Why Trading Firms Trade at Such Low Multiples (and Why That's About to Change)

Trading businesses generate real returns, and the market has still tended to undervalue them. Let me give you my thoughts on why that might be. My sense is that it comes down to the economics having always favored the talent over the platform.

For years the market has paid software companies many times the multiple on trailing earnings that it pays trading firms with comparably consistent earnings. I suspect that gap rests on two assumptions, and I think both are worth questioning.

The market perceives trading revenues as low-quality earnings, so they get discounted even when the best firms produce them with real stability. The market also assumes that trading depends on human capital that is expensive, scarce, mobile, and prone to walking out the door, which means profitability erodes over time.

I expect the market to reprice both assumptions, helped along by AI but not driven by it alone.

The large multi-strats and systematic market makers already make the first point for me. Millennium, Citadel, Jane Street, and XTX have shown that stacking enough uncorrelated sources of alpha produces income that is anything but low quality. And AI is starting to move the second one. It has the potential to shift alpha generation and risk management from a human operating expense to a capital expense. You invest in systems instead of hiring more PMs and traders.

You can already see it. Jane Street, Hudson River Trading, and XTX have announced AI capabilities that would make a frontier lab jealous, and they have done it without burning venture money. They are privately held and self-funded.

I think investors should rethink how they price this kind of earnings. Let me take the two assumptions in turn.

The human capital problem

Trading firms are people businesses. Alpha, research, execution, compliance, all of it runs through human beings who are expensive, mobile, and fully aware of what they are worth. When the firm's edge can quit, buyers discount accordingly. And they are right to.

This plays out differently depending on where you sit in the industry, but the underlying logic holds everywhere.

The multi-strat problem

The multi-strat model is elegant on paper. Combine dozens or hundreds of genuinely uncorrelated alpha sources, run tight risk management across all of them, and you get stable, high-quality returns that hold up across cycles. It is a compelling pitch, and for the biggest players it has worked.

The trouble is what it costs to run.

Every uncorrelated alpha source needs a portfolio manager and a team. Those PMs know they are rare, they know what their returns are worth, and they price themselves to match. The market for that talent has gotten extraordinary. Even when you hire well, the cost of maintaining 50 or 100 independent pods is enormous, and it sits on top of returns before a dollar reaches an LP.

Bobby Jain's launch illustrates the point cleanly. The former co-CIO of Millennium understood the model from the inside as well as anyone alive, and he raised roughly five billion dollars, one of the largest hedge fund launches ever. If the model could work at that scale with that pedigree, it would have. What he ran into instead was that assembling the talent to generate truly uncorrelated alpha eats the economics. The competitive market for portfolio managers had simply gotten too expensive to absorb at that size.

To make the model sustainable you need tens of billions under management just to spread the fixed cost of talent across a base large enough to carry it. Below that threshold the people eat the returns. Most of the value the firm creates gets captured by the individuals generating it, not by the platform or the LPs. That is not a knock on the people. It is just how the math works when your primary asset wears shoes and has a phone full of competing offers.

The bank reality

Banks come at the same problem from a different angle, with a different set of costs stacked on top.

A major sell-side FICC or equities division carries heavy front-office expense. Traders, quants, salespeople, and execution staff all command strong comp, and that is before the middle office, back office, compliance, legal, and regulatory functions sitting underneath the revenue.

The back office inside a big bank is more manual than outsiders tend to appreciate. As recently as 2019, PMs were still faxing trade confirmations so back-office teams could book them correctly. That work ran through large teams across multiple offshore centers. Cheap per person, not cheap in aggregate. Add thousands of back and middle office staff plus the infrastructure they need, and the overhead gets substantial.

The result is a structure where real front-office revenue gets eroded by the fixed cost of running an institution that can satisfy its regulators, reconcile its books, and manage risk across a global footprint. The P&L at the top line is distributed across a very large cost base before anyone calls it profit.

The banks know this. They have invested in automation for decades. But they layered those investments onto legacy systems built for a different era, and the pace of change has been slow relative to the pace at which costs have grown.

The boutique alternative

For smaller firms that cannot compete on the multi-strat model and do not have a bank's balance sheet, the traditional answer has been to go narrow. Pick a niche, build genuine expertise, and sell that expertise as a differentiated product.

It can work. A firm that deeply understands distressed debt, or structured credit, or the technology supply chain can generate real alpha in its lane. The research is valuable, and the expertise compounds.

The problem is that niche expertise creates its own fragility. When your theme is in favor, returns look excellent. When the macro turns against it, there is no diversification to fall back on. A fund built around the AI value chain will have a strong year when AI is the dominant narrative and a hard one when it is not. The expertise is real, but the return profile is cyclical, and that makes life difficult for allocators trying to build stable, diversified portfolios.

So the boutique path solves one problem and creates another. It avoids the talent-cost spiral of the multi-strat model and trades it for concentration risk and cycle dependency.

What changes with AI

The common thread across all three models is that scaling has always meant spending more on people. More pods, more back-office staff, more specialized experts. Getting broader or deeper has always cost headcount.

AI changes that relationship.

When alpha generation, research, strategy optimization, and back-office workflow can run on scalable systems instead of teams of specialists, the cost structure of a trading business starts to look different. Fixed technology cost replaces variable headcount cost. The firm can move into new strategies or markets without standing up a new pod. Compliance and reporting that once needed large offshore teams can run on agentic systems with better data quality and faster turnaround.

The more interesting part is diversification. A system that can learn the structure of a new market and apply proven methods to it does not need a fresh team of domain experts every time it expands. That changes the calculus for smaller firms trying to compete with the multi-strats on breadth without matching them on headcount.

I am not claiming AI makes trading easy or that edge stops mattering. The firms that benefit most from this shift will be the ones that pair genuine domain expertise with the ability to build and run these systems well. That combination is rare. You have to understand how capital markets actually work and have the judgment to build AI systems that hold up in adversarial conditions, and that takes having lived in both worlds. A pure AI-native team without market scars will not get there, and neither will a traditional shop that bolts AI tools onto a legacy stack.

But for the firms that can put both together, the valuation math starts to move. A trading business that scales through technology instead of headcount, with costs that do not grow in step with revenue, looks a lot more like a software business than a traditional fund. The multiples that come with that comparison are considerably more attractive.

The old assumption that trading firms are inherently low-multiple businesses was always a statement about their cost structure, not their returns. That assumption is getting less reliable by the year, and the firms building accordingly are going to be valued on very different terms.

If this message resonates with you and you would like to learn more about what we do, we would welcome the conversation.