Grantham: AI turned Mag 7 monopolies into one war
Jeremy Grantham coined "unmonopoly" on June 2 to describe how AI has collapsed the Mag 7's separate monopolies into a single winner-take-all race — and argues competitive capex at $200B/year will destroy the margins that justified premium multiples, while dormant mean reversion switches back on.
Jeremy Grantham (GMO, co-founder; known for calling the 2000 dot-com bubble and the 2008 housing crash) sat down with Reuters Breakingviews editor Peter Thal Larsen on June 2 for a 44-minute conversation that produced his sharpest articulation yet of why he expects the current AI-driven market to end badly. 1 The most important new concept is one word he invented: unmonopoly.
콘텐츠 카드를 불러오는 중…
The unmonopoly thesis
For the past decade, the Magnificent 7 — Apple, Microsoft, Google, Amazon, Meta, Nvidia, and Tesla — each held a distinct global monopoly. Microsoft owned operating systems. Google owned search. Apple owned the consumer device premium. These were seven separate moats, pointing in seven separate directions. Margins were enormous precisely because competition across those seven domains was diffuse.
AI has ended that arrangement. Every member of the group is now pouring capital into the same race: who will own the AI layer of computing. That convergence is what Grantham calls an unmonopoly — the opposite of having separate monopolies is having everyone fight over one. 2
"Instead of having monopolies, you have an unmonopoly. You have a fight to the death, capitalism, red in tooth and claw, you know, this is going to be it, lots of blood." 2
The financial consequence follows directly. When seven companies with separate moats compete in one arena, margins compress across all seven. Grantham puts a number on the scale of the investment required:
"It doesn't matter if it takes me 200 billion in capex this year, I'm going to do it." 2
He is not predicting failure for the technology itself. He is predicting that the competitive structure now surrounding it destroys the above-average returns that justified high valuations.
Productivity isn't profitability
The second argument separates Grantham from simpler bubble narratives. He accepts that AI will deliver genuine productivity gains. That is actually the point.
"If everybody has a brilliant new machine, everyone will be productive. No one will make particularly good money." 2
His historical parallel: the computer in the early 1970s was a genuine competitive advantage. Within a decade, every competitor had one. The companies that adopted computers first did not capture the gains permanently — the gains became distributed across the economy as cost of doing business. Grantham argues AI is following the same path, at higher speed.
He also flags a specific vulnerability: AI leadership depends on hardware and algorithmic architectures that are themselves subject to disruption. The companies that appear to control the AI stack today have no guaranteed hold on the next architecture. In his view, markets are pricing in a permanence of advantage that the technology's own history actively argues against.
The bubble-then-inheritance pattern has precedent. Amazon fell 92% during the dot-com crash. 2 Grantham's point: Amazon eventually inherited the world, but being right about the transformative technology offered no protection against a 92% drawdown during the bubble's collapse. Investors who held through the crash still lost most of their money for years, even as the underlying thesis proved correct.
Mean reversion returns
Grantham's third argument is structural. Mean reversion — the tendency of profit margins to revert toward historical averages — worked reliably for the century before 2000. It has not worked in the 25 years since. 2
"Mean reversion worked beautifully for the 100 years prior. And it hasn't worked that well for the last 25 years. And monopoly has been one of them, the biggest reason by far." 2
His explanation: from 2000 to 2025, monopoly concentration increased in virtually every industry. Fewer competitors meant sustained above-normal returns. But AI, by forcing the largest incumbents into head-to-head competition, is recreating the competitive pressure that drives mean reversion. The very mechanism that investors assumed was broken is, in his view, being switched back on.
Without ChatGPT, he argues, the 2022 rate-hike cycle would have pushed the S&P 500 down an additional ~25%, completing a full reversion to trend. 2 AI capital expenditure was strong enough to absorb the simultaneous headwinds of a war in Ukraine, aggressive central bank tightening, and an energy shock. When that capex wave eventually normalizes — or when the competitive war erodes the margins of the companies funding it — the deferred reversion still happens.
Why you won't hear this from your broker
콘텐츠 카드를 불러오는 중…
Grantham ends with a structural argument about why large institutions consistently miss or suppress bearish signals:
"The uncertainty of long-term market moves is greater than the client's patience. That's it. Once you've got that one line in your head, you know if you're a big company, you're never going to tell your clients to get out of the market, you're always going to be bullish." 2
GMO's own track record makes this concrete: before the 2000 crash, the firm underperformed the market by roughly 6% per year for two-and-a-quarter years, and lost half its clients before the bubble burst and validated its positioning.
His parting argument is that markets are not prediction machines:
"The market is a coincident indicator. It is not a predictor of long-term streams of dividends and earnings discounted back. That's all complete nonsense." 2
For investors following this thesis, the variable to watch is margin compression among the Mag 7. Grantham's argument doesn't require the AI technology to fail — it requires the competitive investment race to generate sufficient returns to justify current valuations. He thinks it won't.
Cover image: Reuters Breakingviews "The Big View" podcast artwork, via Apple Podcasts
이 콘텐츠를 둘러싼 관점이나 맥락을 계속 보강해 보세요.