Not all bankers get fed at the AI deal feast

The headline numbers resulting from hyperscaler capital raising are staggering and are boosting banks’ profits. But not everyone will be getting fat on the spoils

Large investment banks strive to be diversified businesses. They build sector coverage teams that cover the full gamut of the economy, from healthcare to telecoms to financial institutions (FIG) to industrials. Of course, some groups are larger or more active because they tend to have more activity and generate more fees, but the difference should be an incremental one. In other words — and I’m excluding the specialist boutiques here — you design an investment bank to deal with a wide range of clients, not a single phenotype.

However, sometimes that rough-and-ready balance becomes impossible to sustain. One sector or sub-sector grows so quickly that the equilibrium is lost. It develops a gravitational pull, sucking in huge amounts of capital and personnel. This is happening today.

The artificial intelligence business has transformed financial markets. Hyperscalers are financing the build-out of data centres at a quantum that dwarfs anything we’ve previously seen. Capital raisings that were once $1bn to $5bn now easily surpass $10bn. Forecasts for capital expenditure are into the trillions. The five biggest US banks reportedly collected $114bn in capital markets revenue in the first half of this year, up almost a third, with AI highlighted as the reason for the increase.

On the face of it, this is fantastic news. Bounteous revenue will sustain bonus pools and hiring. It should be a great year for a lot of people. However, just as the 1848 gold rush created widespread social tensions, so too the AI bonanza risks changing quite a lot within a bank, and maybe not for the better in the long term.

When we talk about investment banks, we often assume that the senior leaders have choices about where to deploy capital or staff. The reality is that they don’t have nearly as much agency as is often assumed.

They can strive for a diversified business as much as they like, but if a set of clients is launching mind-bogglingly large deals and generating record-breaking fees, they have to reorient the franchise to service those clients. And they have to do so quickly, because you can’t afford a once-in-a-generation fee opportunity to pass you by.

That means a balance sheet carefully constructed to avoid sector concentration can quickly have large allocations to one area. Senior bankers flock to this area if they can make the change. You throw bodies into the space because the last thing you want is to turn away business due to a lack of people.

In other words, senior leaders struggle to maintain any kind of strategic approach. They wouldn’t put it this way, but they are ambulance-chasing because it is so profitable to do so.

And where business flows, so does compensation. The problem here goes beyond the obvious fact that the bankers working on these mega-AI deals expect bumper pay. It’s also that they will be working to cultivate and own the key relationships and even keep others, including their bosses, as far away as they can without violating internal protocol. In other words, they will be doing everything they can to become indispensable.

Indispensability makes investment banking management that much more challenging because the ethos requires teamwork, sharing and, most of all, the institutionalisation of relationships.

This situation creates other tensions. If you’re a banker in a less fashionable sector, you will be watching your power wither as money and attention are lavished on the hot trends like AI infrastructure. You may think you’re a better banker and have a hard time accepting being effectively subordinate to others who were luckier than you.

This is all a throwback to the late 1990s, when the holy trinity of telecoms, media and technology (TMT) dominated investment banking activity. TMT was to that period what the hyperscalers are today.

TMT teams staffed up, banks competed to finance telecoms capex and underwrite dotcom equity offerings, and the bankers in these spaces demanded higher pay as they were fielding offers from competitors trying to poach them.

When the bubble burst starting in late 2000, much of that process was reversed, at times in a brutal fashion. TMT teams were hit with wave after wave of redundancies, and arguably telecoms and media coverage has never quite recovered.

Banks were forced to provision for billions of bad telecoms debt. Plaintiffs’ lawyers and regulators attacked the underwriting practices of investment banks, resulting in expensive legal settlements, fines and enforcement actions of various kinds. The ambitious Attorney General of New York, Eliot Spitzer, exposed inflammatory emails from research analysts trashing companies they were championing publicly, culminating in the global settlement that transformed the relationship between investment bankers and research analysts.

History does not repeat itself but it does rhyme. None of this is to suggest that the hyperscaler financing will blow up in everyone’s faces. The capital need is real, and no one can deny the underlying demand for computing power. The point is that the sector can become so big that it bends the rest of the organisation to service its will. It becomes too big to fail, or at least to say no to.

Those in the middle of AI-related dealmaking should be counting their lucky stars. Those covering other sectors, and who cannot pivot, might be about to find out what a virtue patience can be.