Zhou Junzhi and Chen Weibin discuss A-shares: The bull market is still in its early stages, AI is the long-term main line, and liquid cooling is expected to grow by 500% by 2025 or take over from optical modules

On September 25th, Chen Weibin, General Manager of the Equity Research Department at Soochow Fund, and Zhou Junzhi, Chief Macro Analyst at CITIC Securities, visited the Wall Street Journal live broadcast room "Celebrity Guest Lounge," discussing the driving forces behind the bull market and the new round of technology themes. For the live replay, please click → Where is the bull market now? Q1: The economic data in the third quarter has slowed down, but the stock market remains strong. What are the main driving forces behind this? Policy, liquidity, or investor expectations? Zhou Junzhi: If we consider this round of market as the starting point of a bull market, it can be traced back to the September 24th market last year. At that time, the trend was already quite clear, but what really excited market sentiment was after July. The economic data in August was generally weak, with investment, consumption, fiscal, and social financing all declining compared to July, yet market sentiment remained high. This led many to feel that the stock market was diverging from the fundamentals. Thus, an explanation emerged: is this a "water buffalo" market, completely driven by liquidity? In fact, if it is a bull market, the early stage is often indeed driven by liquidity, with valuation repairs occurring faster than fundamental improvements. Only then does it enter the profit recovery phase. This is the first point. Secondly, the performance of the bond market also confirms this. Since July, the sentiment in the bond market has been relatively weak, indicating that market expectations for economic recovery are consistent with the direction of the stock market. Although macro data is declining, compared to the extremely pessimistic pricing in April, the current data is not that bad. In April, Trump's tariff increases triggered severe fluctuations in stocks, bonds, and commodities, with pessimistic expectations fully priced in. However, subsequent export and price data showed resilience, and the pessimism did not materialize, highlighting the economy's resilience. From a longer-term perspective, the worst phase of the Chinese economy may have already passed. The drag from real estate is converging, and exports remain strong amid the trade war, supported by capacity going overseas and capital goods exports. The market is beginning to realize that the most pressured moments for the economy have passed, and the future direction is improvement. Price data also confirms this. The year-on-year negative value of PPI has significantly converged, and core inflation remains robust. The market is more focused on signals of "emerging from deflation." Overall, the recovery of risk appetite in the stock market and the convergence of bond market interest rates indicate that the market is not completely diverging from the fundamentals: first, the extremely pessimistic pricing in April has been corrected, and second, in the long term, the Chinese economy has indeed emerged from the trough. Q2: From an industry perspective, what phenomena have you observed that can support this judgment? Chen Weibin: From the perspective of fund managers, the three core factors in the equity market are fundamentals, liquidity, and risk appetite, and the dominant factor in this round is undoubtedly liquidity. The current economic fundamentals are still hovering at the bottom, with weak consumption, yet the stock market continues to perform well, indicating that the driving force mainly comes from loose liquidity. Overseas, the Federal Reserve may still cut interest rates this year and may continue to loosen next year; domestically, bond yields have dropped to a low of 1.6-1.7%, and funds will inevitably seek higher returns. The real estate sector lacks right-side signals, thus the equity market has become the main recipient of liquidity. In terms of industry selection, funds will not be evenly distributed but will seek out those industries with clear trends and fundamentals better than the market. For example, long-term growth sectors such as AI and innovative pharmaceuticals have already been reflected in industry prosperity and leading company financial reports. Therefore, from an industry perspective, liquidity is flowing from the bond market into the stock market, concentrating on these more certain sectors within the stock market Q3: Where is the current bull market positioned? Is it in the early stage, mid-stage, or late stage? How much room is there for the future? Zhou Junzhi: We have studied several bull markets since 2005 and found a pattern: non-ferrous PPI bottoming often corresponds to the stock market bottoming and rebounding, while the peak of black series prices indicates the end of the bull market. The logic behind this is that copper prices have both financial and demand attributes, making them sensitive to liquidity easing and economic recovery; whereas black series prices reflect real demand and only peak when the economy is fully heated. Currently, in April, China's pessimistic expectations were extremely severe, with the market expecting a year-on-year export decline of -5%, but the actual results were much better than expected, showing that economic resilience is significantly stronger than market imagination. At the same time, global liquidity is shifting towards easing, and while the drag from the Chinese real estate sector still exists, the marginal pressure has eased. Therefore, what the market is pricing in is that "the economy will not be worse than the worst moment," rather than "the economy has already rebounded significantly." In other words, the current bull market is still in its early stage. It has not yet entered the "profit-driven" phase and is mainly driven by liquidity and valuation recovery. It will take time to switch to a "profit bull." Therefore, this round of the bull market is definitely not at the end, but rather in the early stage. Q4: Why has the AI/computing power sector been able to strengthen first in this round of market? What are the essential differences compared to 2013-2015? Which specific tracks should be focused on next? Chen Weibin: I believe AI is the "vanguard" of the entire market, but this does not mean that missing its first wave means there are no opportunities left. The reason is that we judge AI is likely to initiate an industrial revolution lasting 5-10 years, similar to the Fourth Industrial Revolution. The U.S. government even characterizes it as an "industrial revolution, information revolution, and renaissance," which is highly qualitative. From a ten-year perspective, we have only gone through two years, still far from the first half. Unlike the thematic speculation of 2013-2015, this round of AI is not just a "story," but a real transformation that improves productivity, with policy support and industrial implementation far exceeding that of previous years. There is also differentiation within the sector: the upstream computing power chain, such as optical modules, PCBs, power supplies, and liquid cooling, each has different life cycles; some have been in progress for two years, while others have just started, such as liquid cooling, which has just been recognized for its value by the market. Downstream applications are flourishing, covering AI office, AI marketing, AI programming, and AI hardware such as glasses, robots, and intelligent driving, with huge future potential. Another direction with long-term trends is innovative drugs, which have a shorter cycle than AI but also have a 2-3 year high prosperity phase. Overall, whether it is AI or innovative drugs, they are still in the early first half. Q5: Is there an opportunity for domestic computing power to catch up? If projects like "Stargate" are implemented, which sectors or tracks might benefit first? Chen Weibin: In this round, the AI industrial chain was initiated earliest by overseas computing power, and the performance has been fully realized, with valuations between a dozen to twenty times, which is reasonable or even low. In contrast, domestic computing power started later, with lower performance realization, relying more on expectations for support, thus having higher valuation levels For investors, this means that overseas chains are "performance-driven + reasonably valued," while domestic chains are "expectation-driven + have huge potential." Choosing which side to be on depends on risk preference. In the long run, achieving domestic computing power in China is a clear direction, and there is strong policy support. In the future, domestic computing power chains still have the opportunity to develop independent markets. Q6: NVIDIA's new generation Rubin will adopt microchannel liquid cooling technology, which is costly and difficult. Which companies are likely to receive orders first? What is the approximate value of a single GW? How should ordinary investors view the opportunities brought by "heat dissipation + hardware upgrades"? Chen Weibin: Let me briefly explain my views on liquid cooling and place it within the broader context of upstream computing power, and then introduce microchannel liquid cooling plates. Many investors consider liquid cooling to be an ancillary industry, not as well-known as GPUs, chips, optical modules, or PCBs, and its importance seems low. This stereotype leads to an underestimation of the significance of liquid cooling. In fact, in terms of importance, we are at the beginning of the liquid cooling era in human history, starting in 2025, which will be the first year of liquid cooling. Over time, the market's understanding of liquid cooling will deepen, just as it once viewed optical modules as merely simple assembly components, only to later realize their importance. Currently, the market sees liquid cooling as an ancillary industry, but in the next year or two, liquid cooling could become a critical bottleneck in data centers. From the perspective of the capital market, liquid cooling is also an important direction worth paying attention to, with a long-term industrial trend. Comparing the growth rates of liquid cooling, PCBs, and optical modules: Liquid Cooling: The industry growth rate is projected to reach 500% by 2025. Optical Modules: The growth rate is estimated to be around 100% to 150%. AI PCB: The growth rate is estimated to be around 150% to 200%. Clearly, the industry growth rate of liquid cooling is higher than that of optical modules and AI PCBs. Moreover, the industry space for liquid cooling is also substantial. Estimates show that by 2025, the liquid cooling industry is expected to reach an industry space of 100 billion RMB, surpassing AI PCBs. By 2026, the liquid cooling industry space may reach or approach the level of optical modules. For these two reasons, liquid cooling is a very important industry with a long-term industrial trend in the capital market. Liquid cooling technology upgrades and the positioning of microchannels After recognizing the attributes of the liquid cooling industry, let's look at technological changes. The liquid cooling technology currently used in data centers is called cold plate liquid cooling. As server power increases, the requirements, value, and cost proportion of liquid cooling will also rise, so cold plate liquid cooling needs technological upgrades to dissipate heat more effectively. The future upgrade directions for cold plate liquid cooling include: Phase Change Cold Plate Liquid Cooling: Upgrading from the current unidirectional cold plate to bidirectional phase change (the cooling liquid transitions between liquid and gas states, similar to household air conditioning refrigerants) Immersion Liquid Cooling: The development from cold plate liquid cooling to immersion liquid cooling, which is further divided into unidirectional (always maintaining liquid state) and bidirectional/phase change (changing from liquid to gas, absorbing more heat). The general direction is to upgrade from the current unidirectional cold plate to bidirectional or phase change cold plates, and then to future immersion liquid cooling. Microchannel Liquid Cooling Plate, which incorporates microchannels into the liquid cooling plate, represents a new technological transformation in the aforementioned upgrade path, although it also comes with higher costs. It is one of the upgrade directions for liquid cooling technology and a potential technological route. This positioning is similar to potential technological upgrade directions in solid-state batteries (oxide or sulfide routes) in new energy batteries and perovskite in photovoltaic cells. It has two aspects of characteristics: Technical Uncertainty: It has the potential to become one of the upgrade directions in the future. Industry Validation: The industry needs to continuously observe and compare which new technology is easier to implement, more cost-controllable, and easier to promote. Technical Route and Value of Microchannel Liquid Cooling Plates Microchannel technology itself also has subdivided technical routes: Basic Microchannel Liquid Cooling Plate: Processing dozens to hundreds of micrometers wide flow channels inside the traditional liquid cooling plate, allowing the coolant to flow through the microchannels to carry away heat. MLCP (Microchannel Lid): Combining the lid and liquid cooling plate into one, with microchannels etched on the lid. Since the chip is packaged on the lid, this design can better eliminate the impact of the lid as a heat dissipation barrier. Chip-Level Microchannel Liquid Cooling: Directly etching microchannels on the silicon material at the back of the chip for heat dissipation. This belongs to a more advanced form, which may have higher costs but potentially better heat dissipation effects. All three routes ultimately need to undergo industry validation, considering costs and implementation difficulties comprehensively. Taking the medium MLCP approach as an example, we can estimate its value: The value proportion of ordinary liquid cooling plates in the entire liquid cooling system is about 30%. If the global AI server liquid cooling field has an industry space of 100 billion by 2025, then the market space for liquid cooling plates would be 30 billion. The value of MLCP microchannel lids is approximately 3 to 5 times that of traditional liquid cooling plates. Assuming that the penetration rate of MLCP can reach 50%, then its future industry space may be in the range of 45 billion to 75 billion. Orders and Investment Opportunities As for which companies can secure orders first, it is still in the early stages. Currently, we are still in the GB200/GB300 era, and MLCP-type microchannel liquid cooling plates are not yet needed. Microchannel liquid cooling plates will only be seen in the NVIDIA Rubin era to determine which companies can secure orders, which will take about a year to gradually realize industrial implementation. For ordinary investors, attention should be paid to this liquid cooling industry with long-term growth potential, especially those companies that have layouts and breakthroughs in technological upgrade directions (such as microchannels, etc.) Q7: What factors may trigger volatility in the current market environment that investors should pay close attention to? Zhou Junzhi: The core of the macro environment lies in assessing the larger context and potential risks. To discuss future risks, we first need to review the starting point of this round of market sentiment (bull market). The starting point of this sentiment was April of this year, when the market had priced in the stories of global demand shocks and the global rift potentially caused by the trade war quite thoroughly. A typical manifestation was the rise in gold prices and the weakening of the US dollar, reflecting the market's deep concerns about the global political and economic instability caused by tariffs and trade wars. Subsequently, as time went on, people realized that the severity of the trade war and global rift was not as serious as imagined, leading to a Taco trade (reversal trade) in the market. However, the current trade game has not yet concluded, so from a risk perspective, we still need to pay attention to the future developments in the trade game. The second core story is the change in global dollar liquidity, which has been discussed since the end of last year. From 2022 to 2024, US assets, thanks to the new round of technological revolution brought by AI, have almost stood out alone, with the main trading direction of global liquidity being to go long on US assets (especially US stocks). Entering this year, we have seen several changes: First, events like DeepSeek have made the market realize that China can compete with the US in the technology sector. Second, Trump's trade war has brought about vulnerabilities in the underlying global political and financial systems. As a result, global dollar liquidity has begun to flow out of the US, seeking assets in non-US countries. Initially, this involved pricing European stocks, European bonds, and the euro, and subsequently, China's renminbi assets also benefited from this round of global dollar "search for an anchor," gaining liquidity support. This is also why A-shares and Hong Kong stocks have enjoyed an increase in global risk appetite. If we extrapolate this trend, the risk points that need to be highlighted in the future remain the fluctuations and volatility of global dollar liquidity, which will impact Chinese assets and even global risk appetite assets. The third point of concern is domestic economic data. In April, the market had priced in a relatively pessimistic expectation for global and Chinese economic data. After August, although some economic data showed a decline, the extent of the decline was significantly lower than the pessimistic expectations in April, so the market has not fully priced in this round of economic data's downturn. The future trend of economic data is worth paying attention to, especially the fiscal factors that people are concerned about. In the first half of this year, China's fiscal expenditure was significantly advanced, and the resilience of the economy, apart from better-than-expected exports, largely came from fiscal stimulus. Entering the second half of the year, especially after August, we have seen the pace of government bond issuance begin to slow down, as a considerable amount of fiscal and government bond issuance has already been overdrawn in the first half of the year Therefore, the future speed of fiscal decline and its impact on economic growth, although it may only be a rhythmic impact, remains a risk point of concern for the market. In summary, trade disputes, fluctuations in global dollar liquidity, domestic economic data, and fiscal rhythm are currently three potential risk factors worth paying attention to. However, the probability of returning to the pessimistic expectations seen at the end of last year or in April this year is relatively low. Q8: With the rapid rise of tech stocks, will it trigger a round of adjustments? If adjustments occur, does it also mean new buying opportunities? Chen Weibin: From an industrial or sectoral perspective, the core factor we need to pay attention to is North American capital expenditure. If a global economic recession or any other reason leads to a decline in North American capital expenditure, this will be the most tangible risk for the current AI mainline and will affect the entire industry. Therefore, we have been continuously monitoring the capital expenditure situation under the North American AI business model. Recently, despite some fluctuations, such as the optical modules and PCB sectors experiencing volatility since September, these fluctuations are more due to normal capital behavior and trading factors: The stock prices have risen significantly over the past few months. The rise in stock prices is primarily driven by fundamentals, resulting from upward revisions in performance, rather than a significant expansion in valuation. The current valuation level remains within a relatively normal and reasonable range. Therefore, if capital expenditure remains normal, or even exceeds expectations, investors need not worry about major systemic risks. The periodic adjustments caused by trading factors and capital behavior are indeed opportunities for investors to choose their entry timing based on their own risk preferences. For directions with long-term industrial trends, such adjustments often signify new buying opportunities. From a business model perspective, we see that North American computing power capital expenditure has formed a closed-loop business model: Positive Cycle: The investment in capital expenditure is no longer based on a FOMO mentality, but rather because investing in capital expenditure can bring about positive revenue growth and a positive cycle in the business model. Driving Force: The growth in revenue prompts North American CSPs (Cloud Service Providers) to invest more capital expenditure, thereby forming a positive cycle. We have observed that the AI business model has entered this positive cycle since the second quarter of this year, thus we have high confidence in future North American capital expenditure. Q9: Looking ahead to the fourth quarter, will the main line of Chinese assets change? What policies should we focus on from the Fourth Plenary Session? Zhou Junzhi: Let me first talk about the key points worth noting from the Fourth Plenary Session and the "14th Five-Year Plan." The "14th Five-Year Plan" outlines the economic and strategic direction for China over the next five years. Looking back at the "13th Five-Year Plan," it framed two very important main lines: First, security; II. High-Quality Development. During the "14th Five-Year Plan" period, we have witnessed the rapid and non-linear development of China's mid-to-high-end manufacturing industry, which has also led to technological breakthroughs in areas such as AI and innovative pharmaceuticals during the transformation and upgrading of the manufacturing sector. At the same time, China's focus and layout regarding security have also increased compared to the past. Looking ahead to the direction of the "15th Five-Year Plan," we can find answers from the scattered clues of last year's Central Third Plenary Session, with the main directions still being two: 1. Technology. Over the past twenty years, China has entered the mid-to-late stage of a model driven by mid-to-low-end manufacturing, real estate, and infrastructure during the "14th Five-Year Plan," and its empowering effect on China's economic growth has significantly weakened. In the future, China will inevitably place more emphasis on the technology direction. 2. Domestic Demand and Reform: The policy system or domestic demand structure derived from real estate and infrastructure over the past twenty years will undergo adjustments. In terms of domestic demand, we will see fiscal and tax reforms, as well as reforms in areas such as people's livelihoods, social security, and education, to stimulate residents' consumption potential. These three (technology, fiscal and tax reform, stimulating domestic demand) are relatively important main lines in the "15th Five-Year Plan." Ultimately, we still need to pay attention to the details of the documents, but in terms of direction, we can continue to explore corresponding trades and opportunities. Regarding whether the market style or main line will change in the fourth quarter, our judgment is that it will not change significantly. The judgment of this round of bull market is a bull market that has emerged from the economic trough, usually accompanied by the rise of emerging industry revolutions and liquidity easing (i.e., the restoration of risk appetite). Referencing the previous period of emerging from deflation (2013 to 2015): the market journey was first to go long on liquidity and restore risk appetite, and then gradually switch to go long on the broad fundamentals. In the "go long on liquidity and restore risk appetite" phase, if certain new economy sectors show significantly higher growth expectations than traditional economic sectors, their pricing will be relatively sufficient. Looking ahead to the fourth quarter, since the fundamental data of China's total economic volume has not yet shown a very strong rebound, the market is still in the first half of the bull market emerging from the economic trough, mainly focusing on going long on liquidity and restoring risk appetite. Therefore, the main line direction will not undergo fundamental changes. However, after the announcement of the "15th Five-Year Plan," there may be funds that stage price certain thematic investment opportunities. The market may need to wait until next year to see more rebounds in fundamental and total volume data before it fully switches to going long on corporate profits and going long on fundamental rebounds. Thus, there will be no significant changes in the market's main line in the fourth quarter. Q10: What role will technology stocks play in such a macro environment? Chen Weibin: We believe that technology is expected to become the main line throughout this round of equity market. This round of technology market is not blind speculation, but rather has solid fundamental support. Unlike before, compared to the "Internet+" round from 2013 to 2015, the rise at that time was more expectation-driven, with very few actually translating into performance This round of AI (artificial intelligence) has solid fundamental support and is expected to become an industrial revolution lasting 5 to 10 years. Currently, the trend is led by the United States overseas, while China is making significant efforts to catch up. Therefore, technology represented by AI is likely to become a main thread running through the entire market. Under this main thread of AI, although we have seen outstanding performance in upstream computing power (such as optical modules, PCBs) in the second and third quarters, there are still many sub-sectors in the AI field that have not fully risen, and its industrial cycle is still in a very early stage. In the future, the following situations are likely to occur: the market will continue to explore new investment opportunities along the direction of long-term industrial trends: In upstream computing power, look for investment opportunities in the liquid cooling field. In downstream AI applications, look for opportunities in AI software and AI hardware. AI hardware includes but is not limited to AI glasses, AI smart driving, AI humanoid robots, and other fields. Overall, we hold an optimistic attitude towards this technology direction with long-term industrial trends. Q11: How do you view the current high positions of the liquid cooling and optical module sectors? Will the progress of OCS and the rise of ASIC chips change the long-term pattern of the Nvidia industrial chain? Chen Weibin: Let me first state the conclusion. After several months of significant increases in optical modules, the performance is more performance-driven, and the current valuation has entered a reasonable range. The sector may need some time to fully digest the profit-taking brought by the past huge increase. After a period of oscillation and digestion, the valuation still has upside potential. Currently, the valuation of optical modules for next year is about 17 to 18 times, while the growth rate of the entire AI industry may be between 30% and 50%, and the growth rate of optical modules/PCBs may be more than doubled. In terms of growth rate comparison, a valuation of 17 to 18 times is still at a reasonable or even low level. In the future, the optical module sector may evolve into a more optimistic valuation system, such as 20 to 25 times, or even 25 to 30 times. However, its most intense main rising wave phase may have already passed. In the future, after the valuation undergoes oscillation and digestion, there is still room for growth, but it requires patience. As for the liquid cooling sector in upstream computing power, I believe it is still in an early stage, somewhat like the optical module or PCB sectors in the first half of 2023. The lifecycle of liquid cooling has just begun to start in the second half of this year, and performance has not yet truly been released. The true release of performance will gradually become apparent next year. Currently, the combined market capitalization of the core companies in optical modules and PCBs may already be at the trillion RMB level; while the market capitalization of listed companies related to liquid cooling, the leading companies are in the hundreds of billions, and all related companies combined may be at the level of 100 billion to 200 billion It can be seen that the liquid cooling sector is still in the early stages of its lifecycle and has not yet experienced significant price increases, making it a direction that investors should pay more attention to. We believe that in the future, among upstream computing power, liquid cooling may take over from optical modules and PCBs, becoming the next key factor in AI investment. As for the progress of OCS (Optical Switches) and the rise of ASIC chips, our understanding is: This industry will always see new technologies emerging to compete, as well as new competitors entering the market. Whether it is NVIDIA (NV) or ASIC chips, they both have their business model rationale in the field of computing power, and it is not necessarily the case that one will overpower the other in the future. The most likely scenario is that, against the backdrop of growing AI demand, both NVIDIA and ASIC will be in a flourishing process, each with its own business positioning, and will grow together with the industry's expansion. Risk Warning: The master class is a platform for selecting personnel from third-party compliance institutions to teach investment research theory courses. The content taught does not constitute buying or selling advice for any specific products or investment recommendations. The opinions expressed in the platform courses are for learning and reference only and do not represent the opinions or views of Wall Street Insights, nor do they address the user's specific investment goals, financial situation, or needs. The market is volatile and uncertain, and the platform is not responsible for any losses incurred by you relying on the opinions or information from the courses. Investment carries risks; please make decisions cautiously