Dialogue with Sequoia Capital's Roelof Botha: AI development is overheating, rapid growth does not equal success

Source: ++Lessons from 20 Years of Venture Capital: Roelof Botha (Managing Partner at Sequoia Capital)++

Compiled by: Daisy, ChainCatcher

Editor’s Note:

In the context of the rapid development of artificial intelligence technology, AI investment continues to heat up, and the decision-making logic of venture capital is also constantly changing. Roelof Botha, Managing Partner at Sequoia Capital, shared his views on the current market stage, his perspective on AI investment risks, and how Sequoia internally addresses decision-making biases during an interview with The Generalist podcast.

He reflected on his experience in crisis management at PayPal, pointing out that in a capital-abundant environment, companies are prone to strategic slack. He also discussed how to assess the long-term potential and market ceiling of early-stage high-growth companies, as well as the importance of adhering to investment discipline in specific contexts.

This conversation provides an entry point to examine the AI startup cycle from a venture capital perspective, covering practical experiences in risk control, psychological biases, and investment strategies.

The following content is a compilation of the interview.

TL;DR: (too long, didn't read)

1. From an overall market perspective, there is no systemic bubble. Of course, in specific areas, such as artificial intelligence, there are indeed signs of overheating.

2. Accelerated company growth is part of a larger trend, not an isolated phenomenon.

3. Rapid growth in early-stage startups does not necessarily mean they have a foundation for long-term success; attention should be paid to market ceilings and sustainability.

4. When investing, if one only looks at the early growth curve, it is easy to misjudge future ceilings. Early investors must consider: can this model truly become mainstream behavior?

5. Rapid growth does not equate to success.

6. Capital excess can lead to strategic slack in companies, while resource constraints can stimulate maximum innovation and execution.

7. People are easily influenced by the information they first encounter.

8. Sequoia reduces the interference of cognitive biases on decision-making through institutionalized processes (such as horizontal comparisons of investment projects and publicly identifying psychological biases).

9. Long-termism and collective judgment are core principles that help Sequoia maintain rationality in volatile markets.

Current Market and Investment Rationality

Mario: Do you think there is a bubble in the current market?

Roelof: By my standards, I do not believe we are in a bubble period. To me, a bubble means that asset prices are generally inflated across various sectors, and that is not the case now. For example, the U.S. real estate market, whether residential or commercial, is still overall weak. We have an internal tracker at Sequoia that updates every Monday, tracking the enterprise value and revenue multiples of about 690 publicly listed tech companies. From this statistic, the overall level is currently around the 60th percentile of the past twenty years. In other words, from an overall market perspective, there is no systemic bubble. Of course, in specific areas, such as artificial intelligence, there are indeed signs of overheating, but overall, that is not the case.

Mario: Besides the weekly tracking report, what other ways do you have to maintain rationality when evaluating investments?

Roelof: In addition to tracking market multiples, we have a tool that records all completed investments of the current fund. This allows us to continuously review our recent investments and compare them with past investments. Humans naturally tend to make relative comparisons in decision-making. If you only look at the projects you encountered in the past month, the standards can easily be skewed by the current environment. However, if you broaden your perspective to the entire fund investment cycle, or even longer-term, you can more accurately judge the absolute quality of a project—whether this company has the potential to become a "legendary company." Sequoia has been around for over fifty years, and the companies we support account for over 30% of today's Nasdaq market value, which requires us to always hold ourselves to the highest standards.

Mario: With the rapid development of artificial intelligence, are there more and more companies standing out?

Roelof: Indeed, the time it takes for companies to grow to a considerable scale today is shorter than ever before. But this is not a sudden new phenomenon; it is the result of continuous evolution over the past few decades. From semiconductors to computer systems, to the internet, to mobile internet, and now to artificial intelligence, each technological leap has lowered the barriers for companies to reach users and accelerated product promotion. Especially today, with over 5 billion people connected to the internet, most of whom own smartphones, this means that new services can almost instantly reach the global market. Coupled with the massive accumulation of data, this provides rich fuel for the development of AI. Therefore, accelerated company growth is part of a larger trend, not an isolated phenomenon.

Mario: In the context of rapid development, is the biggest challenge for investors to judge the ceiling of opportunities?

Roelof: Indeed. The first challenge is to judge the upper limit of market size. For example, flash sales e-commerce (like Gilt, Rue La La), which was very popular fifteen years ago, had an impressively steep growth curve at the time. However, it turned out that the flash sales model did not become a mainstream consumption method but was limited to niche markets. When investing, if you only look at the early growth curve, it is easy to misjudge future ceilings. Therefore, early investors must consider: can this model truly become mainstream behavior? The second challenge is the issue of competitive barriers. Being a leader does not necessarily mean winning in the end. For example, Google was not the first search engine, and Facebook was not the first social network. In many technological fields, it is often the second or third entrants who summarize the experiences and lessons of their predecessors and ultimately achieve surpassing results. Therefore, when investing, we must not only judge market size but also assess whether the company's moat is deep enough to maintain a sustainable lead.

Mario: You mentioned that there are phenomena such as SPV (Special Purpose Vehicle) investments, small self-investments, and large external fundraising in the AI field. Are these signs of a bubble?

Roelof: This phenomenon is mainly concentrated in the AI field. My view is that even if the overall market is not a bubble, specific areas may be overheated. Moreover, I worry that this reckless investment approach could undermine market health, especially for the startups themselves. Reflecting on my own experience, PayPal faced tremendous funding pressure in 2000, and it was precisely because "burning cash" was restricted that the team was forced to focus on innovation, optimizing the business model, and cutting costs. Over-funding can cause companies to lose this sense of crisis, become complacent, and hinder the creation of a healthy long-term business.

Risks and Misjudgments Behind High Growth

Mario: In reality, can founders maintain discipline and rationality after receiving massive funding?

Roelof: It is very rare. Those who can often are founders who have experienced significant crises and life-and-death moments. For example, Brian Chesky of Airbnb faced a near-collapse crisis during the pandemic in 2020. This experience gave him a profound understanding of capital and risk. I believe that people like him, even if they receive abundant funding in the future, can remain cautious and disciplined. However, for the vast majority of founders who have not experienced similar hardships, it is difficult to resist the temptation of complacency brought by abundant capital solely through rational self-discipline.

Mario: It seems that only by experiencing pain can one truly learn the lesson.

Roelof: Exactly, I am the same. In 2000, when I was at PayPal, our monthly losses reached $14 million, and the company had only seven months of cash left. Initially, we spent money like water on expansion, but when the market collapsed and funding channels closed, we were forced to act quickly. During that time, we cut expenses, addressed online payment fraud issues, and adjusted our revenue model. After that, our revenue grew rapidly for three consecutive years. This experience taught me that when resources are constrained, it can stimulate maximum innovation and execution. Therefore, I often advise entrepreneurs to ask themselves: if you only had 12 months of cash flow left, what decisions would you make today? Such a hypothesis can greatly clarify what is truly important and avoid resource waste.

Psychological Biases in Decision-Making and Coping Mechanisms

Mario: When did your interest in decision-making psychology begin?

Roelof: The interest comes from multiple aspects. First, my father is an economics professor, so I was influenced from a young age. In college, I majored in actuarial science, which requires making predictions over decades, training me to think about problems on an extremely long time scale. Most people, like accountants, tend to only look at data from the past year. At Stanford Business School, I took a course in organizational behavior, where I systematically studied heuristics and biases. Since then, I have started reading extensively on this topic. Later at Sequoia, we not only introduced bias recognition in team discussions but also required each investment memo to proactively list potential psychological biases, such as: "Am I overly anxious because I haven't invested in a while? Am I too close to the founder because I know them?" By making biases explicit, we can greatly reduce their implicit impact on decision-making.

Mario: This proactive identification of biases acts as a layer of protection for decision-making.

Roelof: Exactly. If you can openly discuss biases within the team, such as someone admitting, "I feel I might be biased on this case," then others in the team can participate in the judgment more rationally, collectively reducing the impact of biases. We always believe that the team is better than the individual, and collective rationality can remedy the blind spots caused by individual emotions.

Mario: What do you think are the most common and concerning decision-making biases?

Roelof: There are two particularly important ones. The first is the anchoring effect. People are easily influenced by the information they first encounter. For example, if you looked at a company six months ago and found its valuation cheap, but now it has tripled, you instinctively resist investing. But the right question should be: is this a good entry point given today's conditions? Rather than getting caught up in past prices. The second is loss aversion. People often exit prematurely when they have existing gains, fearing the loss of profits, rather than continuing to hold potentially growing assets. This psychology can cause people to miss out on truly long-term compounding opportunities. To combat this tendency, we specifically established the Sequoia Capital Fund, which allows us to hold shares of excellent companies long-term after they go public, rather than distributing them to LPs immediately at the IPO.

Mario: Sequoia's establishment of a long-term fund is, to some extent, a way to correct the shortcomings of human nature through institutional means.

Roelof: Exactly. Establishing mechanisms is one way we combat instinctual weaknesses. Relying on individual willpower is not enough; we must improve the rationality of collective decision-making through structured design.

Investment Review and Long-Term Mindset

Mario: In your investment experience, what are some cognitive biases or erroneous decisions that have left a deep impression on you?

Roelof: There are some. For example, we had the opportunity to invest in early rounds of Twitter but ultimately chose to pass. One reason was our skepticism about the company's growth data at the time, and we did not sufficiently forward-think about its network effect potential. In hindsight, this was an error of over-reliance on short-term quantifiable metrics while neglecting the possibility of long-term nonlinear growth. This is also why we later emphasized in our internal education the importance of valuing both quantitative data and qualitative trends.

Mario: Truly world-changing companies often do not have impressive data in their early stages.

Roelof: Exactly. Legendary companies are often filled with uncertainty in their early stages; their growth paths are not linear and may even appear chaotic. If you only apply traditional financial metrics, you are likely to miss them. Therefore, identifying early potential requires combining data, trends, founder characteristics, and broader market insights.

Mario: In the current AI investment boom, how do you view the rapid growth of startups? What risks lie behind it?

Roelof: The speed at which companies grow today is indeed astonishing. Some companies have achieved hundreds of millions in revenue within just a few months. But the question is whether this explosive growth is sustainable. First, market capacity must be assessed. Some fields have a low ceiling, and growth will quickly slow down after reaching a certain scale. Entering at a high valuation poses significant risks. Second, the competitive landscape. Early leadership does not guarantee long-term dominance. Competitors continually emerge, and user stickiness and network effects are the true moats that can withstand shocks.

Mario: In other words, rapid growth does not equal success.

Roelof: Absolutely correct. Rapid growth is a signal, but it must be evaluated in conjunction with market size, moat, profitability model, team execution, and other factors. Otherwise, one can easily fall into the trap of blindly chasing high growth.

Mario: In such an environment, how does Sequoia maintain calmness and rationality in decision-making?

Roelof: We have several internal principles. First, always adhere to "discounting future value rather than tracing past prices." Even if a project was cheap in the past and has tripled now, if it still seems reasonable from a long-term value perspective, we will not miss the opportunity due to psychological barriers. Second, broad comparisons. Each time a new investment decision is made, we not only compare it with the current case but also horizontally compare it with all investments throughout the fund cycle to ensure standards do not decline. Furthermore, openly discussing potential biases. Each investment memo requires the responsible person to list their possible psychological biases, such as whether they are overly eager because they haven't acted in a while. Through collective discussion, these biases can be made explicit, thereby reducing their implicit impact on decision-making.

Mario: It seems that Sequoia places great importance on self-reflection.

Roelof: Exactly. We believe that only by continuously examining ourselves can we maintain excellent decision-making levels in the long term. The market environment changes rapidly, and any moment of complacency can lead to disastrous consequences. For fifty years, Sequoia's sustained success has relied on this culture of continuous self-iteration.

Mario: What advice would you give to young investors or entrepreneurs?

Roelof: Stay curious and remain humble. Curiosity drives you to constantly learn new things, while humility reminds you to always be aware of your limited understanding. Do not become complacent due to short-term success, nor discouraged by short-term setbacks. Truly great endeavors emerge from continuous trial and error and correction.