The 4,000 stock universe most fund managers can't fully cover - and how Goldman Sachs does

This interview was filmed on 25th March, 2026. 

Small companies are exciting for several reasons that appeal to human nature - they're often innovative, riskier, and have the ability to become the Next Big Thing.

But that gloss can quickly rub off when the lack of coverage and the enormous number of stocks out there are enough for any investor to lose steam.

We sat down with Dennis Walsh, who is the Global Co-Head of Quantitative Investment Strategies at Goldman Sachs Asset Management and co-manages the Yarra Global Small Companies Fund, to share his insights into the often under-researched and widely-dispersed global small cap universe. 

By using a combination of a tried-and-tested investment framework, Goldman Sachs' data infrastructure, and constantly evolving AI tools that allow real-time coverage of about 4,000 small-cap stocks globally, Walsh unpacks how that translated into outperformance of 9.34% above the MSCI World Small Cap Index over the year to end of February.

Watch the interview above for all of the insights or read the summary below. 

INTERVIEW SUMMARY

The case for global small caps right now

The commonly held view is that small caps are riskier than large caps. Walsh doesn't entirely disagree, but argues that the picture is more nuanced at the portfolio level.

"You can actually reduce the total volatility of your portfolio with some allocation to small caps because they're not going to be perfectly correlated with the large cap parts of the market," he says. 

The excess returns on the small-cap side will also be "almost completely uncorrelated" with the large-cap side, a diversification benefit many investors overlook.

There's also a structural argument about where AI disruption risk actually sits. The companies most at risk tend to be the big, asset-light, high-valuation names. 

Small caps, by contrast, tend to be physical, hard-to-replicate businesses in traditional industries - what Walsh calls "HALO companies" (heavy assets, low obsolescence) that are far less likely to be made obsolete by AI technology, as seen in the pockets of opportunity that are opening up at the moment.

Where the opportunities are right now

Walsh sees two of those pockets in global small caps right now. In Japan, the reflation trade is pushing up smaller companies, especially industrials and technology names tied to AI spending that haven't yet been fully priced in.

In Europe, government stimulus is lifting old-economy infrastructure, construction, and engineering companies that have been out of favour for years.

"We think there's still a significant runway for those businesses going forward," Walsh says.

How the system works

The fund's process runs on two broad categories of signals: fundamental and technical.

On the fundamental side, the team looks for high-quality companies at attractive valuations. 

"What's a little bit different about our process is how we define what it means to be a high quality company," says Walsh. "We use a large number of data sets to form a contemporaneous view of the world. And then based upon that view of the world, get a sense for what companies are likely to benefit from the world as it exists today, rather than how it existed over the previous four quarters."

The same goes with valuations. Rather than applying the same ratios across every stock, the team uses metrics tailored to each company's business model and peer group to get a more accurate read on true intrinsic value.

On the technical side, the team tracks sentiment from company management, sell-side analysts, and both institutional and retail investors. 

The other key input is themes - such as understanding in real time which global dynamics are driving returns, which companies have exposure, and whether that exposure has been priced in yet.

"We look for companies that may be tangentially related to those themes, but have yet to be recognised by the market," Walsh says.

What AI is changing

Walsh also explains that their core investment approach hasn't changed, but rather what has changed is how well and how quickly it can be executed.

Using machine learning models that are, in Walsh's words, "incredibly compute intensive and require access to very powerful GPUs at scale," the team processes thousands of data sets daily of text, images and audio.

 "The efficacy with which you can capture an idea has increased pretty dramatically," he says.

The biggest gain through using the models is in understanding context and its ability to read between the lines of earnings calls, regulatory filings, and industry data the way a seasoned analyst would. 

"Not only at a human level, but a human that is very much a professional, experienced financial analyst," says Walsh. "That's incredibly powerful, and that's really where we are today."

An example

Walsh gives an example where the market was pricing a small HVAC company purely as a warehouse cooling business, but what hadn’t been recognised was that its technology was well-suited to data centre cooling, a fast-growing need driven by AI infrastructure build-out.

The same technology was commanding a significant premium elsewhere in the market, but as Walsh notes, "this company is still trading as if it's a simple HVAC company."

The link was only visible by processing information across all 4,000 stocks at once, "reading all of their regulatory filings, reading all of their patent filings", at a scale no human team could replicate.

What the models can't do

The models have their limits. In volatile periods such as now, with conflict escalating in the Middle East, they can't pick up everything.

"Our models don't capture every single dynamic that's happening in the world at any particular time," Walsh says. “In these types of environments, it's incredibly important to have a very experienced and well-tenured team of portfolio managers that can oversee this process.

In stable markets, the system runs with little intervention, but right now, the team is more actively managing risk.

Volatility and portfolio construction

The portfolio deliberately holds from 350 to up to 750 stocks. This is so the team manages overall exposures rather than individual positions, and no single stock can meaningfully shift outcomes.

"One single idiosyncratic stock should never move the needle too far for us," Walsh says, a discipline that allows the team "to focus on making sure that we have highly diversified portfolios that are still expressing our views in a very risk controlled manner." 

For Australian investors considering global small cap exposure, the main takeaway from Walsh isn't a specific country or sector call, rather that AI-driven analysis at genuine scale, with experienced analysts at the helm, is opening up a universe that most managers can only partially cover. 

Managed Fund

Yarra Global Small Companies Fund

Global Shares

View