Late-Cycle Investment Theory: Q3 2025 in 10 Snapshots
From GPU traders to keiretsu partnerships, ten signals that we're living through the 1970s of tech
Late-cycle investment theory suggests that we are living through the 1970s of tech.
Back then, the age of oil, automobiles, and mass production was reaching maturity, putting unbearable pressure on national economies that had expanded rapidly since the Second World War. The shift towards slower growth and stagnant markets pushed institutions—the Bretton Woods system and the welfare state—close to breaking point. To sustain growth for a few more decades, new approaches such as lean production, financialisation, and later free trade had to be embraced.
Today, it is another age—that of semiconductors, computing, and networks—that is reaching maturity. Most of the winners, the Microsofts, Googles, and Metas, emerged during the previous phase of synergy. Now they are the ones allocating vast amounts of capital to the lean production of our time: artificial intelligence. And as in the 1970s, we face a transition towards slower growth and stagnant markets, which puts institutions under severe strain. That said, I’m confident we will again find levers to perpetuate growth, and just as the 1980s and 1990s followed the 1970s, we may eventually see a new era of economic expansion. What form that will take, however, remains unclear.
Now that we are well into October and Q3 2025 is behind us, it is time to revisit the late-cycle investment theory. As before, I will do so by providing ten snapshots of developments that align, at least in part, with the late-cycle framework. As always, I welcome your feedback.
1/ Market concentration reaches historic levels
We keep being told that the US economy is doing well because the stock market keeps rising, regardless of the terrible events happening in the world and the underwhelming state of the global economy. But there are ways to relativise the impression given by this single economic indicator:
First, the stock market may be climbing, but the dollar is weakening. That means U.S. assets are not performing particularly well from a foreign investor’s perspective.
Second, much of the stock market’s growth is driven by passive investing. A recent study found that a $1 inflow into passive funds can create an additional $5 in market value. For now, passive investors such as BlackRock and Vanguard are mainly adding funds, but this could reverse as large generations reach retirement, turning inflows into outflows—with potentially serious consequences.
Third, the stock market is often discussed as if it were homogeneous, when in fact the gains are concentrated in a handful of companies. The so-called Magnificent Seven, together with Broadcom, now represent 37% of the S&P 500.
Where is this headed? Four ideas:
First, stock market concentration is not new. The last time it happened was during the Nifty Fifty era, when the post-war boom highlighted clear winners in the age oil, automobiles, and mass production. Investors then had no better option than to focus almost exclusively on these blue-chip, highly profitable companies.
Second, this echoes Jerry Neumann’s observation that investing in the late cycle is difficult because outcomes are so predictable. We know Nvidia is dominating GPUs. Microsoft, Google, and Meta will endure. Apple, despite challenges in China, will remain strong because iPhones and MacBooks are not going anywhere. Where do you find alpha in such a landscape?
Third, this concentration will not last forever. The Nifty Fifty eventually came back to earth during the 1973–1974 crash, triggered by rising inflation, the oil crisis, unreasonable valuations, and an economic recession. Could we be seeing a similar context today?
Finally, as Gita Gopinath notes in The Economist, the consequences of a market correction today could be far more severe than during the Nifty Fifty or dot-com crashes, because a much larger share of global capital is now invested in US stocks:
Over the past decade and a half, American households have significantly increased their holdings in the stock market, encouraged by strong returns and the dominance of American tech firms. Foreign investors, particularly from Europe, have for the same reasons poured capital into American stocks, while simultaneously benefiting from the dollar’s strength. This growing interconnectedness means that any sharp downturn in American markets will reverberate around the world.
2/ China weaponises rare earth dominance
The main headlines over the past few days have focused on China imposing a stringent export control regime on rare earth materials—a segment in which Chinese players have patiently built a dominant position over the past few decades.
“Rare earths” refers to a group of soft, heavy metals whose properties make them essential for certain electrical and electronic components, notably catalysts and magnets. These, in turn, are found in products as widespread as batteries, electric vehicles, wind turbines, weapons systems, and consumer electronics such as smartphones. In other words, rare earths are everywhere, and much of modern technology depends on them.
Contrary to what the name suggests, rare earths are not actually rare. They are relatively abundant but occur in very low concentrations, which means that transforming them into usable materials requires extensive and costly mining and refining operations. What China controls is not access to the raw resources but the capacity to refine and process them for industrial use.
As Yanmei Xie noted, commenting on an FT article published in August, it makes little sense for any Western private company to try to compete with this dominance. Only sustained state support could make such an effort viable over the long term—not only to bridge the path to profitability but also to shield operations from potential Chinese sanctions under the new regime. In the meantime, China has decided to play its hand, effectively doing for refined rare earths what the United States did under Biden for silicon chips, which are also made from an abundant material: sand.
This move not only takes Trump’s trade war in a new direction but also highlights the breakdown of international institutions and the resulting fragmentation of global markets. That fragmentation, in turn, creates opportunities for a new generation of intermediaries and traders to emerge.
Just as the nationalisation of oil reserves in the 1970s gave rise to Marc Rich, the inventor of the oil spot market, today’s export controls on silicon chips and rare earths may give birth to a new generation of commodity trading empires—the Glencores and Trafiguras of the modern age. These traders would act as crucial brokers of access and liquidity, navigating political barriers, securing alternative supply routes, and keeping critical industries supplied in an era of restricted flows.
3/ GPU trading creates new intermediaries
Another market that appears ready for the rise of new intermediaries in the mould of Marc Rich is that of GPUs—the dedicated chips, mostly made by Nvidia, used first for training AI models and then for processing inferences as millions of users access them.
The way the market works today is simple but inefficient. Companies that decide to build capacity often raise capital from the private credit market, place large orders with Nvidia, wait for delivery, and then install the GPUs in data centres that require vast amounts of energy. This setup is suboptimal in several respects. Capacity is being built without clear visibility on future usage. It is being built without securing the necessary energy supplies near data centres. The system favours large buyers with the ability to raise capital, thereby limiting opportunities for new entrants. In turn, this makes Nvidia highly dependent on a small group of customers, creating mutual reliance that constrains flexibility in the value chain.
To address these inefficiencies, some players are already exploring ways to create a market for trading GPUs on the spot—much as traders did with commodities in the 1970s. In time, such a market could enable the development of sophisticated financial products such as futures contracts, improve long-term matching of supply and demand, reduce the risk of bubble bursting, foster competition across the computing and energy stack, and further consolidate Nvidia’s market power.
This, to me, is another instance of the Marc Rich theory at work—especially since GPUs, as a specialised category of silicon chips, are also subject to export controls.
And I did not come up with the idea myself. I stumbled upon it twice within a short period, which is usually a sign that something is indeed taking shape—once in an essay by Felix Salmon in Bloomberg and again in an Odd Lots podcast featuring trader Don Wilson.
4/ AI suppliers form manufacturing partnerships
In an edition published earlier this year, titled Who Will Be the Japan of the AI Era?, I first argued that AI is essentially an efficiency innovation.