My worst investment ever was probably the first Iraq War. So there's a long goes back a long time. I remember because the date of the invasion was August the 2nd, 1990. It was my birthday. And then seven months later, the Americans finally struck on Iraq. By that time, I had built a very significant long position in oil services, oil stocks, and so on. And I had an absolute nightmare, absolute nightmare performance. My worst performance year ever by a long way. Could have lost my job, really.
Hi everybody. I'm Nicolai Tangen, the CEO of the Norwegian Wealth Fund, and today my guest is one of the most successful investors of our generation. Sir Paul Marshall started Marshall Wace in 1997, and today is one of the largest and most successful hedge funds in the world. Marshall Wace combines man and machine, and they built the world's first alpha capture application, which is like a system that collects trade ideas from thousands of brokers, tracks who gets it right, and puts money behind the best. Now Paul has also written down what he knows about investing in his book, "10 and a Half Lessons from Experience." He's also huge on philanthropy. So let's explore this and other things today. Paul, warm welcome.
Thank you, Nicolai. Great to see you again. It's been far too long.
And I should say, when you say I started Marshall Wace, Ian Wace and I started Marshall Wace. Very much a partnership, and it's remained a partnership and a broadening partnership all through its life.
When you started Marshall Wace, who decided that Marshall was going to be before Wace?
Now, today you run around $90 billion. What is it that Marshall Wace is best in the world at?
Well, it's very presumptuous to say you're best in the world at anything, but I guess what we could claim, I would probably claim two things. We're probably best in the world at alpha capture because we created alpha capture. So that was our unique contribution. Secondly, I would say we may be best in the world at combining discretionary with systematic investing. So over the years, our original route in discretionary fundamental investing has been combined with systematic, and they have actually turned out to be highly synergistic. So our discretionary side has benefited hugely from the systematic side, and the systematic side has benefited hugely from sitting in an environment where people basically are very focused on fundamental and on the market. So it's never ever been remotely detached from the market. In fact, it's a very market-orientated way of doing systematic.
And we'll come back to that. But why have you been successful now roughly 30 years when so many other funds have kind of fallen by the wayside? What do you think is the key to—
I think the single most important thing is continuous innovation. So a lot of people in our industry, perfectly legitimately, started as a very smart guy looking at balance sheets with a Bloomberg terminal, and that's all you needed. And some very smart—not very many, but some—very smart guys have continued to flourish, primarily relying just on being very smart. But we have gone well beyond that and effectively embraced every type of innovation to the point where we're almost now a tech firm rather than an investment management firm. So it's changed out of all recognition because we innovate every day, basically.
Because you do say in your book that most fund management careers end in failure.
Yes. Well, that was a different point really, and that was kind of reflecting the observed pattern of my previous firm. But also a lot of fund management firms, I think, get into the problem of hubris, where they grow and grow, they don't have discipline around capital, and they don't have discipline around or a sense of their own mortality. And so a lot of firms get too hubristic, too big—and size matters, which is another thing in the book. So they get too big in the market, positions own them rather than them owning the positions, and when they're wrong, that can drag their business down and destroy their return patterns.
How are you reminded daily about your mortality? Do you have somebody telling you?
I don't have anybody at my shoulder. But basically, as you know, Nicolai, if you look at the success ratios of managers, a really good manager might have a 53-54% success ratio. So 46% of the time, they're wrong. So every day, one day in two, you go home feeling you're a schmuck. And my wife can testify to that. I can come home, I feel I'm a schmuck, I got something wrong. So you've got to be pretty head in the clouds not to understand that actually it's a very fragile path and it's very difficult.
Well, it's a humiliating job, but yet there are a lot of hedge managers who are not so humbled. But that's another discussion. But let's go back a bit. In 1997, you were head of European equities at Mercury, which was, you know, a tremendously successful company, and then you suddenly decide to walk away and start a fund with Ian Wace. Why?
I'd always wanted to have my own business.
I don't know, that's something in me. I even made the mistake of announcing it within six months of being at Mercury that my long-term goal was to have my own business. But it was Ian who approached me.
And he was the head of sales. So he was doing sales?
No, he was originally head of sales at Warburgs. He was the first person I ever spoke to in the market. He called me my first day at Mercury, and they were part of the same company. Then he became head of proprietary trading at Warburgs, and then he became head of proprietary trading at Deutsche Bank. So he moved away from sales. And so he was ahead of the game really. He wanted to set up a hedge fund in Europe at the time when there were only arguably two other equity long/short hedge funds in Europe. And he approached me, and the timing was perfect from my point of view. So—
Are you the deep-thinking analytical machine, and he's kind of the fast-moving kind of thing, or how did—
I would say that was the original basis. Ian is also very strong on technology. He's a great believer in the power of technology. And he created at Warburgs a trading system which still exists within—I think it's UBS now that ultimately owns it—called Oasis. So he brought an interest in tech, which to be honest, I was not very focused on. And so we are, as I think you know in a little bit, we're extremely complementary and understand each other's strengths and weaknesses.
And you said that you were fund number three in Europe. So there you had already two other firms. They were—
Odey and Egerton, I think, were the two that—
Egerton, FYI, is where I learned my things. And by the way, John Armitage is coming up in a future episode, so stay tuned.
Now, how much more difficult is it to start a hedge fund today than what was the case back in '97? What does it take?
That's a great question. I think the AUM you needed in 1997 was $50 million.
AUM as assets under management, right?
Right. Now, primarily because of the regulations, basically most people estimate it's 300 to 500 to get started and cover your costs. And I think also in those days, my expectation of our business and of the industry was it would remain a cottage industry, highly fragmented, lots of small operations. As I said at the beginning, people looking at annual reports and crunching the numbers and making calls. And that's how I thought it would stay. And now of course, it's become highly concentrated, just like almost every industry has become concentrated. The entry barriers have risen and risen and risen. It's much more dependent on technology and operational capability, financial capability. All of these things matter much more than they did originally.
You got half your start capital from Soros. How did that come about?
Well, we basically just went and knocked on the door in New York, and the meeting was actually with Stanley Druckenmiller. And there's a very funny story, which is too long to tell on this podcast, but—
Somebody said, "Be careful with Stanley. He's like a lizard. He'll sit in the corner of the room, won't say anything, and then he'll pounce." And that's exactly what happened. Didn't say a word. Then at a certain point he said, "Tell me, do you ever use any charts, technicals, when you invest?" And Ian said, "Quite frankly, I've never met a rich chartist." And Stanley said, "Thanks very much." And that was pretty much the end of the meeting. We went outside the room, and the people who had brought us there, which was Morgan, said, "By the way, you just met the world's richest chartist."
Absolutely. Well, we've had him on the podcast. He's very, very successful and very good.
And a good choice, and evidenced as his investment in you guys.
Now, just after you started, you had the Asian crisis. You just went straight into these big things.
I think the Asian crisis was actually when we were marketing. I can remember in New York being—it might have been the day of the Soros meeting—the market was down 7% on one day. But no, first after we launched, we had the 1998 LTCM crisis, which we were fine with, and the Russian crisis. But we didn't have much exposure in either way, so we weren't affected very much by that.
Have you ever been close to failing as a firm?
Yes. Well, not maybe—to some degree. In 2008, we went from $14 billion AUM before '08 to $3.5 billion. And the reason primarily—our performance definitely wasn't worse than other people, but our client base was weak in the sense that it was largely fund of funds because we'd never actually marketed. We never went out like some competitors to institutions and tried to build a firm client base, and we relied on people knocking on our door. So it was only in 2004-05, which was seven years in, that we had any marketing people.
And fund of funds is a fund which takes money from retail investors and invests it in funds like yourself.
Yeah. And what happened was that a lot of them got into difficulty and had liquidity problems. A lot of our competitors gated investments, even though there was no real basis for gating because there was no liquidity issue. We did not have a liquidity issue, so we did not gate. So we became the cash point, and assets under management went down from 14 to three and a half.
To be honest, we were angry with our competitors, but I never felt much angst for some reason. I actually enjoy crises. So it was a moment of—I was actually more excited by what was happening and what was going on in the market than I was stressed by the risk of the business.
How do you cope with stress? Do you have feelings? Do you get angst?
Less and less. One of the criteria—we now do personality tests for all of our hires, and we use the five core personality trait test. One of the five traits is stress tolerance, and I have an extremely high stress tolerance, kind of first or second percentile. So I don't get overly stressed. I actually enjoy crises. They're the best. They're the best.
Do you think your wife would agree with you?
Yes. Brilliant. She would in the evening, but I think she would. Yeah.
When did you first understand that, "Wow, we are onto something big here. We're going to be really successful," and kind of high-fiving Wace and really happy about life?
Well, there was obviously a successful period after alpha capture. We created alpha capture kind of in the 2004 to '07-'08 period. But then the real—and then obviously 2009-10, we were building out of the crisis, learning our lessons, doing things a bit differently, more strategically. And I think it was probably '13-'14 when we were starting to really see the fruits of that in terms of the growth of the business and the solidity of the returns.
You sold a stake in your business to KKR. Well, I think you've done it in two stages, but why did you want to have a private equity firm as a part owner of your business?
They approached us, and they actually approached us twice before we were interested to have a conversation. And I would say that the key element of the whole discussion was that they only wanted to be a minority investor, not majority investor, and that was critical to me. But the key at the time, 2015-16, was that we were a partnership. We wanted to create some kind of way of putting a value on the shares in a real market-related sense. And there was a secondary benefit, obviously, that being endorsed by a firm like KKR was a pretty good endorsement and no doubt helped us with institutional fundraising.
Would you have done it again?
That's a very good question. With hindsight, we feel that we would have succeeded in pretty much the same way anyway. But they've been a fantastic partner to us. They're a great firm. They have a highly entrepreneurial culture at the top of the firm. And so there's never been a moment when we've really regretted it. I think it's very unusual in the sense it's an investment that's worked for both sides.
Let's move on to TOPS, which is your alpha capture system and which is, in a way, what you're famous for as a firm, right? Tell me about the start of it.
Yeah. So the genesis story is that Ian—this was 2002, so five years in—Ian said, "You know, there's got to be a better way of measuring the street, the sell side," because we were giving out commissions, like some people still do today, I think, by a voting system or some kind of qualitative assessment of their research. And it was a big budget. And so he said, "Well, how can we measure this more scientifically?" And so we gave to a young intern called Anthony Clake the job of saying, "How can you measure the sell side?" And what he did, it was the age of the internet, and he said, "Well, let's create an intranet and get all of the salespeople on the sales side—it was initially just Europe—to contribute their ideas via a portfolio system." So we created a virtual—each of them had to run a virtual portfolio for us on a transparent intraday basis. And this, at the time, was a completely radical change. It was almost like a reverse engineering of the relationship between the sell side and the buy side. But very quickly it caught on. I was the big skeptic because I had the arrogance of the buy side. I said, "Well, surely that all sell side doesn't generate very much value," and I was completely wrong. Within a few months, it was clear that sell side were actually creating a lot of value for us, which we could—originally, we weren't even intending to monetize it—which we could monetize and turn into real portfolios that we would manage. After about two years, we launched a real monetized portfolio under the Marshall Wace banner. Yeah.
And you put what, 10% of the capital or something into it?
How does it work now then? How many external contributors do you have to this platform?
Well, it's hundreds. We don't give the exact number, but it's obviously all over the world. And the first phase of the evolution was to take it from Europe to the US, which was in '05, and then to Asia in '06. And that was, if you like, the global rollout. Then the next thing is that obviously—this is how we got into the subject of continuous innovation—we knew we had to develop optimization to extract the best data, to identify how to combine signals, combine contributors, combine signals about contributors, look at patterns, so on and so forth. Then the next thing was to look at our trading because essentially TOPS traded and trades around twice—turns over twice a month, so much more than our fundamental side of our business. So that led us to go into algorithmic trading very early for a fundamental firm, obviously not so early for a systematic firm. And then post-2010, we developed very active machine learning to mine the signals properly and to branch out into all other types of data to bring in to create signals. So TOPS became much, much more than an alpha capture system. It's a proper, very market-orientated systematic strategy.
So with AI, what's some of the mining you can do? What are the types of things you can look at now?
Well, AI, obviously some of the things that people associate with AI actually were already happening with machine learning, like scraping for sentiment, scraping every broker's note for sentiment, scraping social media for sentiment, looking at balance sheet data, all kinds of fairly obvious things you could already do. What AI does is take it to a whole new level, a whole new depth. And so whereas looking at sentiment data might be looking at kind of fairly basic signals and word patterns, AI allows you to look at the whole context of every single piece. But what I would say, Nicolai, is that today, although AI we think is going to be totally transformational of our TOPS system, the biggest change to date has been on our fundamental side, where it's been completely revolutionary. And this is again where the synergy comes in. Because of the infrastructure we have, we've been able to apply it immediately to the way we do fundamental investing and everything to do with distillation of information to start with, multiple varied ways of tools to distill information for our managers so that we absorb all information very rapidly overnight and they have it at their fingertips. But then also we've moved to the point where we already have agentic portfolio managers developed by our fundamental teams by transforming their skills into a portfolio system. And all of that is before we go to recursive self-improvement. So our head of tech thinks that we're going to go from 200 quant researchers to 10,000 quant researchers, by which he means we'll have 10,000 agents. And they'll be recursively analyzing all of the data, improving and developing new ways of implementing the signals into live portfolios.
Paul, how many people are you in the firm now?
And how many people are in tech?
It's interesting, because in the Norwegian Sovereign Wealth Fund, we are 700 and we have roughly 200 in tech.
Oh yeah. So we're moving on a parallel path, except you've got rather more assets than us.
Well, have you increased the numbers in tech since you arrived?
Yeah. Yeah. Meaningfully. Yeah. And on the AI side, of course.
Hey, this is really interesting because it kind of leads into your book, "10 and a Half Lessons from Experience." And there you talk about the fact that you say markets are inefficient, humans are irrational, investment skill is measurable and persistent. The thing is that with people like you operating in this market, aren't markets just becoming more and more efficient? How can a human being beat these markets?
I think although efficiency itself is a kind of convenient myth, it is also the case that markets are getting more efficient. I would agree with that. That's because information—when you and I started, we used to talk about information edge. It's very difficult now to have an information edge because the good firms are receiving and processing the data all the time in real time. But analytical edge still remains, and what you've always got to look at is the ecology of a market. So I think actually what's going to happen with AI is that more retail investors will feel more and more empowered. They'll think that they have just as much information as Nicolai Tangen or Paul Marshall, and they probably arguably do if they're well-organized. Therefore, they're empowered to make investment decisions. So the share of trading done by retail will go up and up. And I would say that you can't necessarily rely on retail to be as skilled as the best hedge funds. So I think it is getting more competitive, but I'm not at all clear whether it's going to get more efficient over the next years.
Is the retail part of the market kind of the market where US professionals can make more money? I mean, are retail investors there to be fleeced?
I think that's always been the theory, right? So you look at the markets with the highest retail share, like China, Taiwan, Korea, US is now growing, and they're likely to be more inefficient because of the way the decisions are made. And that's definitely been the case in terms of where a lot of the alpha you can extract. So I'm saying that's likely to continue. Clearly the retail investor is going to be more empowered now than he or she was in the past, but I still think they won't take decisions, they won't have the same risk management culture, they won't have the same set of criteria as hedge fund specialists.
So 700 people, 200 in tech, that leaves 500 people. In 10 years' time, how many people do you need, you think, with all these machines?
At the moment, think that the headcount—there's no reason why it's going to move very much from here. It went up a lot in the last three years towards 700-750. It's been stable for two years. We think that AI is leading to huge boosts in productivity and the quality of decision-making. But it's not leading anybody to say at the moment—there's one team where they've cut back, one fundamental team where they've cut back the headcount, but that's one out of 15—not leading anybody to think that we'll need less people. We may be needing a different type of person, that's a different question, but we don't think we need less people. We think we're going to get more out of the existing people.
Very interesting. It's exactly our philosophy here as well.
But does it help you trade more cheaply? What does it do on that side?
I wouldn't say that. Yes, but that is not the primary focus. We're not directly competing with high-frequency traders. We haven't made the decision yet to go in that direction. It just helps us to trade. I mean, we've obviously been trading more cheaply forever. Our trading cost numbers are very low, I would suspect, compared to any other traditional hedge fund other than systematic hedge funds. But we're not going down the route of HFT and giant data centers just focused on that yet.
Why does man plus machine beat a machine?
So that saying, which is one of the chapters in the book, originally came allegedly from Garry Kasparov after he was defeated by Deep Blue in 1997, so when we just started Marshall Wace. And the original, the first part of it was: a machine beats a man, but a man plus a machine beats a machine. And I guess the reason I'm confident of that is it presupposes that there are skills that the AI can't do and that will be less good at than a human. And clearly there will be some things that the AI can do better. So we now have a term, "nowcasting," the whole business of assembling data and getting every single possible input together in order to anticipate quarterly earnings and positioning and so on and how stock—that is, we think, getting commoditized. So that is getting a harder space, which never was the main space we operated in. That's a pod shop space. But what will remain, in my opinion, is the skill of judging the market, judging how the market behaves in particular circumstances.
So if you just take the recent fun and games in Korea and in this momentum crash we've seen in the last four to six weeks, obviously what you have, the particular example of Ken Griffin taking out situational awareness, that's a man thing, a machine wouldn't have done. But every good hedge fund in that environment would have looked at this set of circumstances, and they're looking not just at the fundamentals of the stocks, but also of the ecosystem of the market. What's going on here? What is happening in Korea? Who are the buyers and sellers of the Korean market, which has set this all off supposedly? What's happened to the leveraged ETF market? What's happened to hedge fund holdings or foreign holdings of Korea? What's happened to the hedge fund behavior in the United States? Then you get into understanding looking at how risk managers in traditional hedge funds would have been basically de-risking the funds because their vol was too high. So they're systematically de-risking arguably, in our opinion, at the wrong time because of high vol rather than because of anything wrong with the fundamentals. So what I'm saying is there's a whole balance of things which include both the fundamentals and the technical situation and the ecology of the market, which I don't think a machine yet will be very good at. It's the human who brings it all together and sees the opportunity and senses blood in the distress.
So you mentioned a lot of the factors which are impacting the market just now. And you have the machines, and here we have the human, Paul Marshall. How do you interpret today's market? Where are we today?
I think, unlike the FT who said we were in a bubble two years ago, I think that we're not in a bubble yet, but that we will go into a bubble at some point in AI. But at the moment, the fundamentals for me—stepping back, artificial intelligence is the most important thing that's happened arguably in the history of capitalism. I try very hard to find superlatives to try and communicate internally and externally what's going on. The best one is the idea of a Cambrian moment, when there's a huge explosion in biological evolution 500 million years ago. It's that type of thing, or it's like the steam engine. It's an intelligence explosion, and it's huge. And you've got the external data, and you've also got our internal data, and our own use of tokens is exploding at the same rate as you see on OpenRouter or as we see in Anthropic's data. And so all through the year, I've looked at what's going on in terms of the token explosion, and I think we're early days, particularly because in six to 12 months, we're probably going to go into recursive self-improvement, when this is going to be happening 24/7 overnight. We're going to set our computers to work overnight—well, they already do, but they're going to be self-improving overnight.
What is the recursive self-improvement going to do, you think?
Well, I think it's going to be coming up with—whereas at the moment the humans are coming up with the skills or the prompts, and then the skills which we put in the test overnight, the system is itself going to be self-correcting, evaluating, changing the skills, and modifying the skills. So we're only at the very, very beginning. And I think hedge fund management is somewhat at the tip of the spear in terms of AI, so we actually have an early glimpse of the scale of the change. Corporates are much slower to adopt it for all kinds of historic and bureaucratic reasons. Big companies find it difficult to approve things quickly.
So the adoption numbers are going to carry on exploding, and so I think we're in early days. Coming back to the market, then you look at the market and, as you know, you look at the valuations of the stocks in the center of the story: they're cheap. The semi index in the US is at a five- or 10-year low, depending on how you look at it. The Korean market is at an all-time low valuation. Now, there are arguments about the cyclicality of some of the stocks, but essentially the stocks have only moved up now in line with earnings, if that. And so that part of the market is going to carry on powering ahead, in my opinion, and we haven't really seen the big effects yet in terms of the productivity benefits from adoption. So the rest of the market will also benefit from that. So I think we're going to go into a very good period. It will no doubt have its bubble components and it'll get overheated. We already had one overheating moment, but I think that's what we've got to navigate: a market where there'll be a lot of excitement and periodic overheating, but the underlying fundamentals are very strong.
When do you think we'll go into a bubble?
I don't know. I can't predict when we'll go into a bubble, but for the bubble to really happen, it needs, one, I think, rates to stay low and to encourage speculation; two, valuations to get very extended, which definitely hasn't happened; three, increasing leverage, unsustainable leverage, which you had in the Korean ETF market. That was a bubble phenomenon, but it's now corrected. So the numbers have basically gone from $60 billion of leveraged ETF down to 20 in the last five weeks. So you'll get those types of leverage conditions. And I guess if this Nvidia $500 million deal—if they then try and turn those into CDOs of GPUs and sell two-times levered versions of those instruments, those are the types of phenomenon we got in 2007-08. But we're not there yet. I think my path would be: absolutely, we're going to have a big bull market with a bubble top, and you're going to have to navigate all of that, but I don't think we're there yet.
How much higher is the end of the bull market from where we are now?
