When the STAR 50 Becomes the “STAR 25”: Has the Global AI Bubble Burst?

“China’s private markets, China’s public markets, America’s private markets and America’s public markets exist in four different times and spaces. They are parallel universes.”

— Raymond, on the differences in market temperature he felt after returning from the United States

Guest: Raymond | Host of Mossfire

In conversation with: Dong | Huiketang

At a dinner in Silicon Valley, people may become a little more courteous when someone says they are in the compute business: Do you have any new supply? Could you allocate some to me? Back in front of a trading screen, the credit risk of that very same company may be what investors worry about most.

Raymond spent nearly a month in the United States meeting founders, investors and industry partners. His impressions did not fit neatly into the black-and-white question of whether technology stocks are in a bubble. Some businesses are doing very well; others cannot raise their next round. Some sectors are just heating up, while cracks have appeared in others. People inside the industry are competing for resources, while public-market investors are calculating risks. Both sides may have their own reasons.

This conversation between Huiketang and Mossfire took place after a bout of technology-stock volatility that had left investors deeply anxious. The two speakers’ views have not been blended into a single voice: Dong puts more emphasis on ordinary investors not exhausting themselves trying to predict bubbles; Raymond prefers to bring the question back to his own positions, continuing to study entry points and what to do if his judgment proves wrong.

Between being optimistic about a technological transformation and ultimately making money from it lie valuation, the duration of capital, position discipline and a person’s actual response to losses.

The conversation follows, edited and condensed.

1. “The temperatures you see are completely different”

Dong: A trading app gives you very little information from which to imagine a transformation of this scale. You have just returned from the United States. From what you saw, how do the Chinese and American technology industries differ today?

Raymond: I had not been to the United States for a year or two. This time, around China’s National Day holiday, I stayed for almost a month and met quite a few people in technology: founders, investors and partners in the ecosystem.

We normally look at price charts and financial statements. If you can access X and YouTube, you can find most public-market information. But when you sit down with people locally, the state they are in feels different. Whether they are Chinese or from other backgrounds, putting together what they are doing and what they worry about gives you a more complete understanding.

I wrote an article when I returned, thinking of it as little more than a travelogue. What surprised me more was the response: the praise seemed disproportionate. I think one reason is that many people do not have much opportunity to go to Silicon Valley and speak with people there. Another is that people really are very anxious.

Everyone wants to get in and get out at the same time. They think it is a bubble, but also think the bubble has already burst. People cannot work out how to face this market from the position they currently occupy.

My strongest impression was that, depending on which aspect of the American or Chinese market you look at, the temperatures are completely different. I spent more time looking at private markets on this trip. Many areas were genuinely flourishing, and there were reasons for that prosperity.

For example, growth at foundation-model companies creates strong demand for compute, data and all sorts of services. These are the anchor companies in the value chain, surrounded by businesses large and small that serve them. When you meet those businesses, you find that they are doing well too. It is not one company growing in isolation; paying demand is flowing downstream.

Look at China, by contrast. Everyone uses Doubao and DeepSeek, but the number of people who actually pay, and the amount they pay, are different from the United States. The money a company can collect determines how many other businesses that money can support downstream. Having many users does not necessarily mean the commercial ecosystem around them is equally prosperous.

In the United States, some coding-agent companies can still receive backing even if they are not number one, provided their products are differentiated and their revenue is growing quickly. Investors ask whether there is a real business and how it is growing. Ranking second or third does not automatically make a company worthless.

So you cannot use a single index to describe the state of an entire market. Within the same market, some areas are hot, some have already been through a cycle, and some are only getting started.

Dong: Hearing those details feels very different from looking at the Nasdaq and major companies’ share prices every day, then at markets in South Korea, Taiwan and mainland China.

2. “Private markets are still scrambling for compute; public markets are already calculating when defaults might happen”

Raymond: Are there bubbles in American private markets? Yes. One fairly obvious area, in my view, is Neo Labs: the new model laboratories.

For example, you were a researcher at OpenAI or Anthropic. While working there, you develop a new algorithmic idea and think you could build the next OpenAI yourself, so you go out to raise money. Capital was once very accommodating toward teams like these, offering very high valuations at an extremely early stage.

But when the next round arrives and another group of investors has to take over, it is not so easy. The bubble was inflated too much earlier, and later investors are unwilling to step in. I cannot say this segment has already burst, but I can certainly feel cracks. There may still be opportunities, but you cannot treat every laboratory as the next giant.

Robotics feels different again. From what I have observed, Chinese robotics has already gone through a substantial part of its cycle. Leading companies have listed, and their shares have experienced ups and downs. Some related areas in the United States have only become hot recently. Both are technology investing, yet their rhythms are very different.

Some American robotics companies have been doing well for a long time, but enthusiasm for investment in different use cases, data and new laboratories is another matter. By American investors’ standards, I feel capital has not fully entered some areas, so people are moving into them. Before long, we may again see new teams with very high valuations before they have built anything.

“Nothing has been built, but it is already worth a lot”: that is the kind of private-market bubble I mean.

Another area I have not fully figured out is NeoCloud, the emerging compute-cloud providers. I can share observations, but I do not necessarily have a conclusion.

You can clearly see that, although the United States theoretically has a great deal of compute, people still feel there is not enough and continue to compete for it. One intuitive sign that companies such as CoreWeave and Nebius are doing good business is that, at dinner, people are more courteous when they find out you provide compute.

“Do you have any new plans? Could you allocate some compute to me?” Startups struggle to obtain it, and frontier-model companies need even more. The shortage people face is real.

Resource-allocation problems occur even inside large companies. You think your research idea is good, but cannot run the experiments without enough accelerators. You may want to move to another company or raise money on your own. But what exactly is “the compute I deserve”? That is not a simple question either.

It is like someone inside an investment firm telling Dong: “My strategy is very good. Could you allocate another 500 million to me?” Dong would ask: “Why should you get 500 million? Zhang San does not have it, Li Si does not have it, so why you?” When resources are limited, there will always be internal competition.

People in Silicon Valley are still competing for those resources today. If compute capacity keeps being built next year and the year after, could it suddenly become excessive one day? Possibly. But that is a forward-looking question. It should not be conflated with the competition happening today.

Private-market participants are still scrambling for compute, while public-market investors are already looking at CoreWeave’s CDS—credit default swaps—and calculating its credit risk. People in two markets are looking at the same industry but focusing on different questions.

That is why I feel that China’s private markets, China’s public markets, America’s private markets and America’s public markets exist in four different times and spaces. They are parallel universes.

3. “If 2021 was a ten, today might be a seven or eight”

Dong: If the technology industry has a future, why can technology stocks still be in a bubble? I think there may even be a causal relationship: capital sets high expectations precisely because the industry has a future. What is the mood among the people you met who are actually putting money to work?

Raymond: Broadly, they are still optimistic, and there are reasons for that optimism. First, markets have performed well, so people are starting from a relatively favorable level. Second, some people in both private and public markets have actually made money over the past few years.

In private investment especially, when a portfolio company is acquired quickly, paper value turns into money investors can get back. Before that, saying what the company was worth may have referred only to a fundraising valuation. A real exit feels different to the people who provided the capital. When a fund’s DPI—distributions relative to paid-in capital—rises, the next fundraise becomes easier, and LPs, or limited partners, the fund’s capital providers, are more willing to invest again. That feedback encourages investors to back the next company more actively.

A company you invest in might be acquired by a large company eighteen months later. What was a valuation can now be monetized. The capital comes back and is invested again. The cycle turns very quickly, and it feels quite different from the state of private markets in China.

On bubbles specifically, I acknowledge that many technology companies have relatively high valuations. But compared with bubbles we have seen in the past, I do not necessarily think today’s is an especially large one.

I often compare this with 2021. Many stocks traded at high multiples then too, but the main force I felt was liquidity. Today is different: many technologies are undergoing fundamental changes at their foundations.

Those changes may not yet generate especially large returns, but the technological progress itself is more promising than it was then. You know some new things are happening. You cannot simply apply past experience to everything.

If I were to give sentiment and the degree of speculation a very subjective score, with 2021 as a ten, today feels perhaps like a seven or eight to me.

Dong: A-shares might also have been a seven or eight when they were hotter, but have since come back down. The Nasdaq may be in a different state. Even if it is still making new highs, its path may not be as smooth as in its best years. Some investor concerns are already reflected in that.

4. “Our courage comes from the capital behind us”

Dong: What you have just described is actually a complete capital ecosystem. Research, industry and investment reinforce one another, and capital at different stages can connect.

In the United States, different people and pools of money are available at every stage, from very early angel investment through later fundraising to listing and exit. For a while in China, people said “private markets are dead.” It is not that there are no institutions, but the risk capital active across the environment has changed substantially from the past.

There is a rather crude saying: “Liquor gives a coward courage.” For investors like us, where does our courage come from? It comes from the capital behind us.

If that money has little appetite for risk, is judged rigorously over short periods, and demands high annual returns and short-term win rates, how can it give an investor much courage? You may ask that investor to endure volatility for a long time, but the money behind them may not permit it.

When there are few people to take the baton, earlier investors want to list and exit as soon as possible. There are not many other paths to a return besides the stock market. If capital at different stages does not connect, the ecosystem becomes thin, and startups are pushed to move ahead.

People easily assume that capital offers a price because it has already understood the future of the technology. In reality, even founders inside the industry cannot completely predict where they will end up. How long the money behind an investor can last, and what risks it can bear, strongly influence what that investor is willing to believe.

Raymond: Exactly. Industrial development and capital markets are closely connected. I mean capital markets in the broad sense, including everything from angel investment to listings and acquisitions.

Building on Dong’s point, I have a view that may sound counterintuitive: I do not think China’s problem today is that it has a bubble, but that the bubble is not big enough. The title of my previous article was “May There Be Bubbles Everywhere.”

Look at financing and acquisitions in the United States. Very large amounts of money can keep supporting a company’s progress. Behind cases such as Cursor being acquired at a valuation of 60 billion dollars are different rounds of capital, shares in listed companies and expectations for the next stage of the business.

In China, a company reaches a certain stage after early dollar funds have invested, state capital has taken a turn and strategic industry investors have taken a turn. Everyone who can come to the table has done so, and then there is no fresh money left. At that point, the company has to rush toward a listing.

Sometimes a public-market listing by a Chinese company would, in the United States, simply be a financing round led by one fund. The company may not yet be fully mature; the capital behind it just needs an exit.

Not every good American company eventually goes public either. Some are acquired before reaching that point. Counting how many companies ring the opening bell each day may not tell you how active the entire startup market is. You also need to look at acquisitions and other exit channels.

I think China’s acquisition environment is relatively weaker, and two aspects deserve consideration. First, many business leaders are first-generation founders and tend to build things themselves. They see a product and think: “This is not that complicated. I am a programmer too. I can just build one myself.”

Some professional managers in the United States focus more on time to market: how quickly a product can be brought to customers. If an acquisition lets them immediately put a product through their existing distribution channels and grow the business faster, the deal makes sense. They are thinking about how time and existing resources work together, rather than insisting on building everything from scratch.

Second, do listed companies’ own shares have enough purchasing power? When your share price is low, you naturally hesitate to exchange your stock for someone else’s company. Management’s feelings about the same acquisition can be completely different depending on whether its own shares are doing well or badly.

Dong: They lack the capacity for capital expenditure.

Raymond: Yes, and they are reluctant to spend. Industrial development and capital markets often affect each other. If financing is difficult and exit channels are limited, younger companies struggle to grow. Weak share prices at large companies also affect their willingness to acquire others.

But we should not belittle ourselves either. Chinese companies have produced many things over the past two years that have surprised me. When we make our quarterly foundation-model review, one episode may be full of concern that the gap is widening; in the next, we see new models catching up.

The emotions when we record are real. In one episode, we think: “It is over. The gap keeps widening. What do we do?” As Chinese people, seeing our models fall behind makes us anxious. A few months later, new progress arrives and people say: “It looks as though we have caught up again.” It keeps overturning judgments you have only just formed.

Even with all these problems, some founders always manage to achieve results with relatively limited resources. I think those people are remarkable. There are difficulties today, but you can also look forward to further progress.

5. “Where on Earth will all that money come from?”

Dong: Is the scale of American AI capital expenditure evidence of an industry boom, or is it a risk?

Raymond: A great deal has certainly been spent over the past few years. Looking at it today, you feel it is a sign of industry prosperity. For instance, data centers need so many workers, many places struggle to find them, and construction itself faces various constraints. Those needs are right in front of us.

But I understand the concerns too. Move the timeline a few years forward, and the question changes.

Suppose a trillion is spent this year, another trillion next year, and investment continues. If those costs are spread over four or five years, more and more depreciation will accumulate. Without getting into the fine detail, just consider the broad numbers: how much revenue will ultimately be needed to support that much annual depreciation?

At this point, many investors and ordinary people ask: where on Earth will all that money come from?

Once the numbers become large enough, you cannot simply discuss one company gaining another subscriber or another business making API calls. You have to use a much more macro framework: How large is the white-collar market? What is the total wage bill across society? How large is GDP? Has the pie you are drawing become larger than you can comprehend?

If revenue never reaches that scale while many data centers have already been built, supply could exceed demand. Once supply becomes excessive, prices can move very quickly. If demand at the anchor companies falters, companies downstream will also be affected.

So I think the capital market’s concern about this is real and has a basis. I cannot say it is completely unreasonable simply because I see people in Silicon Valley competing for compute today.

Conversely, the other question is: could we actually create revenue on that scale?

A friend once told me that, over the past few years, people often seemed to turn pessimistic in the second half of the year. They had already seen that round of technological progress and did not know what would come next. Then, after a while, another astonishing development appeared.

The spread of coding capabilities through Claude Code, or the changes brought by DeepSeek before it, has prompted more people to use these tools. Technological breakthroughs make people consume more tokens, the basic units in which models process and generate content.

That moves the capabilities available across society up another step. We may not yet be able to imagine the next step.

We should not use the scale of an old civilization to estimate what a new civilization will look like.

Dong: Yes, that is well put. Expectations are being refreshed more quickly too. Just when you think a model has reached its limit, it produces something surprising again. From outside the industry, it is hard to know whether an apparent gap in expectations comes from our own lack of understanding or a genuine change in fundamentals.

6. “Before discussing bubbles, ask whether you have a position”

Dong: If it is hard to predict when a bubble will burst, should ordinary investors still spend a great deal of effort studying it?

My inclination is that, if your daily work and life are already demanding, you should not take on a question this enormous as well—unless you have exceptional curiosity, research ability, data and tools. Otherwise, even if the best economists and investors sat down together, they might not agree on the timing.

High crowding and the existence of a bubble are also different things. Crowding reflects the composition of investors and the concentration of trading more than anything else. A falling market does not immediately prove that the technology has no future, or that a particular bubble has been punctured.

Raymond: I would add something. Identifying bubbles is a perennial topic. If you say, “I think there is a bubble, so I will not invest,” that is certainly a decision you can make. But suppose the market changes dramatically six months later and you realize you have missed it. You still have to decide all over again whether to get in.

I often think about it through a different framework. Before discussing what a bubble is, first ask: Do you think the index is high, or that your own position is too large?

There are at least two situations here. First, you have a position. Second, you do not.

If you hold a position and think it is too high, you need to think specifically about whether to trim it or how else to handle it. If you do not hold a position, the question is whether to enter. If you are not entering now, spend time researching a more suitable level at which to do so.

For example, if my exposure is zero and I want to increase it, the entry point is very important to me. I need to study it. Simply debating when a bubble will burst or when its collapse will be over may not help me act.

In reality, the moment a bubble bursts may also coincide with a market low. It is hard to suddenly build your entire position when everyone is at their most fearful. People say beforehand, “I will buy if it falls.” When that moment actually arrives, they may feel completely different.

So I still want to spend more time learning and trying to understand where tops and bottoms might lie. Even if it means making a guess, you have to try. Otherwise, when the event actually arrives, you will still not know how to respond.

I prefer to think in practical terms and in relation to my own positions, rather than stopping at a judgment about bubbles.

Dong: I take your point to be that preparing to act is more important than endlessly trying to predict an event. Whatever your current situation, first work out what you will do if a certain scenario occurs, how you will protect your money, or how you will obtain what you want.

You can use any decline to prove that your prediction of a bubble was correct, but that does not mean you can act correctly. There is still a large distance between a judgment and the action you ultimately take.

7. “Asset allocation cannot precisely help you sell at the top”

Dong: Listeners also asked what asset allocation actually does. Does it reduce day-to-day volatility, or prevent losses you cannot withstand?

I think it does both. Start with the most basic point: an asset that is difficult to predict and highly volatile should not determine the outcome for your entire household’s wealth.

You may love technology stocks and believe they have a great future, but you do not know where the top or bottom is, and cannot guarantee that every judgment will be correct. Allocation first provides protection so that one difficult decision does not damage the whole.

That protection can then allow you to go further. Someone who invests a large share of their net worth and suffers a sharp decline is not experiencing the same thing as someone who invests only an amount they can bear.

The former watches their entire wealth shrink and struggles to stay calm. The latter may see that investment fall too, but other assets remain relatively stable and the overall loss stays within an acceptable range. That makes it more likely they can keep waiting.

Rebalancing follows the same logic. When one asset class rises and becomes an increasingly large share of the portfolio, you can take some money out and return to a structure you can accept. Do not try to capture every rise from beginning to end, only remembering risk control at the last moment.

What it cannot promise is a precise exit at the top. That is not possible. But it can keep the effect of a major swing on your overall life and state of mind within a more manageable range.

A barbell strategy and asset allocation need not be seen as conflicting. Allocation is the broader concept, and a barbell can be one approach within it. What matters is why you have divided your assets that way, and whether the other parts can protect you if something goes wrong.

8. “It is not a personal-finance class; it is a meditation class”

Dong: Are you still investing in technology stocks yourself?

Raymond: Of course. It is what I do. I think what you just said makes a great deal of sense; it is almost a point about planning your life. Of course we like making money, but a prerequisite is being able to sleep at night.

Everyone has a different sense of what lets them sleep at night. Under normal circumstances, holding some technology stocks does not make me especially uncomfortable.

If you have the time and ability to research companies, you can learn a little more even if you are not a professional investor. Knowing what a company actually does, what its finances look like, the management team’s style and having your own understanding of its future can give you more peace of mind.

For example, if you have spent a great deal of time understanding Nvidia and know why you own it, a sudden fall in the share price may make you ask whether the price has become more attractive, rather than immediately panicking and selling at a loss. The more you understand, the more comfortable you may feel holding it.

The second point is to avoid FOMO—buying because you fear missing out—as much as possible. When everyone else wants to buy, can you do something that follows the crowd a little less? That is difficult for human nature, but I think it is a direction worth trying to move in.

Especially in relatively smaller markets, large price swings can make it harder to restrain yourself. In a particular moment, you think: “No, I have to get on this ride. If I miss it, I will never move up the social ladder.”

Not necessarily. Once you get on, you may move down the ladder instead.

So keeping your hands under control matters too. Ultimately, it resembles training your own state of mind. It is not a personal-finance class; it is a meditation class.

Dong: A listener asked a very practical question: if living expenses are covered by stable cash flow and there are sufficient savings, can they sell all their other assets and invest in the high-risk assets they believe in?

I would still start with the proportions. What exactly does “all” mean? Would losing this money affect your life and state of mind? If you have genuinely arranged everything your household needs and can afford to lose the portion you invest, that is a different situation.

Raymond: I would add that “I will put the money in, stop looking, and consider any loss money I did not need” is often easier to say than to do. I have met enough wealthy people not to believe that having more money necessarily makes this possible.

Some are even more particular: why have I lost another thirty cents here and fifty cents there? In the end, it is like a game, a scoreboard. Watching your score fall is like losing another two marks on an exam. It makes you unhappy.

Loss aversion has a powerful psychological effect. So I do not quite believe that financial comfort alone means someone can truly forget about an investment altogether. Getting to that point is hard.

Dong: I understand. Another common feeling arises when you look back twenty years later and discover that the fluctuations that caused you so much pain were just a small kink in a very long upward curve. Then you regret not having held on all the way.

Many investment questions therefore come down to choosing an approach that fits you. No method can keep you from feeling regret at every point in time.

9. “If A-shares rise after you switch, will you switch back?”

Dong: How should people with losses in technology funds and people preparing to enter approach things differently? I still come back to discipline. Existing holders should first consider what share of their total assets this investment represents and whether the rest is stable. New investors can consider entering in stages and setting an upper limit.

Another question is whether switching from A-shares to American stocks can reduce risk. When we mentioned someone wanting to sell A-shares and buy the Nasdaq, I noticed you smiling. Do you think switching can reduce risk?

Raymond: I would first want to understand whether they already have overseas assets and want to allocate them in the United States, or whether they are buying an ETF linked to that market.

Dong: I think they mean buying an ETF.

Raymond: The American and Chinese markets are very different. But if they have decided to switch, my next question is this: if A-shares suddenly surge afterward, will they switch back?

Going back and forth like that, could they not get hit on both sides?

For some assets, gains are concentrated in a relatively small number of trading days. If you happen to be out of the market then, the long-term result can be very different. But how do you know in advance which days those will be?

Dong: It is the same with A-shares.

Raymond: So, when we say we are optimistic about something over the long term, in a sense we are saying we will remain invested and be there when the important gains occur. If we keep switching, we may miss the rise in this market and the rise in the other one too.

I think we ultimately need to establish an overall framework.

Dong: People also ask whether assets such as the Nasdaq and Bitcoin will be hit simultaneously by liquidity and risk appetite when risk appetite undergoes an extreme reversal. They certainly will.

Things that do not normally appear highly correlated will become highly correlated when risk appetite reverses sharply. It is also difficult to predict when capital will undergo that kind of change. Switching markets therefore does not mean escaping all common risks.

10. “By the time you see the signal, it may already be too late”

Dong: When upstream sectors in the supply chain correct first, is that directly related to a turning point in the industry’s overall cycle? How do changes actually propagate through the AI supply chain you observe?

Raymond: This involves many companies. Let me begin with some relatively simple things that are easier to observe.

We look at rental prices for GPUs such as the H100 and even the A100 as a reference. Those are prices someone in the market is willing to pay for compute. It is understandable that new chips are expensive. If older models are still being used and people are still willing to pay for them, that also carries information.

If compute capacity really starts to become excessive, new supply arrives and users first move toward newer, more suitable equipment, older supply will face pressure. Will rental prices for those older chips change significantly? That is something to track.

Likewise, some memory spot prices are worth watching. The supply chain builds stock and inventory layer by layer. If supply and demand change downstream, the response may become amplified as it propagates upstream. That is what is commonly called the bullwhip effect.

When you see spot prices move, it sometimes means participants in the chain have already begun to realize that waiting any longer to sell may be too late. When many people act together, prices can change very quickly.

But by the time you see it, it may already be late. To see it earlier, you need to go deeper into the supply chain. That becomes more complicated and harder for ordinary investors to carry out. I think these are things worth tracking.

Another example is Citrini Research looking at construction of the Stargate data center. They sent drones to film the site in considerable detail. You can see what equipment is connected outside the buildings, which logos appear on it, which areas are under construction and how far along they are.

When people discuss whether a data center can be built, they worry about power, equipment and cooling. What you can learn by actually looking at the construction site differs from what you get by discussing it in front of a screen.

What struck me was how detailed the research had become: identifying equipment and conditions on site. To get one step ahead in the market, professional investors will commit substantial resources and send drones to monitor things almost as though they were fighting a war.

For ordinary listeners, that is a heavy burden. Starting with price information we can obtain consistently already adds another layer of understanding beyond watching an index.

Dong: Similar methods exist in other investment fields. Institutions studying agricultural commodity prices use drones, satellites and remote-sensing data to examine weather, crop growth and yields. They cannot send researchers across that much farmland, so they have to use these tools.

It is therefore very difficult to understand something at its roots and find leading information the market has not yet seen. Institutions with greater financial resources and research capabilities may already be doing that work.

You can certainly study industry fundamentals out of curiosity. But there are still many steps between researching a company more deeply and managing your own wealth well. Finding a leading indicator does not solve all the decisions that follow.

11. “I started researching AI and ended up researching gas turbines”

Dong: People also ask which matters more: timing, stock selection or asset allocation. I think they serve different functions, so you cannot choose only one. Allocation protects the whole; it is hard to get timing consistently right; and stock selection depends on whether a particular industry offers room to make a difference.

What is your own impression? How large are the differences between individual technology companies today?

Raymond: Rather than discuss particular companies, I can talk about the overall market. American stocks have offered many opportunities over the past two years, but they have also been difficult to trade, with more pronounced volatility than before.

Some influences come from outside. First it might be tariffs, then war, then oil prices. You encounter many things beyond what earlier textbooks or your own experience prepared you for. Some changes exceed my understanding, yet the market seems able to keep moving forward.

Another change is the speed of social media. The time between information appearing and prices fully reflecting it can be very short. You may not have much room to react. Sometimes you think you have found alpha—an opportunity to outperform the market—only to discover later that it is not so easy to capture.

But I also think people can make greater use of AI tools to learn. If you do not understand a company, even asking a few questions every day will teach you a little more. Progress comes step by step. Over time, your feel for markets, industries and companies will change.

Especially over the past two years, what the term “technology stocks” contains has changed dramatically. For the preceding decade, technology companies meant Tencent, WeChat, Alibaba and Xiaohongshu. These were products we used ourselves. At least we knew what they did.

That is no longer the case. Now people constantly talk about optical modules and PCBs, or printed circuit boards. A year ago, many people may not even have been familiar with those three letters; now they are being forced to become PCB experts.

The United States is similar. Semiconductors have so many specialized segments, and then there is energy. Studying AI draws you into subjects you never imagined studying before.

A couple of days ago, I made an episode about gas turbines in the United States. Two years ago, I could never have imagined that studying AI and technology stocks would eventually lead me to gas turbines.

But life really has changed. This is the new normal: everyone needs to learn new skills.

If you're working on this too — or you think we've got it wrong — write to us at [email protected]; if you'd rather not write, just leave your email below.