Yesterday’s post drew a wider response than I expected, and I am grateful for it. Most readers found the argument thought-provoking and timely; a smaller number found it unnecessarily alarming and short on substance. Both reactions are fair, and I want to address the second group directly rather than let the disagreement sit unanswered.
When the numbers stop explaining themselves
Every investing career eventually runs into a stretch where the spreadsheet stops being a reliable guide. Valuation models still produce an output, but the confidence behind the output erodes, because the inputs themselves — growth rates, discount rates, terminal assumptions — are being asked to do work they were never designed for. In those stretches, experienced investors tend to fall back on something closer to pattern recognition than arithmetic: a sense, built from having sat through a few cycles, that the character of the market has shifted even before the numbers confirm it.
My sense is that we are in exactly that kind of stretch. It is not a single data point driving that view; it is the accumulation of several stress signals appearing at the same time, each individually explainable, but collectively unusual.
The signals worth naming
Fiscal sustainability. In several major economies, government deficits and debt-servicing costs have reached levels that would once have triggered a sharp market response. Instead, they have been absorbed with surprising calm, largely because interest rate cycles and demand for sovereign debt have so far cooperated. That cooperation is not guaranteed to continue indefinitely.
Sticky inflation. Headline inflation has come off its post-pandemic peaks, but the underlying components — services, wages, shelter — have proven far stickier than central banks initially projected. This matters because it narrows the room policymakers have to cut rates aggressively if growth weakens, which is precisely the tool that has rescued markets in past downturns.
Stagnating growth. Several developed economies are showing growth rates that are positive on paper but weak beneath the surface — propped up by government spending, a handful of mega-cap companies, or one-off factors, rather than broad-based private investment.
Stretched growth valuations. A narrow set of companies, heavily concentrated in AI-linked infrastructure, is being priced as though its current growth trajectory extends cleanly for a decade or more. Multiples of this kind have historically required near-perfect execution to be justified, and near-perfect execution is a high bar over a long horizon.
Rising credit stress. Pockets of the credit market — leveraged loans, private credit, and select corporate sectors — are showing early signs of strain: widening spreads, downgrades, and refinancing difficulty. Credit stress is usually a leading indicator rather than a lagging one; it tends to show up before equity markets acknowledge a problem.
Geopolitics and trade. Tariff uncertainty, shifting alliances, and logistics disruption are adding a layer of unpredictability to corporate planning that did not exist to the same degree a decade ago. Supply chains built for efficiency are being rebuilt for resilience, and that transition is expensive and slow.
Volatility across unrelated assets. Semiconductor equities, silver, bitcoin, and the SpaceX IPO have each delivered sharp, painful drawdowns in recent months. What is notable is not any single move, but that the losses have appeared across asset classes that are not normally correlated — a sign that liquidity and risk appetite, not asset-specific fundamentals, may be the common thread.
AI’s unresolved business model. Artificial intelligence has genuinely disrupted several legacy industries, and that part of the story is real. What remains unproven is whether the AI buildout itself — the capital expenditure, the compute, the infrastructure — will generate a return that justifies its cost on any reasonable timeline. Disruption of others’ business models is not the same as having a durable one of your own.
The balloon in the Dark Room
None of these signals, on its own, is a crisis. That is exactly what makes the moment harder to read, not easier. It feels like a group of persons standing in a dark room holding pins in their hands, knowing there is a balloon somewhere above, without knowing which pin — or whose hand — will find it first. The probability that no pin ever finds the balloon is not zero. But it is not the base case either, and treating it as the base case is where I think the complacent reading of this market goes wrong.
Why this could be a combination, not a repeat
Markets rarely repeat a crisis exactly; they tend to borrow ingredients from more than one. My working view is that if a serious correction unfolds from here, it is more likely to resemble a blend of three earlier episodes than a clean replay of any single one.
2008 and 2000, combined. The 2008 crisis was fundamentally a credit crisis — leverage built on assets whose value was assumed rather than tested. The 2000 crisis was a story of capital pouring into business models that had not yet proven they could generate durable cash flow. The current AI investment cycle carries elements of both: enormous capital expenditure, increasingly financed with debt and vendor-financing arrangements, directed at business models that remain commercially unproven at scale.
1997 and 1998, revisited. The Asian financial crisis and the collapse of Long-Term Capital Management were both, at their core, stories of a strengthening US dollar exposing leverage and mispricing elsewhere in the system. A materially stronger dollar today would put similar pressure on precious metals and crypto assets, both of which have attracted significant speculative positioning over the past two years.
1987, echoed. Black Monday was less about a single fundamental trigger and more about mechanics: rising rates, unresolved trade and currency disputes, stretched valuations, and — critically — automated trading systems that amplified a decline rather than cushioning it. The scale and speed of algorithmic execution in markets today is far greater than it was in 1987, which means the mechanical amplification risk, if anything, is larger, not smaller.
A downturn that borrows credit stress from 2008, unproven business models from 2000, dollar-driven deleveraging from 1997, and mechanical amplification from 1987 would not look identical to any one of those events. It could plausibly be sharper and more disorienting than any of them individually, precisely because it would be arriving from several directions at once rather than one.
What the twelve months before every crash have in common
It is worth revisiting how each of these episodes felt in the twelve months before they began, because the emotional texture tends to rhyme even when the underlying causes differ. In 1996, 1999, and 2007, the dominant mood was not fear — it was irritation with anyone who raised the possibility of a top. Alan Greenspan’s irrational exuberance remark in December 1996 was widely dismissed at the time, and the market ran for several more years before the reckoning arrived. The same pattern held in 1999, and again in 2007, when concerns about housing and credit quality were treated as the fringe view rather than the mainstream one.
I do not think 2026 sounds meaningfully different from those years. The specific worry differs each cycle, but the response to the worry — dismissal, followed by irritation, followed eventually by recognition — has been remarkably consistent. That consistency is itself part of the argument, not a footnote to it.
A closing note
None of this is a prediction with a date attached, and it should not be read as one. It is a personal, opinion-based on reading of conditions that I believe warrant more caution than the current mood in markets reflects. Markets can, and often do, remain irrational for longer than any individual view accounts for, and I could simply be wrong. Regardless, I would like to be cautious and brace myself for the “reset”. Also, I completely respect the disagreement.