The Halo effect

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Good morning. The last few letters (and the letter below) have dealt with the economics of the AI future. One crucial aspect of this which has been missing from most recent discussions is retraining and career switching. It is extremely difficult, by all accounts. I’ve switched careers twice (academia to finance; finance to journalism). While I survived the switches, it was rough, and I had every conceivable advantage — education, savings, connections, portable skills, youth. Many who stand to be displaced by AI won’t be as lucky. If you have any thoughts on this economic challenge, I’m keen to hear them: unhedged@ft.com. 

The Halo effect

Regular readers will know that Unhedged loves a good acronym. So we were delighted to discover “Halo”. From the FT earlier this week: 

Utilities, energy and materials stocks have emerged as winners from the AI anxiety gripping Wall Street, as investors fleeing sectors seen as vulnerable to disruption seek businesses with tangible assets . . . 

“All these capital-light businesses that could scale historically are also the ones that could be easily disrupted,” said Guillaume Jaisson, European strategist at Goldman Sachs. On the other hand, “capital-heavy businesses are difficult to replicate, it takes time”, Jaisson said. “They are more insulated from the risk around AI,” he added, labelling the buoyant sectors as “Halo” stocks: heavy asset, low obsolescence. 

This story is intuitive. At a time when the future of feather-light businesses based on ideas — paradigmatically but not only software businesses — has been called into doubt, people want to buy something tangible and heavy. And certainly S&P 500 sector performance since late last year conforms to this general pattern. Below are the classically “heavy” sectors: energy, materials, industrials and utilities, all outperforming the index.

And here are the classically “light” technology and consumer discretionary sectors, underperforming the index:

But is the Halo story true? Are investors really chasing capital intensity, a property that for some years has been a Wall Street synonym for “low profit”? 

I have to disagree with Mr Jaisson on one thing: capital-light businesses such as software have not been easily disrupted historically. The opposite is true, which is what makes the present moment so striking. Asset-light tech businesses have various features that have made them — unlike traditional heavy industry businesses — harder and harder to displace as they grow. The canonical text on this topic, which Unhedged never tires of citing, is W Brian Arthur’s “Increasing Returns and the New World of Business”. Arthur contrasts businesses that process physical resources with those that process information, a distinction that overlays roughly, but by no means perfectly, with the asset heavy/asset light distinction. Information processing businesses enjoy increasing returns and wide competitive moats, Arthur argues, because of a combination of high development costs, negligible unit costs, network effects (information products improve as more people use them) and high switching costs (he calls this “customer groove-in”). 

It’s illuminating to think of the AI threat in terms of these features of information-processing businesses. AI machines clearly bring development costs way down (“vibe coding”, and so on). They may well bring switching costs down, too, as we move from a world of job-specific software programs and platforms to a world of general-purpose digital assistants. That is to say: we may never have a painful “switch” from one customer relationship or resource management platform to another. Instead, general-purpose AI machines simply start doing more and more things, such that the platforms become less and less useful. Legacy software will not be defeated in head-to-head competition so much as it will withers away from neglect. And once switching costs have been lowered, network effects are easier to reproduce. Information-processing businesses, in this hypothetical world, have much lower barriers to entry. 

So maybe it would be better to not say that investors are seeking asset-heavy businesses per se. Instead, they are seeking businesses that have a resource-processing element, instead of or in addition to an information-processing elements. The barriers to entry in resource-processing businesses are not necessarily that they require owning a lot of expensive assets. They are hard to get into for other reasons. Utilities, energy and materials businesses are surrounded by a thicket of permitting and regulatory barriers, for example; and manufacturing businesses entail a huge amount of practical knowhow accumulated through experience.  

Looking at the S&P 500 stocks that have done well under the recent market regime, there are a lot of resource-processing businesses that are not particularly asset-heavy, as measured by things such as the assets/revenue or capex/revenue ratios. Hotel chains such as Hilton or Marriott have long since moved to an asset-light model, but even in AI world, people are still going to need places to stay (the resource hotels process is, fundamentally, people who are asleep). Airlines, freight and logistics companies are surprisingly asset-light, relative to their level of revenues, but they move stuff around that will still need moving even when the vehicles involved are autonomous. Energy companies and utilities do own a lot of assets. But that is not the critical thing; what matters is that they deal with something other than pure information. 

One good read

Regulating private equity?

FT Unhedged podcast

Can’t get enough of Unhedged? Listen to our new podcast, for a 15-minute dive into the latest markets news and financial headlines, twice a week. Catch up on past editions of the newsletter here.