DeepSeek founder Liang Wenfeng rarely gives public speeches. Among the few that exist, two have drawn particular attention—both are “Dark Undercurrent” interviews: “DeepSeek Unveiled: A More Extreme Story of Chinese Technological Idealism” and “The Frenzy of High-Flyer: The Invisible AI Giant’s Path with Large Models.” These have been circulated and analyzed by many.
Going even further back, there was a keynote speech he gave in 2019.
On August 30, 2019, serving as head of High-Flyer, Liang Wenfeng delivered a highly regarded keynote entitled “Outlook on the Future of Quantitative Investment in China” at the 10th China Private Fund Golden Bull Awards in the “Quantitative Investment and FinTech Sub-Forum.”
This speech caused a huge sensation among institutional investors in China and went viral at the time. It was also Liang’s first public roadshow. In it, he did not discuss strategy or Phantom Quant’s operations. Instead, he shared his reflections on the development of quantitative investment within China—a kind of educational introduction to the broader financial community. He included a lot of carefully estimated data; one can sense the effort put into it.
At the time, he made a bold assertion: from an investment standpoint, on the technical side, “programs already far surpass most human capabilities.” Moreover, he stated that the progress of quantitative private funds aligns with Moore’s Law—doubling in investment capability every 18 months. A quant portfolio manager on Zhihu commented, “He explained quite well what quant is and what the future of asset management looks like to those outside our circle. And for those inside the field, he also spoke up on behalf of quant and cleared some misconceptions.”
Below is the full English translation of Liang Wenfeng’s speech on that occasion:
The Future of China’s Quantitative Investment Through the Eyes of a Coder
First, to predict the future of quantitative investment in China, one approach is to look at the current landscape in the United States—our “teachers.” In U.S. asset management, there are two main trends: one is the increasing indexation of mutual funds, and the other is the gradual move toward quantitative approaches by hedge funds. In other words, overseas hedge funds are somewhat equivalent to China’s securities private funds.
Initially, hedge funds were not quant-based at all. This chart here shows the top 10 global hedge funds by AUM (assets under management) in 2004—most were not quant-driven. But last year (2018), looking at the rankings, quant funds occupied the majority of top spots. Bridgewater was number one, AQR number two, and Renaissance number four. Over the past decade or so, quantitative funds in the United States have gradually become the mainstream in hedge funds. Many people even conflate “hedge fund” with “quant fund.” Because we ourselves are a hedge fund, today I will focus on quant funds within hedge funds.
From the American experience, quantitative private funds can reach very large scales. The world’s largest hedge fund, Bridgewater, manages around 1 trillion RMB. Meanwhile, large Chinese quant firms currently manage only 10–20 billion RMB. There is, potentially, tens of times’ worth of growth ahead. Is it really possible for a private fund in China to manage 1 trillion RMB? I think it’s absolutely possible. China’s economy will eventually be on par with that of the U.S. A top domestic team should be able to manage 200–300 billion. If the stock market expands and the derivatives market matures, that might become 400–500 billion. Add global markets, and 1 trillion could be within reach.
What do all these quant firms overseas actually do? Are they all doing high frequency? Obviously not. High frequency cannot accommodate large amounts of money and isn’t the mainstream in asset management. The answer is that they do everything—from global macro to fundamental equity, from equity price-and-volume to commodities, to bonds. The main arenas are equities and bonds. Bridgewater, the largest global hedge fund, is a macro quant shop; AQR, the second largest, focuses on equity fundamentals. The more low-frequency the strategy, the higher the capacity. Essentially, every strategy humans have undertaken is now being done by quant methods. Meanwhile, Chinese hedge funds are still primarily in price-and-volume strategies—lagging behind the U.S. If we look at America’s experience, in terms of strategy types, we still have a lot of room to grow.
How exactly do we differentiate between quant and non-quant?
Given China’s current situation, let’s define quantitative investment. Some say that “quantitative investment means using automated (programmatic) trading.” That isn’t correct, because many quant firms place trades manually, while many traditional mutual funds have programmatic trading systems (like mature VWAP systems).
Others say it’s “using quantitative methods for research.” That’s also not accurate. Modern investment research often relies on quantitative methods, so that definition covers nearly everyone.
Another common notion is: “Subjective managers focus intently on individual stocks, whereas quant managers don’t.” That’s not true either—at least our firm studies individual stocks very carefully, and so do our counterparts in the U.S.
So what is the real distinction?
The key difference is how investment decisions are made: by quantitative (program-based) methods or by human judgment. It’s not about how you trade or conduct research but about your decision-making process.
Quant firms also have plenty of traders and researchers, but you’ll notice that they don’t really have a “portfolio manager” in the traditional sense. Their “portfolio manager” is essentially a cluster of servers. When humans make investment decisions, it’s an art—it relies on intuition. When a program decides, it’s a science—it aims for an optimal solution.
People often ask: “If quant investing takes over, will humans still be needed?” Of course! You’ll need a lot of programmers and researchers.
Let’s look at what quant investors in China are doing.
We estimate that total quant assets in the Chinese market are somewhere between 250 billion to 500 billion RMB. Over half of that is allocated to equity strategies, followed by commodity CTA, and the rest is minimal. Historically, equity strategies have slightly outperformed commodity CTAs. Let’s focus on equity strategies today. The table here (not shown) is what we and our peers estimated—though it might not be exact, the overall picture is about right. If you’re looking to invest in quant, pick a manager in line with that table.
In equities, we typically divide strategies into four categories: the most important one is Daily (interday) price-and-volume models – These are often referred to as multi-factor or alpha models. This is the largest category, roughly 200 billion RMB in size. The second is Intraday mean-reversion or T+0 models – Commonly known as “stock T0,” a few tens of billions or more.
The last two categories are fundamental models and event-driven models, but they are not the focus for now. These figures pertain to private funds. Additionally, in the mutual fund (公募) space, around 120 billion RMB is allocated to fundamental quant strategies. However, today we are only discussing private funds.
All four types can be profitable. Traditionally, all models were multi-factor systems that capture alpha via stock selection and timing. Before 2017, multi-factor models were considered a “cure-all,” and many tried to emulate WorldQuant’s method of employing large teams to dig up as many factors as possible. The competition was over who had the most effective factors. By 2017, it was already very difficult to discover truly innovative factors. After that, the industry began to shift, and the traditional multi-factor framework was gradually replaced by AI. By 2019, it started to be overtaken again by newer “ensemble” frameworks.
As private fund managers, our investors have high expectations. If we underperform the index by less than 25% in a year, that’s considered poor. Competition among private funds is fierce. Every week, we get updated performance data from our peers: how many points each competitor outperformed. If we fall behind, clients call us right away. The pressure is intense; I believe everyone in the industry feels it. But that pressure forces us to improve our investment capabilities, to keep refining strategies—if you slack off, you’ll quickly fall behind. Of course, our management fees are much higher than those of mutual funds, so the outperformance is expected.
People often ask: “Whose money do quant funds actually make?”
The answer is straightforward: quant funds earn the money that used to be made by traditional human (discretionary) investors. Those investors can be divided into technical and fundamental camps. Concretely, at the moment, quant is taking over much of what was previously earned by the technical analysis side. Who did technical analysts make their money from originally? In many cases, from market inefficiencies that have now been mostly captured by hundreds of billions in quant strategies, drastically increasing market efficiency. Over time, it’s only going to get harder for human traders because algorithms keep getting better. Here in 2019, algorithms in technical strategies have already far surpassed most human experts.
As a whole, the quant private fund industry progresses roughly in line with Moore’s Law—it doubles in capability every 18 months. However, over the last few years, the average performance of quant funds hasn’t really changed much because the market has been getting more efficient in tandem. The logic is: if investment capabilities doubled while market inefficiency stayed the same, the profits should be double. But because the market is becoming more efficient, that increase hasn’t translated into twice the gains.
One indicator of rising market efficiency is that even top human traders find it harder to profit. Another is that previously effective quant strategies gradually stop working. Quant investment still has plenty of room to evolve, so we expect China’s equity market efficiency to continue improving in the coming years. It’s a historical trend—unstoppable.
A frequent concern is: “If the market becomes extremely efficient, will no one make money?”
From the U.S. experience, markets never become completely efficient. Otherwise, hedge funds would vanish altogether, and there would be no one to provide liquidity or price discovery. Instead, markets reach a near-efficient equilibrium—such that hedge funds can cover their operating costs and investors’ capital and risk costs. Worldwide, hedge funds are not exactly a “massively profitable” industry—certainly less so than primary-market ventures or real estate. At our current stage in history, we’re still quite far from full efficiency, so we don’t need to worry about that scenario for at least the next few years.
Let’s conclude with two predictions: a short-term and a longer-term outlook. If these turn out to be accurate, quant returns will remain healthy for years to come.
The short-term prediction is for the next one to two years. Over the next one to two years, industry improvements should come from multi-strategy integration. Multi-strategy integration is not simply diversification. Diversification works like this: with 4 billion RMB in capital, 1 billion is allocated to Model A, 1 billion to Model B, 1 billion to Model C, and 1 billion to Model D. The drawback of this approach is that the return becomes the average of the four models. What we refer to as multi-strategy integration is overlaying strategies: the same 4 billion RMB is simultaneously used for Model A, Model B, Model C, and Model D, ultimately forming a large, all-encompassing strategy that does not belong to any traditional category.
Last year, the combination of interday alpha and intraday T0 performed well, but it has now fallen behind. More strategies and more advanced integration methods are now required. This sounds logical, but in practice, it is very challenging. The difficulty does not lie in the strategies or the technology itself but in the business logic within private equity firms. Every model requires a dedicated team. Previously, a single team could manage tens of billions in AUM, but now, multiple teams are needed to manage the same amount, significantly increasing costs while company revenues do not increase proportionally. However, from our observations, this trend is already underway. If you do not do it, others will. The best-performing private equity firms in recent times are all multi-strategy firms.
We expect this process to accelerate. As market efficiency improves and returns decrease, relying on a single strategy to achieve strong returns will become increasingly difficult. In the future, strategies will be extremely complex, workload-heavy, and have high barriers to entry. Quant firms that cannot organize multiple teams will struggle to survive. Quantitative investment will concentrate in leading firms, which have the resources to implement these more complex strategies. We believe there is still significant room for development in multi-strategy integration. Based on our own progress, it will take at least another one to two years to complete. If this prediction holds, quantitative private equity firms should still see decent returns over the next one to two years.
The long-term prediction looks at the next three to five years. One day, the volatility in technical signals will diminish, and technological advancements will reach a bottleneck. At that point, quantitative investment will inevitably begin taking over the profits that fundamental investors once earned. In terms of fundamentals, market efficiency is still relatively low, leaving significant room for growth. Fundamentally driven quant investing is technically feasible. Some argue that since each company’s fundamentals are different, they cannot be quantified, but this is incorrect.
First, the U.S. has already achieved fundamental quant investing, so why can’t China? Second, if technical factors can be quantified, why can’t fundamental factors? Around 2015, fundamental quant became popular among private equity firms for a period. At that time, market efficiency was not as high as it is now, so traditional multi-factor frameworks could still generate profits. However, starting in 2017, returns declined, and private equity firms focusing on fundamental quant lost competitiveness and were gradually phased out, though mutual funds continued their efforts. Private firms must elevate fundamental quant to a higher level. This mission will not be accomplished by the old generation but by a new group of more capable individuals using more sophisticated and refined methods.
Our current products have already incorporated fundamental quant models, and the results have been promising, but we have only applied traditional methods so far. To take it further, a more meticulous approach is required, which incurs significantly higher costs than technical models. To reach the level of AQR, we conservatively estimate that team costs would exceed 1 billion RMB annually. This means the process must be gradual. However, if a quantitative private fund eventually manages 100 billion RMB in AUM, this cost becomes justifiable, and the business model remains viable.
Fundamental quant still has a long way to go. To reach the current level of technical quant, it will likely take several more Moore’s Law cycles. But this day will come within our lifetime.
The final question: If hedge funds end up making the money that used to go to both technical and fundamental human investors, what’s left for ordinary people?
Again, we look to our American “teachers.” Hedge funds only earn returns related to volatility, liquidity, and pricing. They don’t “take away” beta—market returns. The largest U.S. hedge fund, Bridgewater, manages around 1 trillion RMB; the largest U.S. mutual fund, BlackRock, manages 45 trillion RMB. Compared to mutual funds, hedge funds are dwarfed. When the market is efficient, you can just buy the index. That’s real value investing. The bulk of wealth still remains in the hands of the public.