How to shield your portfolio if AI goes ka-boom

You may have a different view of the world from me — and the chances are that you do. I don’t have a television and spend most of my hours either in the woods or out at sea. The other day I saw a photo of the new Fed chair. I thought it was Novak Djokovic.

What matters when investing, however, is not your outlook. You’ve got as much chance of being right as anyone. It’s whether your portfolio is consistent with it. Do your positions align with your view?

Readers may remember me writing about this in relation to my allocation to US equities over the years — in particular when I had none. I worried at the time that no matter how much I preferred UK or Japanese stocks, owning only them did not make sense.

Because if I thought the S&P 500 was overvalued and was eventually proved correct, my other equity funds would come crashing down with it — such is the US’s heft. Being 100 per cent in shares and zero in the US was internally inconsistent.

Hence, when the prolonged bull market gave me the heebie-jeebies last year, I sold every stock I owned. It’s also why, when I decided to dive back in, I made a small purchase of US equities. So long as I owned some, I could justify my other holdings.

I remain OK with the logic. The trouble is, I still posit that artificial intelligence companies are too expensive. The related debt binge also looks familiar to many of us who have invested in booms and busts before. Ray Dalio, the billionaire founder of Bridgewater Associates, said this week that AI is a “classic bubble” and that we’re “approaching” the point where these tend to pop.

What if you agree with him? Do you have to torch the rest of your equity holdings for your portfolio to be internally consistent? Luckily, not any more, it seems. Indeed, I probably didn’t even need to.

It was different when we were emerging from the pandemic and AI was becoming a thing. Back then, even the stocks least connected to AI — your supermarkets, utilities and energy companies — had a rolling correlation with a global AI index of 0.5, according to Bloomberg data. That is, they more or less moved together.

Since that 2022-23 period, the correlation has drifted south and is now minus 0.4. In other words, when the Googles and Nvidias have dropped recently, the likes of Duke Energy and Coca-Cola have risen.

Naturally, the top quintile of names as measured by their links to AI still has a strong 0.8 positive correlation with Bloomberg’s AI index (BAIAT if you want to look it up). But the gap between them and the least-linked quintile is the widest it’s ever been.

What does this mean in real money? If we take the worst 10 per cent of days in terms of returns, the AI index fell about 2 per cent on average. Had you owned the fifth of stocks with weakest links to AI, however, you would have been flat over those days.

Thus, it is now possible for an investor who thinks AI is nothing more than dotcom for lawyers, accountants and children cheating on their homework to own a basket of companies that may offer them some protection if the bubble bursts.

Let’s be realistic, though. The down days in the calculation above are nothing like a proper walk-out-the-office-with-your-possessions-in-a-box mega bust. I was there during the 2008 and 2020 meltdowns, and I can tell you every stock was burnt.

Moreover, Dalio and others believe the pin that will make the market go pop is higher borrowing costs. Even if there is no crash for a while, this will also hurt defensive sectors such as utilities, whose valuations tend to move conversely to interest rates.

Are there other ways to shield your portfolio from an AI-mageddon? Being globally diversified helps a bit. On those worst days for AI stocks again, the MSCI All Country World Index fell about 1 per cent on average — roughly half the drop.

Every global benchmark includes the US, which has almost a 0.9 correlation with the AI index. The UK, on the other hand, is much less influenced, mainly because the few companies it has left are banks or oil producers. Bad AI days since 2022 have resulted in only a 0.3 per cent drop in the MSCI UK index (Bloomberg data once more).

Safer still is China, only falling 0.1 per cent. Malaysia, India, Australia and Europe ex-UK provide good insulation too. It reminds me a bit of a bank that I once worked for that escaped the subprime wipeout not because it was smart, but because it was so late to the party.

Japan doesn’t come out brilliantly on this analysis due to making too many semiconductors. Likewise, my overweight position in South Korea. As for Taiwan, watch out below!

If you still want exposure to the US, but don’t want a portfolio choking on food companies, another approach is to buy an equal-weighted S&P 500 fund. Or simply rebalance your AI holdings often and ruthlessly.

Another approach would be to counterweight AI with gold, cash or short-dated bonds of some description. For investors convinced that superintelligence is about to meet its kryptonite, why not make lots of money while the rest of us scream by buying put options on the Nasdaq, say, or on individual stocks? There are plenty of short or inverse ETFs available too, but I’ve written about the danger of these before.

Personally, I’ve kept it simple. A huge US underweight, a massive UK position, some bonds and loads of yen that historically soars when it’s goodbye kitty.

The author is a former portfolio manager. Email: stuart.kirk@ft.com