Grantham's 70% Warning Is a Serious Voice in a Thinly Sourced Debate

TL;DR

• Jeremy Grantham, GMO co-founder and veteran bubble-spotter, told CNBC on June 26 that the US stock market is the most expensive in American history, pointing to a modified Buffett Indicator of roughly 235% — well above the ~150% level that alarmed investors during the dot-com era [2].

• He predicts a potential 70% decline in AI stocks as valuations revert to historical trend, arguing that AI, however significant, will not rewrite investment fundamentals and that over-investment in new technologies has historically led to speculative collapses [1][2].

• GQG Partners manager Kersmanc separately argues the market remains 'very much' in an AI stock market bubble, countering the bull case that massive chip and hardware orders disprove bubble dynamics — those orders, he says, are part of the bubble, not evidence against it [3].

• Every major claim in this story is single-sourced across three Tier-2 outlets. No Tier-1 primary source, no independent corroboration of the 70% figure or the 235% Buffett Indicator reading, and no on-the-record response from the companies or analysts whose valuations are being questioned.

• The signal is low-confidence: the argument is intellectually serious and the voices are credible, but the evidentiary base does not yet support treating this as confirmed rather than as one prominent investor's well-reasoned thesis.

What happened

On June 26, Jeremy Grantham — the co-founder and long-term investment strategist at GMO, a man who has spent six decades telling investors they are wrong about valuations — appeared on CNBC with a blunt assessment [1][2]. The US stock market, he said, is now the most expensive it has ever been in American history [1][2]. His evidence centres on the modified market-cap-to-GDP ratio, commonly known as the Buffett Indicator, which he pegs at roughly 235% [2]. For context, that same metric was considered alarmingly high when it crossed 150% during the dot-com era [2].

Grantham's track record on such calls is not trivial. The bundle notes that he has spent six decades in this business and that his history suggests investors should probably listen [1]. He presented what he describes as a dual framework for understanding the current moment, though the reporting does not fully unpack what that framework entails [2]. The core of his thesis, however, is clear enough: valuations are at record highs, and historically, bubbles fully revert to trend. By his statistical reading of past speculative episodes, that reversion points to a potential 70% correction in AI-related stocks [1][2].

Separately, and reported by a different outlet, GQG Partners manager Kersmanc offered a corroborating view from a different angle [3]. Where Grantham focuses on aggregate valuation metrics and historical precedent, Kersmanc addresses the specific counter-argument that bulls have been making: that the enormous increase in spending on the artificial intelligence buildout — the huge chip orders, the data centre construction, the hardware procurement — is proof that there is no bubble, because real money is being spent on real infrastructure [3]. Kersmanc's response is that this spending is not evidence against a bubble; it is part of the bubble's mechanism [3]. The orders, the capex, the supply chain pressure — all of it, in his telling, is consistent with the dynamics of a speculative mania rather than a refutation of one.

These warnings arrived against a backdrop of tech stocks experiencing one of their worst weeks in a year, according to the summary framing of the story. The timing is notable: Grantham and Kersmanc are not making abstract academic arguments. They are speaking into a market moment where the stocks they are questioning are already under pressure.

What it actually means

The real story here is not that two investors said the word 'bubble.' It is that two investors with different methodologies and different platforms arrived at overlapping conclusions from different directions, and neither is making a casual or throwaway call. Grantham's argument is structural and historical: valuations are at extremes that have no precedent in American market history by his preferred metric, and the historical record of bubbles is that they fully revert to trend — not partially, not gracefully, but completely [1][2]. The 70% figure is not a guess; it is, in his framing, a statistical inference from how previous speculative episodes have resolved.

The intellectual weight of Grantham's position comes from his insistence that AI, despite its genuine significance, will not rewrite investment fundamentals [1]. This is the crucial move in his argument. The bull case for current AI valuations implicitly assumes that this technology is different — that the normal rules of mean reversion and competitive erosion do not apply because the technology is so transformative. Grantham rejects that framing. He argues that over-investment in transformative technologies has historically led to speculative collapses, not permanent value creation [1]. He draws a distinction between lasting infrastructure — the kind that retains value over decades — and today's technology, specifically citing chips that may become obsolete quickly and may not provide long-term value sufficient to justify their investment cost [1].

This is a sharper argument than it might first appear. The bull case for AI hardware spending often treats the buildout as analogous to railway construction or telecommunications infrastructure: enormous upfront cost, but durable assets that underpin decades of economic activity. Grantham is effectively saying that the analogy fails because the assets in question — semiconductor fabrication capacity, specialised AI accelerators, data centre configurations tuned to current model architectures — have a shelf life. They are not tracks in the ground; they are equipment that may be superseded by the next generation of the technology itself, potentially within a short horizon. If that is true, then the capital expenditure that bulls cite as evidence of real commitment is actually evidence of a race to deploy capital into assets that may not hold their value — which is, in Grantham's framework, precisely the dynamic that defines a bubble.

Kersmanc's contribution from GQG reinforces this from the demand side [3]. The argument that huge chip and hardware orders prove the bubble is not real rests on an assumption: that the spending reflects durable, rational demand rather than fear-driven competitive positioning. Kersmanc is saying that when every major player in an industry simultaneously orders enormous quantities of a scarce input because they cannot afford to be left behind, that is not the market functioning efficiently. That is the market in the grip of a dynamic where the spending itself becomes the symptom [3].

Hype deconstruction

This is where the analysis must be most careful, because the story's signal strength is low and the claims are almost entirely single-sourced. The 70% decline figure, the 235% Buffett Indicator reading, the characterisation of Grantham's 'dual framework' — all of these come from individual outlets reporting on a television appearance, with no independent verification of the specific numbers [1][2]. The Buffett Indicator is a publicly calculable metric, but the modified version Grantham uses and the specific 235% figure have not been corroborated by a second source in this bundle.

More importantly, this is not a prediction with a timeframe. Grantham is making a statistical argument about what full mean reversion would look like if it occurs. He is not saying the market will fall 70% next quarter, or even next year. The reporting does not specify any time horizon at all [1][2]. Readers who encounter the '70% decline' headline without that context may interpret it as an imminent forecast rather than a conditional statistical inference about the endpoint of a reversion process that could take years. That distinction matters enormously for how actionable this information is.

It is also worth noting what this story is not. It is not a consensus view. It is not the product of a broad survey of market participants. It is two voices — one very prominent, one less so — making a case that runs against the prevailing market narrative. The absence of any counter-argument in the bundle is itself a weakness of the reporting, not necessarily a weakness of the market. There is no on-the-record response from the companies whose valuations are being questioned, no alternative reading of the Buffett Indicator from another strategist, no bull-case advocate given space to rebut the bubble characterisation. The reader is getting one side of a debate presented as the story.

Finally, the framing of tech stocks having 'one of their worst weeks in a year' creates a narrative coherence that may be misleading. Grantham and Kersmanc's arguments are structural and long-term. A bad week for tech stocks is a short-term price movement. Connecting the two implies a causal or confirmatory relationship that the sources do not actually establish.

Stakeholder landscape

The most obvious stakeholders are investors holding concentrated positions in AI-related equities — semiconductor companies, hyperscale cloud operators, and the broader cohort of technology stocks whose valuations have expanded on AI expectations. For these holders, Grantham's 70% reversion thesis, if it were to play out, would represent catastrophic drawdowns. The argument that chips may become obsolete quickly [1] is particularly pointed at the semiconductor sector, where current valuations implicitly assume that today's leading products will retain their pricing power and relevance for an extended period.

GQG Partners, as an asset manager publicly stating a bubble view, has a reputational stake in being right [3]. Fund managers who call bubbles publicly and are wrong face career consequences; those who call them and are right gain enormous credibility. Kersmanc's willingness to go on the record suggests either a high conviction level or a calculated reputational bet — or both.

The companies themselves — the chipmakers, the cloud platforms, the AI model developers — have an obvious interest in the bull narrative holding. Their ability to raise capital, retain talent, and justify capital expenditure depends on investor confidence that the spending is building durable value. Grantham's argument that today's chips may become obsolete [1] is not just a valuation critique; it is an attack on the core premise that current capital deployment is creating lasting assets.

Retail investors who have piled into AI-themed ETFs and individual tech stocks are perhaps the most vulnerable stakeholders, because they are the least likely to have the risk management infrastructure to withstand a 70% drawdown in a concentrated sector. They are also the least likely to distinguish between a conditional statistical argument and an imminent forecast.

Cross-layer implications

There is a non-obvious connection here between Grantham's argument about chip obsolescence and the broader question of capital allocation efficiency in the AI buildout. If Grantham is right that today's AI hardware may be superseded quickly [1], then the enormous capital expenditure currently underway across the industry is not building a durable productive base — it is building a throwaway infrastructure that will need to be replaced, potentially at similar or greater cost, within a short cycle. That has implications far beyond equity valuations.

It would mean, for instance, that the energy and resource commitments being made to support current AI infrastructure — the data centre power contracts, the cooling systems, the semiconductor fabrication plants — are being sized for a generation of technology that may not be the generation that ultimately matters. The physical infrastructure might outlast the computational architecture it was built to house. That is a misallocation not just of financial capital but of physical capital: land, power, water, specialised labour. The cost of that misallocation would be borne across the economy, not just by the companies making the investments.

This also connects to the competitive dynamics Kersmanc is describing [3]. If every major player is simultaneously ordering enormous quantities of current-generation hardware because they cannot afford to fall behind, and if that hardware is indeed going to be superseded, then the industry is collectively engaging in a capital expenditure race that destroys value even as it creates it. The spending is real; the orders are real; the revenue for chipmakers is real. But the value being created may be ephemeral, because the assets being built have a short useful life and the competitive advantage they confer is temporary. That is the mechanism by which a bubble can have real economic activity underneath it and still be a bubble.

What this means for you

If you hold concentrated positions in AI-related stocks or technology-focused funds, the most practical takeaway is not to sell everything tomorrow — it is to stress-test your portfolio against a scenario where the sector experiences a deep, prolonged drawdown. Grantham's 70% figure is a statistical inference, not a forecast, but it gives you a concrete number to use in your own risk modelling. Ask yourself: if the AI stocks in my portfolio fell 70% and took five to seven years to recover, would that be survivable? If the answer is no, the position size is too large regardless of whether Grantham is right.

For Australian investors, the direct exposure may be less obvious but is still present. Many Australian superannuation funds hold international equity exposure that includes US technology stocks, often through index-tracking strategies that mean your retirement savings are weighted toward exactly the companies Grantham is flagging. It is worth checking what percentage of your super fund's growth or balanced option is allocated to US tech, and whether that allocation has grown passively as those stocks have appreciated — which is what happens in market-cap-weighted indexing.

If you are considering new investment in AI-themed products, the Grantham and Kersmanc arguments suggest extreme caution. The bull case — that massive spending proves the thesis — is precisely the argument Kersmanc identifies as a bubble symptom [3]. That does not mean the bull case is wrong, but it means you should not treat capital expenditure figures as independent validation of investment merit.

Uncertainty ledger

The 70% decline figure is single-sourced and presented without a timeframe [1][2]. It would change the analysis significantly if Grantham specified a horizon — a 70% reversion over two years is a very different proposition from a 70% reversion over a decade.

The 235% Buffett Indicator reading has not been independently corroborated [2]. The modified version Grantham uses may differ from standard calculations, and without knowing the modifications, the figure is hard to evaluate. A second source confirming the number and explaining the methodology would materially strengthen the claim.

The characterisation of Grantham's 'dual framework' is mentioned but not explained in the reporting [2]. Without understanding what the two components of the framework are, readers cannot evaluate the argument's internal logic.

The absence of any bull-case counter-argument in the bundle means the story is one-sided. A rebuttal from a credible strategist arguing that AI valuations are justified — or that the Buffett Indicator is structurally flawed as a valuation metric in a services-dominated economy — would provide essential context.

The relationship between the bad week for tech stocks and the bubble arguments is unclear. If the price decline is the beginning of the reversion Grantham predicts, that is one thing. If it is unrelated noise, connecting it to the bubble thesis is misleading. The sources do not establish which is the case.

Bottom line

Grantham and Kersmanc are making a serious, internally coherent argument that deserves attention — but the reporting behind it is too thin and too single-sourced to treat as confirmed diagnosis rather than informed opinion. The 70% figure is a statistical inference about full mean reversion, not a timed forecast, and readers who treat it as the latter will make poor decisions. The most defensible response is not to dismiss the warning or to act on it wholesale, but to use it as a prompt to examine your own exposure and ask whether you are being adequately compensated for the risk that two experienced investors might be right.

Sources

1. Editorial Team. (26 June 2026). Jeremy Grantham warns AI boom drives US stock market to record highs, risks historic decline. Crypto Briefing.
2. Monica L. Correa. (26 June 2026). This is 'the most expensive stock market in American history - GMO's Jeremy Grantham (SP500:) | Seeking Alpha. Seeking Alpha.
3. Tom Lauricella. (26 June 2026). GQG: We Are Still Very Much in An AI Stock Market Bubble. Morningstar.