HyT Capital's Sun Yelin: Most Leading Companies That Have Survived Past Cycles Were Born in Bubbles
"Investors must be wary of cognitive traps."
The global technology investment landscape is currently undergoing a profound structural transformation. A new wave of technological revolution, led by AI, is advancing at an unprecedented pace. As capital market valuation systems fluctuate wildly, investors are caught in a mix of confusion and anxiety.
On one hand, AI project valuations are soaring and market enthusiasm is running high. On the other, historical experience seems difficult to apply directly — industry cycles, founder profiles, and the pace of technological evolution are all exhibiting entirely new patterns. Against this backdrop, how to navigate cycles and identify companies that can truly create long-term value has become a major question that every tech investor must confront.
At the 20th China Investment Annual Summit, Sun Yelin, Founding Partner of HyT Capital, delivered a speech titled "From Bubbles to Compounding: The Reallocation of Investment Opportunities."Drawing on the firm's ten years of investment practice, Sun Yelin systematically laid out his thinking on the current situation.
Sun Yelin pointed out that AI is the only scientific and technological revolution in human civilization to date that is capable of autonomous work, and therefore must be viewed through an entirely new lens. He urged investors to be wary of cognitive biases, broaden their aesthetic spectrum, and remain committed to founders who have long-term vision and the ability to achieve either extreme technological breakthroughs or extreme product definition. At the same time, he emphasized the need for ample patience and the capacity for sustained heavy allocation, accompanying companies through the three major growth cycles of technology, product, and organization.
The following is a transcript of the on-site speech, edited and organized by ChinaVenture:
Good morning, everyone. This year marks the tenth year of our fund's investment practice, so we've been constantly thinking about what to do in the next decade. The next ten years will be very different — especially since the beginning of this year, we've made a significant shift in focus. Our requirement for the team is: if it's not AI itself or AI-driven, we don't even look at it. We are fully committed to what we believe is the biggest trend of the next ten to twenty years — possibly the largest technology cycle opportunity in our lifetime, or even in human history. And we've been thinking about what to do in this context.
After speaking with many outstanding entrepreneurs and investors, I've come away with one clear observation: no matter how brilliant the person, everyone is trapped in some form of cognitive bias — the only difference is the size and depth of the trap.
The reason is simple. First, our lives are very short. China's investment history is only about twenty years old, and hard-core technology investing has only been around for about a decade. We iterate our cognition within a very short sampling, validation, and feedback cycle — compared to the long arc of history, this represents only a tiny fraction, perhaps less than a percent or even a thousandth of a percent.
Second, human cognitive models are largely set by around age 25 — this is a statistical pattern. For the vast majority of people, the improvement in cognitive models after 25 is limited. In addition, we are constrained by sources of information, the number of people we interact with, and other factors. As a result, most of us are trapped in cognitive biases to some degree. I, too, am constantly discovering my own cognitive limitations.
In the AI era, entrepreneurs are particularly eager for that "Aha moment." In the past, when building products, we also experienced sudden flashes of insight that revealed solutions to problems and led to significant breakthroughs. Since I started investing, this has also happened frequently: suddenly realizing that my past cognition had been severely constrained, and discovering a new way of thinking.
Learning from the Automotive and PC Industries: Bubbles and Endgames
In China's investment community today — whether in dollar or RMB funds — robotics is the hottest sector. Robotics is essentially AI with a hardware carrier, and we are also strong believers in this space. It may be the sector that will produce the largest number of great companies in China over the next decade — high in quantity, vast in growth potential, and with high ceilings for great companies. But it is precisely this strong consensus that has allowed many projects to easily reach valuations of RMB 10 billion.
Internally, we have been thinking and reflecting on what patterns tend to emerge each time such a phenomenon occurs. Historically, there have been several major hardware categories.
One is automobiles. From the first car in 1895 to 1950 — just 50 years — nearly 1,900 automotive companies emerged in the US, but ultimately the industry converged to just three: Ford, General Motors, and Chrysler. How did these three develop? Ford and GM were founded about 13 years after the first car appeared. About a decade later, GM's market share began to climb from second place, mainly because Sloan's company was merged into GM, and Sloan became its president. It took nearly 30 years after the first car for the third major American automotive brand, Chrysler, to emerge — founded by a former core GM executive.
The second major hardware category is PCs. Apple was indeed a pioneer, a testament to Steve Jobs's extraordinary foresight. Founded in 1976, Apple was one of the earliest startups in this space. Hundreds, if not thousands, of companies followed. By around 2000, the industry had converged to a few dozen, with roughly ten holding the majority of market share. Today, only a few brands remain — highly concentrated. How did market share shift? Apple once held over half the market. But after IBM introduced the standardized PC architecture and decoupled the system, Apple's share dropped rapidly, and new companies like Dell and Compaq emerged.
That's the industry perspective. Now let's look at capital bubble cycles. We've roughly tallied a few major cycles: railroads, electricity, automobiles and oil, computers, and the internet. Among them, the PC bubble saw a relatively shallow decline of about 20%–30%. The others all saw drops of 80%–90%. The pattern seems to be: hardware-driven bubbles tend to have smaller declines, while software-driven bubbles tend to collapse more dramatically. SaaS was another cycle — the era of "software-defined everything." About a decade later, the value of software is rapidly disappearing; it's still software-defined everything, but now AI defines software.
We've listed the companies that survived each major bubble cycle. Proportionally, the vast majority were founded before the bubble burst — with only a handful, like Facebook, founded after the internet bubble. This is historical data — we're not rushing to conclusions.
Using Dynamic Hypotheses to Navigate the AI Era
We have roughly analyzed the characteristics of founders of great companies that have survived past cycles: age, education level, type of institution, work experience, and so on. The results are highly diverse — there is no fixed rule like "must invest in top-school graduates" or "must invest in young people." People with all kinds of external profiles have succeeded. If we were to vaguely abstract from a very small sample: hardware-related and complex business model categories — such as automotive and e-commerce — require industrial foundations, supply chain solutions, and so on, and founders in these areas tend to be slightly older. For frontier tech software categories, founders tend to be younger. There are two exceptions in hardware: one is Dell, who was very young — perhaps because by the time Dell was founded, the PC industry chain was already quite mature, with significantly reduced technical barriers and supply chain complexity. The other is the early visionary of the PC industry, Steve Jobs.
We have also compared recent AI investments with historical patterns and found that history does not simply repeat itself — this cycle is moving much faster. The same amount of capital invested in infrastructure has seen the time cycle shortened by an order of magnitude. What we see as the essence is this: in the entire history of human civilization, there has never been a scientific and technological revolution capable of autonomous work like AI. This is the only one. We need to view this technological revolution cycle through a completely new lens, and cannot directly apply historical data to today and the future. Because a software and hardware system that can autonomously perceive the environment, make judgments, make decisions, and execute closed-loop operations is emerging at an extremely rapid pace — this is a particularly different kind of productivity revolution.
We don't actually have the answers, so we approach every sector judgment and project decision with great caution. Internally, we use an approach to guide our investment logic — that is, forming a set of important hypotheses, continuously collecting new data based on those hypotheses, refreshing our logic, and continuously updating our hypotheses.
VC's Aesthetic Expansion and Commitment to Long-Term Value
I've recently had a major reflection: my past so-called successful experiences have also limited my range of thinking. I was fortunate enough to join a fast-growing tech company on my very first day of my career, competing alongside the world's best tech companies — so we've always held ourselves to a very high aesthetic standard in investing. After reviewing all the historical data, we found that a new major hardware industry typically begins with dozens of companies emerging, then slowly consolidates over a period of ten or even dozens of years. This led to a new insight: as VCs, while we maintain high aesthetic standards, we need to broaden the spectrum of our aesthetic judgment. At the same time, we are more convinced than ever that only a few companies will survive the cycle. So we must both maintain a certain aesthetic spectrum and have the patience and capacity for sustained heavy allocation.
What kind of companies can survive the cycle? Over the past ten years, we have been continuously refining our approach to "identifying great companies that can survive cycles" — trying as much as possible to avoid founders who lack long-term vision and long-term passion. Regarding founders, while we do broaden the aesthetic spectrum, we believe three aspects are particularly important:
First: people with long-term vision.
Second: in an era of heavily technology-driven change, two types of founders are especially important: those who achieve extreme technological breakthroughs — we need to study such teams carefully, because in this era, extreme breakthroughs attract great customers and talent, and can rapidly build technical and commercial moats. And those who achieve extreme product definition breakthroughs — founders who combine long-term vision with strong capability, delivering either extreme technology or extreme product.
Third: reviewing all the great companies in history — despite their diversity — one thing remains constant: they must navigate three typical cycles: from 0 to 1, from 1 to 10, and from 10 to 100 — the technology cycle, the product cycle, and the organizational capability-building cycle. The underlying quality behind all of this is openness and a growth-oriented mindset. If you pay attention to a recent in-depth interview with DJI's Wang Tao, he is a classic example of someone who moved from the technology cycle through the product cycle — defining the aerial photography drone — and then entered a long organizational building cycle, becoming the truly exceptional company it is today.
That concludes my brief sharing with you today. Thank you.
The above article is republished from [ChinaVenture].
Back
Next article
