BEDROCK Podcast E20 - Looking Ahead to 2026, AI Investment Trends: US-China Friction, Bubble Divergence, and Niche Opportunities?

As a global investment firm, we continue to bring our audience the latest observations on global investment themes. In this episode, we are joined by TC (CIO) and researcher Tracy to discuss the recently heated debate around an “AI bubble” and related industry topics.

Topic 1: The Federal Reserve’s December Rate Decision and Outlook for 2026

Claire:Before we dive into AI, let’s start with a topic that has drawn even more attention over the past couple of days. The Federal Reserve released its rate decision and Summary of Economic Projections earlier this morning. We’d like to first hear TC’s perspective, since TC regularly shares our macro views on our public platform.

TC:This can be described as a "hawkish cup"—while interest rates were cut, the guidance for the yield curve next year remains hawkish. The current guidance only addresses one rate cut, but there's significant disagreement, particularly regarding the opinions of voting committee members on rate cuts in 2026. A common analogy is "a string of candied hawthorns"—each point has several representatives, but no clear consensus. The extent of the rate cut next year, the current state of the US economy, and the future direction of interest rates are all points of contention. On one hand, the economy and employment are facing pressure, but on the other hand, inflation cannot immediately return to the Fed's 2% target. Coupled with multiple rate cuts this year, the significant disagreement on whether to further lower interest rates is understandable. Because if rates are cut again, the 10-year Treasury yield could fall further below the inflation target. However, I think that the Fed's rate hikes and cuts next year are unlikely to be a major factor influencing market expectations. In other words, the Fed's interest rate decisions are unlikely to be a critical variable and will likely fluctuate. Overall, we believe that the resilience of the US economy and the possibility of inflation still exist; you could even argue that we've entered a relatively stable phase. Therefore, while macroeconomic disruptions are not nonexistent next year, we may want to focus more on structural opportunities—the investment opportunities themselves—rather than relying on expectations of a sudden surge in macro liquidity. This is my initial thought; Tracy, you can add more details.

Tracy:I completely agree. When people talk about entering a rate-cutting cycle, they easily think of the two past instances: the 2008 financial crisis and the pandemic. Both of those involved financial crises or other crises, so the rate cuts were coupled with significant central bank easing measures. Therefore, people readily associate rate cuts with those two events and assume it will be a major boon for investment. However, I think the situation is completely different now. In a crisis, the Federal Reserve desperately needs to reverse the momentum, which was a huge shock. But now there are various constraints. For example, inflation hasn't simply declined smoothly; and while the economy is slowing, it's only just begun to slow, starting from a very strong level. Furthermore, if we were to engage in the same kind of massive stimulus as before, we would face fiscal constraints and the risk of a weakening dollar. Significant rate cuts would have many side effects. Therefore, given these significant side effects, it's unreasonable to insist on large-scale rate cuts and quantitative easing (QE) at a rapid pace. The US decision-making process is based on a multi-person committee, not a single person making the decision. Therefore, I think there is a high probability that the future interest rate change (resolution) will be relatively moderate.

Claire:Okay, we can hear that BEDROCK (we) still believe that the Fed's monetary policy next year is unlikely to be a primary or significantly impactful factor in our investment decisions.

Topic 2: Investing in Chinese and U.S. Chip Companies Amid Sino-U.S. Tensions

Claire:TC has already started mentioning Trump, and we're discussing AI today. Just yesterday, Trump and his US government approved Nvidia's sale of H200 chips to China. We'd also like to know how the current friction in US-China relations is affecting AI industry investments, especially in computing power, in both countries. Previously, people might have focused more on the technological gap in computing power between China and the US, but given the friction, how will investors choose—for example, whether to invest in AI and semiconductors in the A-share market or in the US?

TC:I think Chinese semiconductor companies, such as Moore's Threads which recently went public and Cambricon which has seen significant growth and is highly sought after, filled the market gap during the period when Nvidia was banned. From this perspective, we agree that the market believes they are very capable. However, objectively speaking, I think we still need to recognize that a technological gap still exists between them and advanced American companies. The current US strategy and TSMC's strategy of opening up semiconductors to mainland China are quite similar: opening up relatively outdated products to China, but these outdated products still have technological competitiveness and influence in China, while preventing the most advanced cards from flowing to China. Therefore, I think we cannot underestimate the competitiveness of Nvidia's previous generation cards. So, the landscape next year, or for a longer period thereafter, as long as the US and China don't completely break off relations, will likely be a similar situation: internationally advanced previous generation products will compete in China, but will still be competitive because they may have advantages in software and ecosystem, while Chinese chip manufacturers are making progress, creating a competitive environment.

Tracy:Let me add something. After all, the semiconductor industry is now of national strategic significance. The future landscape of the semiconductor industry isn't solely determined by the market; it also requires consideration of many national interests. While the US has allowed some Nvidia chip exports to China, this is a two-way street. The attitude of the Chinese government also plays a crucial role, and companies will ultimately make their choices based on various factors. For example, without interference, even if Nvidia chips are a generation behind, they might still be purchased by Chinese companies. However, if there are long-term concerns (such as the Chinese government's stance), companies' choices will change. In this situation, Chinese companies won't solely rely on Nvidia in their chip choices; they still need to develop their own semiconductor industry. This, to some extent, provides opportunities for Chinese semiconductor companies. In short, the semiconductor industry isn't purely market-driven, and China's market demand will continue to have a significant influence on global semiconductor demand in the long term. Therefore, if Chinese chip companies can establish a relative advantage in the Chinese market, the demand from China will be significant, not only domestically but also in the global chip market. This opportunity warrants continued attention. From this perspective, it will also influence investors' pricing decisions for Nvidia. If, in the long run, investors continue to believe that Nvidia's chip sales in China will face various obstacles, they will discount the proportion of Nvidia's future revenue from the Chinese market. The fact that H200 chips can be sold in China will have a relatively significant short-term impact on Nvidia, but its long-term pricing impact will be relatively limited.

Topic 3: What are the sources and disagreements regarding the AI ​​bubble?

Claire:We just discussed the new developments in the computing power segment of the AI ​​industry chain, specifically the progress of Nvidia and Moore's Threads. Today, we'll mainly analyze hot topics in the AI ​​industry. Recently, many people have been asking, "Will the AI ​​bubble burst?" This big question can be broken down into many smaller questions, such as: What is the meaning of this bubble? If there is a bubble, where is it located? What factors caused it? If this bubble is going to burst, what core variables or conditions will cause it to collapse?

Let’s start with TC, and then Tracy can expand on specific points.

TC:Let me first share my thoughts on the AI ​​bubble: In my understanding, AI's ability to solve real-world problems has reached a very high level. For example, in intelligence tests or problem-solving tests—any test that doesn't require highly forward-thinking or creative imagination—AI's problem-solving skills are comparable to a graduate student or a very experienced professional. What it needs more of is engineering modification and adaptation to various industries worldwide. We believe this process will take 10 years, or even longer, to gradually permeate people's lives, just as the internet improved people's lives in the past. We are very confident in this. Therefore, in terms of the long-term application prospects and market space of AI, we don't think it's a bubble at all; it's still in the early stages of industry development. After all, AI has only been widely used in people's lives for about two years. However, the current focus on and debate surrounding the AI ​​bubble stems from the perception that the industry's capital expenditure is too high upfront, especially relative to current application revenue. Therefore, I believe the discussion on the AI ​​bubble should be divided into two parts: AI applications and AI capital expenditure (CAPEX).

• From a CAPEX perspective, money invested now will only be recouped in the future. Is investing hundreds of billions annually too much, or too early? I believe this is the core of the controversy. Comparing capital expenditure to company revenue, which we call Capex Intensity, the intensity of investment is extremely high, several times the revenue generated by the company's business. Although AI company revenue has also doubled—for example, OpenAI's revenue has doubled—its overall revenue remains relatively small, which is a major point of contention. We believe that in the long run, Capex Intensity will inevitably revert to a proportion of revenue generated, rather than several times the size of revenue. However, it's unclear when this will happen, or whether it will remain high, allowing revenue to double or continue to grow, thus lowering the long-term proportion. This depends on many variables, such as whether application growth is faster, and macroeconomic factors, including interest rates—all these factors have an impact. We tend to believe that betting on the total amount (i.e., the total capital expenditure of the AI ​​industry) carries relatively low certainty, as mentioned earlier, given that capex intensity is generally quite high.

• Therefore, while maintaining confidence in the application and development of AI, we tend to be relatively cautious about the overall scale, focusing more on whether there are structural opportunities within the massive annual volume of several hundred billion. For example, as AI clusters grow larger, are there structural growth opportunities in areas such as communication, storage, and liquid cooling? Grasping these structural opportunities and identifying companies with growth potential is more meaningful than simply betting on the beta of the AI ​​industry. Tracy, you can add further details.

Tracy:Let me add something. When discussing the bubble issue, one thing that's easily brought up is the widely discussed financing chart of companies like Oracle, OpenAI, and Nvidia, showing how they "step on their own two feet." This chart depicts the leveraged relationships between large modeling companies, chip companies, and other funding providers in the market. Therefore, people easily draw parallels between this chart and the disastrous situation that occurred in China's real estate industry, where excessive leverage was followed by deleveraging.

1. When people compare the AI ​​industry bubble to the real estate bubble, they overlook one point: the most crucial exogenous factor is the scale of AI applications and the revenue they generate. Currently, the revenue of each model company is still very small, at the level of tens of billions or hundreds of billions. The key is to judge whether the AI ​​industry, whether through leverage or mutual financing, can eventually grow from the current hundreds of billions to a trillion-dollar level. This is very different from real estate. Real estate is a model where industry revenue does not expand, housing prices fall, and the overall industry collapses. But AI is a model where industry revenue grows exponentially. As TC just mentioned, in terms of model capabilities and industrialization prospects, there is still no obvious ceiling. If we calculate in various ways, it is easy to calculate an industry revenue scale of trillions. Is there a bubble in the AI ​​industry? On the bright side, its prospects are that of an industry with trillions of revenue.

2. So, everyone can calculate how much money is reasonable to spend if it can generate trillions of revenue in the future? For example, we can easily see how many years in the future everyone will spend 3 trillion in capital expenditure CAPEX. How much revenue do you need to generate for this 3 trillion CAPEX? Let's do the math. If the CAPEX is all invested in by others and you rent the computing power, you will need to spend about 1 trillion per year. If you can generate more than 1 trillion in revenue, you can use this 1 trillion to buy computing power without making a profit. That's roughly how the math is calculated. Based on our current observations, the AI ​​industry is still in a process of exponential revenue growth, and it is expected to reach trillions of dollars in revenue in the next few years. It is reasonable for it to spend trillions of dollars in the next few years. However, a very important premise is whether the AI ​​industry can actually achieve trillions of dollars in revenue.

3. In a recent interview, Ray Dalio mentioned whether there is a bubble in AI investment. He said, from the perspective of a CEO, how do you make decisions? For a CEO, the past few years have seen 10 times the revenue growth every year. Now you definitely need to make some preparations for the future. Even if you may not be able to continue the 10 times revenue growth in the future, it may still be a very fast growth. Then your current infrastructure is not enough. How do you plan your investment? The current infrastructure investment is not like you can build a data center today and it will be completed tomorrow. It may take two years to complete. So as a CEO, you must make some investments in advance for the explosive growth of revenue in the future. Therefore, advance investment is very necessary.

4. Now, it may take trillions or even several trillions of dollars in cumulative capital expenditure. Will there be enough money for this in the next few years? If we count on our fingers, in fact, if we rely on those large cloud companies, they have a strong ability to generate cash flow. They used to have relatively little debt, and they can increase their debt. So, relying on those large companies alone, they may be able to support trillions of yuan in investment over the next few years. In addition, there is some social financing. If we calculate it fully, we can still hope to solve the problem of funding sources of several trillion yuan. So, the "bubble" is not about the company running out of money after investing to a certain stage. The key is whether the belief in the trillion-yuan level of the future industry is still there, and whether the industry is developing in this direction.

5. I think it is reasonable for everyone to discuss the AI ​​bubble. Now it does have some aspects that look like a bubble. For example, when model companies only have tens of billions or hundreds of billions of yuan in revenue, they have already announced that they will invest several trillion or even trillions of yuan. What are their financing channels? We just calculated that one financing channel is revenue support, and another financing channel is to raise funds in the bond market. But what do bond investors look at? They look at the future cash flow, whether it can pay my interest and my principal. However, if you are currently at the level of several hundred billion in revenue, and you want bond investors to believe that you will have trillions in revenue in the future, you are essentially asking bond investors to act as VCs. This is what people will question, and I think it makes a lot of sense.

6. Finally, the fluctuating attitude of the market towards the industry landscape of large AI model companies can also help to understand the fluctuating attitude of the market towards the AI ​​industry bubble. We can estimate how much overall investment in the AI ​​industry is reasonable from the changes in the large AI model landscape. For example, after Google's Gemini 3.0 was launched some time ago, everyone thought that Google would take the lead immediately and that OpenAI would go bankrupt. In fact, the capital expenditure of these large model companies also has an impact on whether there is a bubble in AI industry investment. For example, there are currently 4 large model companies investing funds to compete, such as Google, OpenAI, Anthropic, and xAI. If the winner of the large model is already very clear, and only Google is left, then the scale of investment by Google alone and the investment by the 4 model companies will definitely be different. If we take an extreme view, if OpenAI has already gone bankrupt today, then this industry does not need to invest trillions in capital expenditure, then the bubble is really a bubble. Since it will take a long time for these four major model companies to determine the winner, continued capital expenditure is essential. From this perspective, the AI ​​industry may not be such a big bubble.

Topic 4: AI company management now needs to lobby more investors to pour money in.

Claire:Thank you, Tracy. You just answered a question we were going to discuss later: whether the overall landscape of the AI ​​industry is set. It seems the outcome is still uncertain. To summarize, we believe that a "bubble" essentially describes something that has reached its limit and is about to burst. If we're looking at the trillion-dollar market potential brought by the penetration and application of AI, we're still in the early stages and far from reaching that limit. Therefore, we don't think the current AI industry is a bubble. However, I understand why many people are worried about a bubble. In the past few years, internet giants, whether it's Metacritic in the US or established companies like Tencent in China, have offered buybacks and dividends to value investors when their AI capital expenditures were too high. Investors found this acceptable. Now, the market is concerned about whether these internet giants can continue to provide such good returns to valuation-sensitive investors. Therefore, for these more cautious investors, the AI ​​industry may already have a bubble.

TC:To add to what Claire and Tracy just said, I'd like to point out that, judging from the current capabilities of their models, large AI model companies still have much to do, such as long-text and multimodal computing, where they can improve. Through our interactions with these companies, we've observed that they have a very strong incentive and willingness to continue investing if they can secure further funding. However, their existing cash flow, including current reserves and remaining funds from previous rounds of financing, is far from sufficient to support further expansion of capital expenditures. Therefore, the current issue also involves the other investor concerns mentioned by Claire and Tracy: AI companies need to lobby others to continue investing. During this lobbying process, these newly acquired investors may not necessarily share the same deep-seated belief in the future as the AI ​​model companies. For example, the difficulties encountered by Oracle (ChipCloud) are closely related to this; because their own cash flow can no longer support such high capital expenditures, people are questioning whether these investors can continue to invest. I think this question is entirely reasonable. AI companies will indeed face many challenges in this regard, and these doubts will likely recur. For example, if their model technology advances in a couple of days, or their ARR (Annual Recurring Revenue) increases, and new investors regain confidence and their belief in AI strengthens, then if there are more doubts about AI applications, AI companies will face difficulties in fundraising, and the talk of an AI bubble will resurface. Given such high Capex intensity, it's difficult to definitively say whether AI is a bubble or not; both sides have ample evidence. On one hand, ARR is rising rapidly; on the other hand, AI model companies themselves are struggling with insufficient cash flow. It's understandable that new investors don't have the same level of confidence when seeking funding. Therefore, the debate about an AI industry bubble will continue, but from this perspective, it can also be argued that we haven't reached the stage of widespread public enthusiasm for the AI ​​industry.

Tracy: Yes. It's quite interesting. Not long ago, everyone thought OpenAI was the king, placing huge orders everywhere, expanding the competition among large model companies from vying for users and improving model capabilities to building an industry ecosystem. For a time, everyone thought OpenAI was amazing, but after a few days, Google demonstrated its powerful hardware capabilities, model capabilities, and ecosystem integration capabilities, and seemed to be in the lead. Suddenly, public opinion shifted to OpenAI, and people thought the company was going to go bankrupt. So, in this situation, it's quite difficult to keep investors from all walks of life firmly committed to investing so much money. In the process of competition, there will likely be various disturbances and doubts.

Topic 5: Structural Opportunities Are More Attractive Than Beta

TC:Therefore, I think it's better to seize opportunities presented by Alpha. For example, the usage of next-generation Nvidia cards or next-generation TPU solutions could increase by 2 or 3 times. If opportunities like these can be found in connectivity, storage, and other areas, it will reduce concerns about the fluctuations in the AI ​​industry's beta. Regardless of the debate over whether AI is a bubble, I'm confident about one thing: even if the industry experiences setbacks, it's not a case of current investment going unused in end-users. Every penny invested and every GPU invested is being fully utilized, or even insufficient—at least currently, it's insufficient. So, even if predictions for the future of AI fluctuate or even decline, it's not a crash, not a zero-sum game. AI isn't a purely conceptual bubble where no one is using it and all the cards are idle. AI isn't a conceptual bubble in that sense. The real bubble we're seeing now is simply a question of whether the pace of investment is too fast, whether we need to slow down, whether we need to increase or decrease our investment. This is completely different from a mirage-like bubble. First, we need to define this clearly. So, if it's not a mirage-like bubble, but just a matter of pace, we can seize more structural opportunities. Because the overall market size is already very large, it's already several hundred billion US dollars. Even if Tracy mentioned that another 3 trillion will be invested in the next few years, whether that 3 trillion is revised down to 2.5 trillion or 2.7 trillion, it's not that important for some investment opportunities, because as long as we firmly believe that this 3 trillion won't disappear overnight, that no one will invest anymore.

Tracy:If we look at the AI ​​industry in a hierarchical diagram, the top layer is applications, the middle layer might be models, and the foundation is infrastructure—these are some rough classifications. Just looking at infrastructure, as we just discussed, it's currently worth hundreds of billions, and could potentially reach trillions in the future. And if you break it down into the various sub-segments involved, there are many. For example, will ASIC chips take off? Will storage take off? For instance, if you're building a data center, you need graphics cards, storage, network connectivity, cooling, and power supplies—so there are many things you can invest in.

TC:For example, we just mentioned that there's a lot of debate about infrastructure, but I think there's no debate about AI applications. There's definitely still a lot of room for growth; it's just that it's not clear yet who will capture the market and who will get the pie. For example, there's a lot of debate right now about whether Google or OpenAI will get the pie. I think the space for AI applications is still very large

Tracy:Another point is that currently, the monetization of a large number of AI applications is still primarily To C (consumer-facing), meaning they are monetized through chatbox subscriptions. If this ultimately becomes a multi-trillion dollar application market, I believe it's impossible to rely solely on the business model of users interacting with chatboxes and paying subscription fees. There will definitely be new developments, such as services for enterprises and other forms of monetization. I think the probability of many new opportunities emerging from this is very high.

Claire:Tracy just reminded me that AI ultimately has many applications. Previously, we might have seen small companies specializing in AI agents for specific vertical markets, building a model and implementing a viable business model to generate revenue from end-users. However, now we see large internet platforms like Alibaba, ByteDance, and Baidu—these platforms already possess significant traffic and monetization mechanisms. Their self-developed models give them a natural advantage in directly accessing the traffic pool of specific vertical agents. Tracy, do you think any promising small companies specializing in vertical AI agents will emerge? If so, in which field? If not, among the giants we've mentioned, which one do you think will succeed, or how would you predict which one will emerge victorious?

Tracy:First of all, if this market is a huge multi-trillion dollar market, I find it hard to imagine that every single penny of it would be earned by just a few large companies. This involves the difference between easy and difficult questions in investment. For example, the relatively easy question now is about these companies that build basic models. Many of them not only have modeling capabilities, but they also already have a large user base. For instance, OpenAI now has 1 billion users. This user base is enormous, and they use it daily with high engagement, so its advantages are already very obvious. In addition, Google already has a very strong ecosystem, which has a massive user base and a lot of data, much of which is unique. This advantage is also very clear. So, the easy question now is about the AI ​​application foundation we just discussed, and then the infrastructure. If you select these leading model companies and say they will be very successful in the future, that's likely to be correct. This is an easy question. If this market is worth trillions, these large companies might dominate it. If it's a few trillion, then there will definitely be many opportunities left, though they may be challenging because the companies that will capitalize on these opportunities haven't yet emerged. You don't know who they will be, but I believe there's a very high probability that companies outside of these large model companies will seize these opportunities. For example, while most data is concentrated in the hands of these large companies, a lot of data is not. For instance, enterprise data may have other considerations, such as privacy or security, so this data might not be handed over to model companies. This will create new demands, which may be met by new companies, or by companies that currently provide excellent enterprise services. It's a challenge right now, and these opportunities may not have emerged yet, but when enterprise-level applications develop, we can seize these opportunities.

Topic 6: How should we view the power industry investment opportunities brought about by the current hot AI industry?

Claire:To summarize, our current focus is still on the simpler aspects of AI industry investment. Both TC and Tracy have been emphasizing seizing more niche opportunities. I'm reminded of a recent warning from Microsoft CEO Satya Nadella, who stated that "the problem facing the AI ​​industry right now isn't computing power, but electricity." Therefore, we've seen some people viewing energy and power stocks as investment opportunities lately. I'd like to hear your thoughts on "the energy investment opportunities brought about by AI in the power sector?" Or, have you been paying attention to these opportunities and considered participating?

TC:We've been looking at the power sector for a while now, not just the generation side, but also the consumption side, and how to reduce energy consumption. I think the power sector is a complex and challenging topic. Our company's goal is to find companies that generate compound returns, which requires them to consistently demonstrate compounded returns on the consumption side. For example, the blockchain mining companies that have been hyped in the market are essentially a one-time monetization of electricity resources. It's difficult for us to fully incorporate these types of companies into our framework. Traditional power companies, including some with thermal and nuclear power, also face significant challenges. While they can monetize their electricity resources and achieve revenue growth, expansion is difficult. They already have several thermal power plants, and it's hard to acquire more. Therefore, many are now shifting towards gas turbines and more flexible methods. Finding good targets in the power industry is something we're working hard to achieve, but it's likely difficult to pinpoint. After all, the demand for electricity is strong, but extremely dispersed, making it difficult for a monopolistic power company to emerge in the AI ​​era, unlike the large platforms that can emerge in the internet sector.

Claire:Does Tracy have any other ideas about AI-powered investment opportunities?

Tracy:There's not much to add, but I can share something. For example, consider a 1Gwatt data center. Due to the performance iterations of NVIDIA cards and continuous optimization of other aspects of the data center system, such as network and cooling, a phenomenon occurs: each generation of data centers may double or triple its token processing or task processing capacity compared to the previous generation. However, it's still just 1Gwatt. If you run it at full capacity, it's 1Gwatt multiplied by time and electricity price. That is, with the same power consumption, the task processing capacity of a new data center doubles or triples with each generation. For us, it's probably easier to find companies whose demand growth can, to some extent, exceed the value growth of cards, and whose improvement can surpass that. This is the source of alpha, the source of structural opportunities. Looking at it now, we believe storage is this kind of alpha opportunity; whether electricity is an option is another matter.

Claire:It is indeed more related to our investment framework. As Tracy just mentioned, we may be more focused on opportunities in storage and network connectivity than in the power sector.

Topic 7: The landscape of large model companies is still unsettled; what are the advantages of each?

Tracy:Regarding the question of whether the competition among model companies is settled, which I only briefly mentioned earlier, I don't think it's settled yet. If we were to elaborate on the advantages of each model company, we could discuss it in more detail. In this industry, some companies are experiencing rapid revenue growth. For example, Anthropic is very strong and clearly ahead in the B2B and coding fields, with its revenue growing tenfold annually. Although it's small in size, its growth rate is very high. Do you think it will leave the game? Not at all. Then there's OpenAI, which is criticized for creating a huge bubble, attracting everyone, but now it's fallen behind. However, the model competition is fierce. Now Google has released a model that seems to be ahead. Does that mean OpenAI won't take any action? Actually, I believe OpenAI is also rapidly iterating its model capabilities. The outcome for large model companies isn't solely determined by whether their models are currently leading or lagging behind. OpenAI has already built some moats. For example, as I just mentioned, it already has 1 billion users, who use it every week, resulting in high user stickiness. Once you develop a habit, if you randomly ask an ordinary person, can they clearly tell you whether Gemini or Chatgpt is better? I think most people can't tell the difference. So if they can't tell the difference, then among the billion users, aren't the majority still highly engaged? Does that mean OpenAI is finished today? While its revenue hasn't grown tenfold every year, it's still growing exponentially. And then there's Google, whose leading position is undeniable. And even xAI, a very small but relatively recent startup in the large modeling field, has it been eliminated? I don't think we should reach this conclusion. I believe that among the few of us here, Gork (xAI's model) has very little user data and very little revenue. From this perspective, you might think it's not doing well. However, if you look closely, you'll find something very interesting: Gork's positioning is very different from other model companies. For example, it has X (social media) data. While the volume of X data may not be huge, compared to the data obtained by other model companies like Google, X data is highly unique—it's real-time data. If you want to understand what's happening immediately and make analysis and judgments based on that, then Gork might actually be better. Furthermore, Elon Musk's goal isn't like other large companies that pursue political correctness or positive values; he pursues ultimate truth. Given this different set of values, could xAI, as a large model, have the opportunity to gain access to unique scenarios and unique user needs within its unique positioning? I think it's hard to say. Even if xAI collaborates with Tesla to develop physical intelligence, will Google necessarily be stronger than OpenAI or xAI when it comes to physical intelligence? I don't think we can draw that conclusion. So, with four major modeling companies now, it's difficult to say which one is weak.

Claire: Tracy just mentioned that the overall model landscape is still uncertain, and she also explained how we understand the leading companies we're paying close attention to, and even the model companies that others might not be so optimistic about, and what unique opportunities we see. This shows that we at BEDROCK really break down things very meticulously when making investments, and that our daily research work is very thorough. And let me give ourselves a little self-promotion. In summary, considering that the general public has only been discussing the AI ​​industry since the end of 2022, it's only been three years (Tracy added that the industry has been doing this for many years). We feel that many issues, if examined closely, haven't reached a point where there are definitive conclusions yet.

Topic 8: Looking Ahead to 2026—Is Tech Still Attractive?

Claire: We've basically covered today's topic. AI is a technology sector, and our company has always been quite adept at global technology investments. With the end of 2025 approaching, everyone is thinking about investment opportunities in 2026. I'd like to ask TC first: from the perspective of an investor skilled in technology investment, especially with a global vision, do you think there will be better investment opportunities than technology next year? Or, if we continue to allocate our portfolio to technology, will there be any shifts to specific sub-sectors? Could you share some extended insights on this?

TC:First, for our company, which operates on a single-strategy fund, we primarily invest in one type of company – what we call compound growth. We spent a considerable amount of time searching for these companies, and we believe they are most likely in the technology and consumer sectors, especially technology, which may account for about 80%. This situation isn't expected to change drastically this year or next. Our investment horizon isn't defined by years, especially considering the significant growth potential of AI, as discussed earlier. Furthermore, other technology sectors we invest in also have great potential, so I don't think 2026 will be any different; in fact, some new developments in technology sectors might even exceed our expectations. I think technology and consumer sectors will definitely remain our primary investment focus. As for other sectors, I don't think they fit our company's investment framework. It's difficult for me to determine whether their investment opportunities will surpass or not surpass those of technology. For example, everyone knows that gold prices have risen significantly this year. If I were to predict whether gold would rise or fall at the end of 2024, I would have absolutely no idea, and it would be difficult to reach such a conclusion. Therefore, it's hard for me to conclude that technology will definitely be the best next year, or definitely not. My approach is simply to suggest that if we're looking for companies with high compound growth potential and better long-term returns, the technology and consumer sectors offer the most opportunities – the places with the most potential. However, as everyone knows, there are many fluctuations across various industries every year, and many unexpected opportunities arise. Therefore, I don't think I can give you any definitive advice on this.

Tracy:To add to that, we've been focusing on technology and consumer goods. Consumer goods haven't performed particularly well this year because they're currently overshadowed by the looming threat of unemployment due to AI, especially among the potential main force of unemployment – ​​the middle class. This group represents a significant structural opportunity in our past consumer investments: companies that effectively cater to the emerging niche needs of the middle class. At least this year, the middle class is constantly worried about job and income issues, impacting their long-term spending power and significantly affecting consumer investments. However, looking towards 2026, many things will likely materialize, making the impact on consumption clearer. Given the strong negative expectations for 2025, many consumer companies have already seen significant declines, potentially creating a lower starting point. I think there might still be some opportunities in this area next year.

TC:Another point I've always wanted to make is that we're quite interested in internet and technology opportunities in various emerging markets. In the past few years, because the US dollar has been exceptionally strong and US interest rates have been high, all global capital has flowed into the US. The US also has many technology companies, like those in AI, that have absorbed a large portion of global funds. Will this area see some changes next year? I don't know, but I want to say that many companies still have very good fundamentals. However, because of the situation described earlier where US stocks have absorbed most of the global capital in the past few years, their valuations have remained relatively low. From this perspective, we can also find some areas with great potential.

Claire:To summarize for the audience, from a diversified asset allocation perspective, there might be some other opportunities next year. However, because of our main investment approach and research perspective, we don't look at a single year or sector for a better opportunity and then immediately switch industry allocations. We still believe that the technology and consumer sectors have many companies with long-term competitive advantages and growth prospects that can generate compound returns. Therefore, we will continue to adhere to this framework to find good companies for our investments.

TC:Simply put, the future of the world will still be driven by technology. Furthermore, the tech industry naturally fosters the formation of monopolies, so it's not without reason that many exceptionally high-performing companies with compounded returns over the past two or three decades have originated from technology-driven consumer goods. I believe many trends will continue to repeat themselves in the future, leading to monopolies and further growth. Many key factors in this analysis haven't changed significantly, and I think focusing on these trends is crucial. As for other areas, I can't say whether there are opportunities; perhaps there are, perhaps not.

Claire:Okay, thanks TC for the addition, it further reinforces our confidence in investing in the future of technology. That concludes today's podcast. Thank you for listening! We also welcome you to continue following BEDROCK Investments in the future, where we'll bring you more interesting, fresh, and in-depth investment information! Bye!