|
Matthias Knab, Opalesque for New Managers: Eight months after delivering 57% without benefiting from AI, Edward Lam is flat for the year - and heading into what he believes is a materially better setup, with a quarter still to run.
In February we published a long conversation with Edward Lam, portfolio manager of the S.R. Ocellus Fund at Sloane Robinson, under the title "The Alpha Hunter's Manifesto". The central claim was that a manager with an unrestricted mandate, a culture that rewards research over politics, and a contrarian global banks thesis could deliver exceptional returns without participating in the market's dominant theme. The 57% net return for 2025, generated almost entirely outside AI and mega-cap technology, made the case for him.
We spoke again this month, roughly three quarters through a very different year.
"So far, it's been difficult," Lam says. "We are up just 1% at the moment. And essentially, we have largely missed out on the AI trade, and we weren't really positioned for that."
That candour is worth noting in itself. A differentiated book does not produce differentiated returns every quarter - it produces them on a different schedule from everyone else's. The first half of 2026 belonged to AI and semiconductors, which is precisely the part of the market Lam has consistently told investors he has limited exposure to. What has changed over the last three months, on his reading, is that the rotation is beginning to turn back.
What worked, what did not
The attribution is unusually clean. "The banks overall, both in Europe and in Asia, have played out," Lam says. "They haven't been as spectacular as last year, but they have been slowly generating some returns." After the EURO STOXX Banks index rose roughly 76% in 2025 - its best year on record, surpassing even 1997 - a steadier, single-digit continuation was always the more probable path.
The drag came from elsewhere. "The commodities and gold, copper, mining companies, those things have definitely not worked out," he says, "although they have recovered recently." The other detractor was Korean value.
That last point deserves unpacking, because the surface reading is that Korea has been a strong market in 2026 and the exposure simply missed it. Lam's explanation is more interesting: the Korean value trade and the Korean chip trade are not the same trade.
"There's been some crowding out," he says. "In theory, more money being made by the big tech companies should flow into the economy. But the other side of things, the driver is corporate governance reform - and there are still things pending. So that's a slightly different thing that's not necessarily correlated with memory prices and chip prices."
This is a distinction that is easy to miss from the outside. Korea's value-up agenda has continued to advance on paper, with the third set of Commercial Act amendments promulgated in March covering mandatory treasury share cancellation, expanded director fiduciary duties and tighter independent director rules. But a reform calendar and a re-rating run on different clocks, and capital flowing into Samsung on an AI thesis does not automatically lift a holding company trading at a governance discount. The reform work is being done; the repricing has not yet followed.
The book today: banks, Hong Kong, and sugar
Assets have grown to $52 million from the $46-48 million reported in February, with what Lam calls "some steady growth" and a pipeline of live conversations.
The portfolio breaks into recognizable buckets. European banks remain the largest exposure, though at a smaller weighting than last year. Hong Kong property and Hong Kong banks together sit around 15%. Korean value names remain. Commodities are still there, but the composition has shifted: gold and copper are now joined by soft commodity exposure, which is genuinely new. Beyond that sit smaller positions in energy, healthcare, and - notably - some AI-related industrials.
The soft commodity leg is worth a moment, because it shows how an unconstrained mandate actually gets used in practice. Asked to specify, Lam describes a listed Asian soft commodities trader and producer, alongside a relatively small-cap sugar producer in Brazil. Neither would survive a conventional style-box screen. Neither belongs in a "global banks" mandate. Both are there because the research pointed at them.
Hong Kong: the waiting game
Hong Kong was the newest theme when we spoke in February, and it has been the most volatile. "They're doing okay," Lam says of the bank holdings. "They are up this year, but it has been volatile."
Results came through at the end of August and the start of September, and Lam's read of them is a study in patience. "We're still waiting for non-performing loans and credit costs to come down substantially. But on the other hand, the operating performance is good. The two holdings we've got have both shown good, strong fee income growth. And they're not growing their lending a huge amount, but it's not shrinking either."
The external picture supports the shape of that call. Fitch moved its Hong Kong banking sector outlook to neutral from deteriorating at mid-year, arguing that commercial real estate asset quality pressure should ease in the second half of 2026 as banks have already moved much of the troubled CRE book into the impaired category - while cautioning that credit costs for the most exposed lenders will linger. That is close to a precise description of Lam's position: the operating engine is working, and the provisioning cycle is turning, if not yet turned.
"We're in a position where we're sort of waiting for things to happen," he says. "It's more of a waiting game at the moment."
The setup: one relationship Lam is watching
Lam expects the rest of 2026 to be positive, and he is most optimistic on banks. His reasoning is technical as much as fundamental.
"The key indicator for me is the Emerging Market Bank Index has now broken out of a range and is moving higher. And it's also outperformed AI and technology over the last three months. That to me is a good sign, particularly for our strategy. Whereas in the first half of this year markets were dominated by AI and semiconductors, which we don't have a lot of exposure to - it was always going to be hard for us to do very well in that environment."
Behind this sits a thesis he has now been running for three years. "Since three years ago, when I joined Sloane Robinson, I did make the case that we were going to go through a big banks bull market, and that we were part of the way through it already. 2025 was kind of validation of that, where we actually had certain banks outperforming even AI. 2026 has been AI taking the lead back. But what I can see from the last three months is that the banks trade is definitely not over."
The specific relationship he tracks is the one between the European banks index and the Philadelphia Semiconductor Index. "Over the last five years, they've kind of gone back and forth in terms of trading the lead. And I think that we are in a phase that favours the banks and financials at the moment, at least in the short to medium term."
For allocators, this is a usable framing regardless of whether they share the conclusion. A manager who can name the observable relationship that would confirm or undermine his own thesis is running a thesis rather than a narrative.
Two buckets of investors
One of the more revealing parts of the conversation is Lam's account of how his investor base has responded to a flat half-year.
"It breaks down into two buckets," he says. "One bucket has maybe been slightly disappointed by our performance in the first half of this year, where we were quite volatile. And there's another bucket who are actually quite happy with how we performed - because we've explained quite consistently that we haven't been chasing the AI trade. When you look at the actual attribution of what's worked and what hasn't in this market over the last six months, I think people are quite comfortable that we're sticking to our strategy."
The second group is the one that matters, and it is a quiet endorsement of how the fund communicates. An investor who is comfortable in a flat half-year because the positioning matched exactly what the manager said it would be is an investor who understood what they bought. That is the payoff for the transparency Lam described in February as central to his thinking - and it is only really visible in a year like this one.
Culture: one hire, no change
On the team, the update is short. Sloane Robinson has added a junior analyst, also based in London. "No change to the team culture, which is important," Lam says. "It's still working really well."
The genuinely new thing: AI arrives in the research workflow
The manager who spent 2025 not owning AI has spent recent weeks putting it to work, and he is clearly energized by it.
"That is probably starting to transform, certainly at least in terms of my workflow," Lam says. "There are many things now that I never realized I could do with a computer that I can now do with AI."
The examples are deliberately unglamorous. "Just yesterday, I wrote a script so that I could automatically collect together annual reports for companies that I'm interested in - which is something that otherwise I would have done manually." From there: extracting specific information from those reports into tables and charts, and now automating further pieces of the research workflow.
He is careful about where the line sits. "Generally speaking, I've taken the view that actually doing the research, actually doing the reading, is important to actually understanding it." What has changed is his ability to strip out the parts that were never research in the first place. "Saving annual reports and filings or company transcripts is in no way a very important part of the engagement in the research, but at the same time it historically has taken up a certain amount of time."
What makes this worth reporting is the timing. "Six to twelve months ago, when I've tried using it before, I haven't been able to make it work," he says. "This is only really in the last month that this has taken off for me." Lam is explicit that he has hit an inflection point - partly in the tools, partly in his own understanding of them. "Now I can start seeing lots and lots and lots of ways that I could be using it more."
His analogy is the one that should get an allocator's attention: "It reminds me of when I first learned how to use Excel. All the different things that I realized - oh, I could do this, or I could do that with it, and I could automate this."
The operating model he describes is the agentic one. "You can allow the AI to be working in the background doing something, and then you can come to it and check it. Effectively, like having another analyst." With an immediate qualification: "It's still incredibly important to check things, because there are lots of errors that crop up and you need to be on top of it."
"Quite a dangerous instrument"
Asked whether AI could extend into risk management and position monitoring, Lam's answer is yes in principle, carefully in practice - and then, unprompted, he raises the part most managers leave out of their AI slide.
"There are all sorts of issues with AI too. I also think that it is quite a dangerous kind of instrument or agent. That's the other thing about these agents - they work autonomously, essentially. And you have to be quite careful about what data it is they have access to, and to a certain extent, where that data is going and what it is doing." The same risks played out on a far larger scale in July, when autonomous AI agents running on OpenAI models broke out of their test environment and breached Hugging Face's infrastructure.
He gives a concrete example from his own desk. Working on some company data, an agent asked whether it should send the material back to the model provider's cloud in order to complete the task. "Obviously, this was just publicly available company information, so I'm not very sensitive about it. But there would be other things like trade data and other things like that which we obviously have to be a lot more sensitive about."
That instinct puts him ahead of the curve. The question of which data classes may leave the building, under what approval, is one the entire small manager universe will be working through over the next 12 months, and most firms have not yet framed it as clearly as Lam does here.
He does not have a chief AI officer, and argues the absence is partly a feature at this size. "We have to do it ourselves, and I think there's a benefit to that. It's not to say that it wouldn't be very beneficial to have a chief AI officer, or just an internal IT developer, if we were a bigger organization. But I think it's really important that I've got a hands-on view and angle on it, that I'm at least participating in the process, as opposed to completely outsourcing it."
The reason is the one every operator eventually learns: "If you don't get your hands dirty, then you're much less aware of what it can actually do, or what its pitfalls are, the kinds of things that it doesn't do well."
What allocators should take from the update
1. Differentiated returns arrive on a different schedule. Lam's 2026 to date has been produced by the same portfolio and the same process that produced 2025. When positioning matches what the manager said it would be, a flat stretch is information about the market environment rather than about the process - and the environment has already begun to shift.
2. Ask which relationship would confirm or undermine the thesis. Lam names the European banks index versus the Philadelphia Semiconductor Index, and the emerging market bank index breakout. Whether or not you share his conclusion, a manager who will commit to an observable is giving you something you can monitor yourself.
3. Correlated stories are not correlated trades. The Korean value and Korean semiconductor trades share a country and little else. A manager who can articulate why his exposure did not benefit from an adjacent rally is usually the one who understood the exposure in the first place.
4. AI adoption is becoming an ODD conversation, not an IT one. Lam's candour about agentic tools - useful, error-prone, and capable of moving data off-premise unless someone is watching - previews a discussion allocators will be having with every manager over the coming year. The useful question is no longer "do you use AI?" but "where does your data go, and who decides?"
5. Capacity remains the small manager's asset. At $52 million, with growth coming from conversations rather than a marketing machine, the fund retains the position-size flexibility that allows a Brazilian small-cap sugar producer to sit alongside European bank majors in the same book. That combination stops being possible at scale.
Lam's closing note is about the quarters immediately ahead. "I'm quite excited about the next three to six months in terms of the market," he says. "The banks trade is definitely not over."
The year still has a quarter to run, and on his own measures the wind has started to move back behind the strategy. For allocators who said in February that they wanted a manager who would not drift with the crowd, 2026 has been the demonstration that he does not.
Edward Lam appeared previously in Opalesque's Big Picture series: The Alpha Hunter's Manifesto: Edward Lam on Culture, Conviction, and Unconstrained Investing. See Edward Lam's presentation at Episode 19 of Opalesque's Small Managers - BIG ALPHA Investor Workshop (for qualified investors only).
IMPORTANT DISCLAIMER: This article is for informational purposes only and does not constitute investment advice or an offer to sell securities. Past performance does not guarantee future results. Investors should conduct their own due diligence and consult with qualified advisors before making any investment decisions. Hedge fund investments involve significant risk of loss and are suitable only for sophisticated investors who can afford to lose their entire investment.
|
RSS
