This week, the World Uncertainty Index hit 106,862, the highest level ever recorded. It reminded me of a line from Elroy Dimson, a Cambridge finance scholar best known for his work on longterm investment returns: Risk means “more things can happen than will happen.”

 

Risk, in other words, isn’t just what you can measure. It’s what you forgot to imagine.

 

Right now, we’re very good at imagining what can go wrong with AI. That’s easier than imagining what might go right. Humans are wired that way. Negativity bias is an evolutionary feature, not a bug (see the chapter “The Seduction of Pessimism” in The Psychology of Money by Morgan Housel). We overreact to bad news, shoot first and ask questions later.

 

But commoditisation (of cognition) is not the same thing as collapse. More often, it’s how technology lowers costs, expands access, raises productivity and, over time, creates entirely new categories of consumption and economic activity.

 

The market, however, doesn’t seem ready for that yet.

 

Disruption Cuts Both Ways

 

Nor, it seems, is the market ready for the idea that if Anthropic’s latest model materially raised confidence in AI’s disruptive power, then the suppliers of that AI — and the owners of those suppliers — should be the beneficiaries.

 

Microsoft owns roughly 28% of OpenAI. Amazon owns a significant stake in Anthropic. Yes, for some businesses this moment is existential. But not for all.

 

Excellence requires adversary. Take Walmart. There was a time when ecommerce was meant to destroy it. Today, Walmart commands a ~$1 trillion market cap and trades at over 40x earnings, a ~60% premium to its principal disruptor, Amazon.

 

Companies facing disruption don’t quietly wait to be wheeled out to the hearse. They reassess, respond and often come back fighting.

 

As investors, we have to do the same. That means resisting instinct, fighting emotion with logic, identifying the constants and asking the hard questions. The answers may not help us predict the future, but they may help us prepare for the shocks still to come.

 

With that in mind, here are seven questions I’m grappling with.

 

1. Diversification or Doubling Down?

 

The Weekender has discussed the merits of markets and sectors beyond technology for years. A mean reversion from AI to its supply chain — from intangible to tangible — felt overdue. At one point, you could buy all listed miners, plus years of future copper and seaborne ironore supply, for a fraction of Apple’s market cap. That was an intriguing setup.

 

Similar comparisons existed between the FTSE 100 or TOPIX (Tokyo Stock Price Index) and Nvidia.

 

Valuations matter.

 

When we first mentioned the FTSE, it was the leastfavoured asset in the market (per the Fund Manager Survey), with cash-covered dividend yields comparable to private equity distributions to paid‑in capital (DPI) that year. Today, it’s among the more favoured. Like many assets, it’s moved from contrarian to consensus.

 

Big regime shifts can last years, and I’m happy to stay on the ride. But it’s critical to understand the risks. Many nonUS largecap tech sectors are also levered bets on AI — sometimes even more so, particularly where capital intensity is high.

 

If you’re seeking regional diversification from US megacap tech, you won’t find it in Taiwan (due to its high concentration in semiconductor company TSMC), South Korea (Samsung, SK Hynix), China’s “new productive forces,” or even necessarily in energy, materials, or industrials — the very gating constraints of AI.

 

As ever, diversification starts with defining the risk you’re trying to mitigate. If that risk is AI itself, it’s not obvious that regional exposure diversifies you — it may simply double you down.

 

2. What Could Kill AI?

 

We could.

 

We’ve previously compared this period to the dotcom era and concluded that even if this is a bubble, it’s still early. Manias don’t end in scepticism or fear. They may end in euphoria, new paradigms, and aggressive reratings.

 

But there is one difference that unsettles me: AI is not popular.

 

In the 1990s, roughly twothirds of survey respondents viewed the internet favourably. Today, fewer than half feel that way about AI — fewer still trust it.

 

The media is saturated with headline risks: Citrini Research’s report “The 2028 Global Intelligence Crisis,” Matt Shumer’s essay “Something Big Is Happening,” the widely viewed AI Confidential with Hannah Fry on BBC Two; and the case of Jaswant Singh Chail, where an AI companion chatbot reinforced delusions ahead of his attempted assassination of the Queen of England.

 

These stories hit the amygdala. Fear triggers flight. Sell first, ask questions later. And sometimes, we — the humans — push back against the machines. In protest.

 

There are roughly 3000 data centres under construction in the US. Not all will be completed — not without a fight. Activist groups have already blocked an estimated $64 billion of projects. The Financial Times has described a “revolt in Maga heartlands,” where residents see data centres as giant robots that make locals poorer and Californians richer.

 

AI is becoming political.

 

And politicians, when votes are at stake, tend to listen. Left unattended, this creates friction and slows progress. Regulation would certainly do that, though, as history reminds us, it can also act as a catalyst, just as the Telecommunications Act of 1996 did for fibreoptic investment.

 

But I doubt we get there without a charm offensive first: one that highlights AI’s positives and reframes adoption as a strategic imperative — the need to win at all costs against our adversaries, whoever they now may be.

 

Whatever the outcome — protest, progress, or regulation — there is one rather comforting realisation: for all these machines’ capacity to learn, AI isn't casting votes itself.

 

Yet.

 

3. What’s the Best Hedge Against Capex Cuts?

 

Suppose the voting public — us — succeeds in slowing AI’s pace. Capex growth slows. What then happens to the AI trade?

 

Markets have derated companies for rising capex, focusing on lower nearterm cash flow and reduced shareholder returns — even when that capex is largely backlogdriven. If capex slows, cash flow rises, capital intensity reverses, and buybacks return.

 

Paradoxically, the best hedge against capex cuts may be owning the providers of that capex — the hyperscalers themselves. The very assets investors are rushing to diversify away from.

 

Said differently, could the Magnificent Seven begin to behave defensively? On valuation, some already do. Several trade on lower multiples than traditional defensives like Staples and Utilities.

 

That seems a little odd. At least to me.

 

4. Is the US Market Now RangeBound?

 

Another Dimson lesson: over the very long run, most equity returns come from dividends and real growth, not multiple expansion.

 

Valuations have life cycles. After secular bull markets — when both earnings and multiples rise — markets often enter rangebound phases. Prices go sideways while multiples compress. The period from 1968–1982 is the textbook case.

 

The mistake is confusing the trend of the system with the opportunities within it.

 

History shows that even in rangebound markets, certain characteristics can outperform. Earnings growth matters more than multiple expansion, for example. Margin of safety matters more than momentum. Quality and value become critical factor traits to consider.

 

So what could drive a derating today, and provide more structural reasons alongside the fact mid-term election years tend to be poor for markets?

 

Three themes dominate client conversations:

1. Mix shift: from assetlight to assetheavy businesses, warranting lower multiples.
2. Deequitisation reversing : fewer buybacks (capex) and more mega IPOs erode the scarcity premium.
3. Disruption risk: especially on longduration assets, where most of the value lies in future cashflows (think software companies moated by intangible assets)
    

5. Could the Next Wealth Shock Come From Higher Earners?

 

Much attention is paid to the lower leg of the Kshaped economy. But what if the real risk lies in the upper leg?

 

The transmission mechanism this time may run through the top decile of US households,  those who drive nearly half of consumer spending, own most equity wealth, and hold disproportionate exposure to private equity, venture capital and crypto.

 

They are also the cohort most exposed to AIdriven pressure on labour income, job security and asset values — simultaneously.

 

What began as wealth concentration could end as businesscycle concentration if AI delivers on its promise.

 

6. Will Bank Deregulation Bring Competition to Private Credit?

 

And if it does, will the same return premium apply to a sector that charges investors for the right to change their mind?

 

The credit default swap (CDS) market doesn’t seem convinced.

 
(While on banking regulation, Michelle Bowman’s speech on mortgage market reform is worth attention. She’s asking the right questions: how to support onbalancesheet bank lending and potentially reverse the steady migration of mortgage activity to nonbanks over the past 15 years. Assuming a stable Treasury market, this feels like a sensible regulatory conversation — one that could ultimately improve mortgage availability).

 

7. Is Crypto Still Coming?

 

Crypto has gone quiet. Winter does that. But quiet isn’t the same as irrelevant.

 

The next catalyst may be the Clarity Act, alongside the Genius Act, which would bring regulatory framing to digital assets, particularly stablecoins. Prediction markets currently assign roughly 70% odds to passage.

 

US deposittakers are lobbying to cap stablecoin yields to protect deposits. They’ll likely succeed. The tradeoff may be allowing nonUS issuers to offer yield — potentially disintermediating local banks while accelerating stablecoin adoption as both a payments and savings instrument.

 

Usage is already surging. In January alone, stablecoin turnover accounted for nearly onethird of last year’s total volume. Circle, the secondlargest issuer, recently reported 70% year-over-year revenue growth — numbers unlikely to trouble the US Treasury, given stablecoins are typically backed by Tbills.

 

And bitcoin? It may benefit too. Senator Cynthia Lummis (R-WY) has been clear: once the Clarity Act passes, the idea of a Bitcoin Reserve is back on the table.

 

Spring may be coming.

 

And so…

 
I don’t know the answers to these questions. But risk has never been about predicting what will happen — it’s about preparing for what we failed to imagine. And right now, these feel like the questions that matter.

 
I would love to know your answers.

 
Have a great weekend.  
 
Gary