright. The AI investment boom contains both a structural shift and a cyclical super-cycle, and confusing them is how capital gets destroyed.
The structural claim is strong. AI changes the economics of specialization before it changes the economics of the firm. Organizations can rely on fewer narrowly specialized employees and more broadly capable individuals whose output is amplified by models, or they can create new specializations around supervising, integrating, and evaluating machine output. Either way, the unit cost of cognition has shifted down in a way that will not revert on its own. This is a regime change, comparable to electrification or the spreadsheet - not a sentiment cycle.
The cyclical claim is equally strong, and it applies to the equities, not the technology. Capital spending on AI infrastructure follows a classic investment super-cycle, and the evidence of rotation is already visible: twelve Nasdaq-100 stocks doubling outside the mega-caps, the "Impressive-493" outperforming since November, and the migration of returns from platforms to suppliers. The correct synthesis is uncomfortable for both bulls and bears. The technology is structural; the valuation cycle is not. An investor can believe AI will transform the economy and still believe that the current dispersion of winners is a cyclical event that will mean-revert. The mistake is to let a correct view of the technology justify an incorrect view of the price.
History gives the investor a map. At the dot-com peak, the ten largest S&P 500 companies accounted for about 29% of the index's weight; by the end of 2025, the top ten made up nearly 41%, more than double their share a decade earlier. In the early 2000s, the top ten occupied roughly 20% to 25% of the index. Concentration has reverted before. It need not revert soon, and it need not revert fully - but the burden of proof sits with anyone claiming that a 41% top-ten share is a permanent equilibrium rather than a point on a cycle.
The Counter-Thesis: This Time, the Moat Is Real
The strongest argument against the Smithian reading is that AI concentration is qualitatively different from every previous technological concentration. The case runs as follows: training frontier models requires capital, data, and compute on a scale that only a handful of companies can access; the resulting models improve with scale and usage, creating a feedback loop earlier technologies lacked; and the switching costs for enterprises embedded in a model ecosystem are far higher than the switching costs between pin factories.
This is a serious argument, and it deserves more than a straw man. It is backed by the observable reality that the same seven companies have dominated the AI build-out, and by strategists who argue the mega-caps' underperformance has gone too far. But it fails on one Smithian test: it confuses scale with permanence. The great chartered companies had state-backed charters. The railroads had first-mover geography. The software platforms of the 2010s had network effects that looked unassailable until regulation, privacy changes, and new interfaces eroded them. Scale is a defence, not a law of nature.
The falsifying signal is specific. If, five years from now, the ten largest S&P 500 constituents still account for more than 40% of the index's weight while earning returns on invested capital well above their cost of capital, the concentration-is-permanent thesis stands. If that share reverts toward the early-2000s range of 20% to 25% while the broader market continues to rise, the Smithian reading - that capital diffuses and the surplus migrates to the next bottleneck - is confirmed. Watch the share, not the price.
What Investors Should Take Away
The practical lessons from Panmure House are not subtle, which is why they are easy to ignore. First, do not confuse a transformative technology with a transformative investment. Smith's pin factory transformed pin-making; it did not make every pin-factory investor rich. Second, watch who owns the bottleneck, not who owns the headline. Power, silicon, and distribution are the current scarce complements; they will not be the scarce complements forever. Third, remember that holding too much cash is itself a risk - a point the Edinburgh roundtable emphasized - because inflation and opportunity cost erode the investor who waits for a certainty that never arrives.
The anniversary speakers framed the moment in similar terms. Baroness Moyo of Knightsbridge, delivering the Adam Smith Lecture at Panmure House, said the anniversary:
"This anniversary of Adam Smith's Wealth of Nations comes at a time that we need to critically review the relevance of economics; when we must challenge and renew the ideas and ideals that have governed the global economy and the capital markets for 250 years."
The review she called for applies as much to portfolios as to policy.
Time horizons matter. In the short term, sentiment and positioning will drive the AI complex, and the rotation between mega-caps and infrastructure suppliers is likely to continue. Over the medium term, earnings will separate the companies actually monetizing AI from those merely invoking it. Over the long term, the technology will diffuse, margins will normalize, and the winners will be the firms that owned the complements at each stage of the cycle.
Three scenarios frame the path ahead. In the base case, AI spending continues to grow but at a decelerating rate, the top of the index drifts lower as a share of the market, and returns disperse across the infrastructure chain. In the upside case, a killer application emerges that re-concentrates value in the platforms, and the mega-caps reclaim leadership. In the downside case, capex overshoots demand, a shakeout hits the suppliers, and the entire complex reprices before the long-term trend resumes. Each scenario is consistent with the technology being real.
The closing judgment is one Smith would recognize. Markets do not reward those who correctly identify the future; they reward those who correctly identify what the future is worth today. The Wealth of Nations turns 250 this year not because it predicted AI, but because it described the one thing AI will not change: capital's habit of chasing a profit until the profit is gone.
Explore more exclusive insights at nextfin.ai.
