Portfolios are now alarmingly focused on one trade — AI

 Financial Times Europe 10 Oct 2026 Katie Mar tin katie.mar tin@f com

 The hot new invest ment theme was AI once — a shiny oppor tun ity for money man agers of all sizes will ing to put their cap ital to work in a trans form at ive tech no logy. It still is that, sort of, if you look past the warn ings of a robot apo ca lypse, the destruct ive “rogue” hack ing agents that look more like a fea ture than a bug and the legal risks circ ling around AI labs. (“Oops, our agents did it again” is wear ing thin as an excuse.) Des pite all that, no one can reas on ably doubt the once-in-a-gen er a tion money mak ing cap ab il it ies of the eco sys tem, from chips to data centres, and that’s the bit that pro fes sional investors are paid to think about. But the AI trade is no longer new and it is no longer “a” theme. Instead it has morphed into “the” theme — a vast, suf cat ing blob slathered on top of the global fin an cial sys tem, starving everything else of oxy gen, dis tort ing mar ket reac tions and leav ing every port fo lio, every where, alarm ingly reli ant on the for tunes of just one bet. Stocks, real estate, infra struc ture, energy, emer ging mar kets, some vin tages of private equity . . . they’re all, to a large extent, the same trade. 

The sheer dom in ance of this thing over global stocks was on emphatic dis play this week. The oil price con tin ued to smoulder at $100 or so a bar rel, and traders again fired up their warn ings that logist ical con straints in the Strait of Hor muz opened up a very real chance of a push to $200. Bond prices were stuck in the deep est rut in dec ades, with the 10-year bench mark US gov ern ment bond yield wedged com fort ably above 5 per cent. Either or cer tainly both of these things would usu ally pull stock mar kets lower. Not a bit of it. Instead, the S&P 500 index of US stocks hit a new record, as did the tech-heavy Nas daq. We’re all by now accus tomed to a little mar ket dis son ance, but this is, at a min imum, extremely odd. Sli cing up indices in weird ways to suit a nar rat ive is cheat ing. But we need to do it any way to under stand. So, yes, the S&P is at a record. But if you strip the AI com pon ent out of the index, the remain ing rump has dropped by 7 per cent since the end of August. You get a sim ilar pic ture if you look at the S&P on an equal-weighted basis, which helps to smooth out the over sized impact of enorm ous tech stocks.

The gap between the top-heavy typ ical index and its more even-handed cousin is not new, but it has now hit its widest point in nearly 24 years. Threequar ters of the stocks in the S&P fell last month — tech stocks and the energy stocks that reflect the data centre buil dout really are doing all the work. Who cares? A lot of people don’t, as long as the line on the index goes up. The prob lem is that investors are strug gling to work out how to diver sify prop erly, how to shield them selves if something goes wrong — a key part of their task in man aging other people’s money. Many are spooked by just how cor rel ated port fo lios turn out to be across a range of sup posedly diverse asset classes.

 For now, the whole world is all-in. Non-US money has fun nelled into dol lar-denom in ated assets at an extraordin ary pace over the past dec ade, and increas ingly, it is not even hedged for cur rency risk. In other words, the AI boom has almost by acci dent become a major prop not just for the US stock mar ket but for the dol lar too. Without ques tion, US cor por ateearn ings, not only in tech but espe cially in tech, jus tify the excite ment. If any thing, US stocks are pretty cheap right now— stock prices have not kept pace with spec tac u larly rapid earn ings growth, seem ingly held down in part by rising bond yields.

 But we still, col lect ively, have a very poor read on how AI will pan out. To some extent, every one is guess ing here, but private equity firm Bain said in a recent report that annual spend ing on AI infra struc ture could reach $1.5tn by 2031. Extraordin ary stuff. Even more extraordin ary, though, Bain added that “sus tain ing this level of invest ment would require an AI mar ket approach ing $6tn annu ally”. Really? An industry worth one-and-a-half times the UK’s entire annual eco nomic out put every year? This is strug gling to pass the smell test.

 Poli cy makers seem uncer tain what they are deal ing with here too. US Fed eral Reserve chair Kevin Warsh has been of the view that the pro ductiv ity stem ming from the use of AI will, fairly quickly, pull down infla tion. But oth ers at the cent ral bank, recently includ ing Mary Daly and Lisa Cook, are veer ing towards an oppos ing view, that the build-out poses an upward risk to infla tion in the months ahead. So we’re left with a strik ingly twospeed US stock mar ket beloved of pretty much every investor on the planet, dom in ated by an AI build-out that is, in prac tical terms, nearly impossible to hedge or avoid.

We don’t really under stand the longterm eco nomic implic a tions, but the under ly ing com mer cial assump tions seem rather heroic. And we’ve all, delib er ately or oth er wise, hitched our for tunes to the idea that it will work out just fi er be right.