Mu Yiling's Latest View: Logical Deduction and Future Trend of the Next Market Rally
The recent market volatility has intensified, with mutual influences between domestic and overseas markets drawing widespread attention. A few days ago, at an institutional roadshow, renowned strategist Mu Yiling shared his perspectives.
How should we define this round of correction?
Has the AI-led tech market rally come to an end?
After this round of correction, what is the most likely form of the next rebound or upward trend?
How should we allocate and position our portfolios to cope with the current situation?
Mu Yiling broke down the logic step by step and put forward his own judgments (this article is the full transcript of the roadshow, written in first-person perspective).
Key Quotes:
1. In early June this year, the capital expenditure of global cloud vendors soared to nearly 1% of global GDP, which is an extremely striking macroeconomic data. It is equivalent to a huge national-level demand emerging out of thin air across the world.
2. Huge demand drives massive investment, which in turn boosts massive credit support and enormous debt, inevitably triggering inflation in the global market and highlighting the cyclical nature of the technology industry.
3. From 2006 to 2007, coal, steel and shipping were the absolute "darlings of the market". Back then, the market was concerned about global shipping space and the area of the Pacific Ocean, just as people are calculating the computing power demand of the United States now.
4. Nowadays, it is those start-ups that support Token consumption in overseas markets. Their stock prices have mostly fallen back to the levels at the beginning of the year since this year, indicating severe divergence in the financing market.
5. The global rental price of computing power assets is a very important observation indicator. If it continues to remain stable and grow, the computing power market and industry demand will be in good shape; otherwise, we need to stay alert.
6. At the current stage of equity investment, we should take stable assets (such as dividend assets) as the core position, and keep part of cash, so that we can appropriately participate in some rebounds when the market fluctuates extremely.
Two Logic Paths for Market Trend after Stabilization
After the market goes through a period of correction, there are usually two scenarios for the rebound market:
The first scenario is accompanied by improved fundamentals. If the rebound is based on this logic, we need to see the market volatility keep decreasing.
What does declining volatility mean? It means that the previous panic has dissipated, capital no longer blindly sells off or chases the rise, and everyone starts to calm down and act in line with the actual fundamentals. This is a mild and relatively solid foundation for a rebound.
The second scenario shows no sign of improvement in fundamentals. Under such circumstances, the volatility has to expand for the market to hit the bottom.
Why? Because expanding volatility means the previous panic is going through its final, extreme release. Currently, the technology sector is the first to correct in the market. When market sentiment is extremely low, panic selling floods the market, and at the same time, bottom-support capital enters the market, this drastic collision will likely bring obvious rebound opportunities.
Back to the coping strategies and allocation ratios, I believe at the current stage of equity investment, we should take stable assets (such as dividend assets) as the core position. When the market is panicking and volatility is extremely high, appropriately participate in some rebounds to increase returns.
In addition, in the current highly volatile market, you should always keep a portion of cash (for example, 10% to 20% in money market funds). This is not a bearish view on the market, but a preparation to cope with high market volatility. When the fundamentals are not completely deteriorated and the market sees irrational sell-off, this cash will deliver extremely high efficiency in bottom fishing.
The Cyclical Characteristics of the Technology Market Rally Should Not Be Underestimated
Having gone through this round of AI-centric tech market rally, everyone is discussing whether it has reached the "last leg of the bull run" stage. At the strategy meeting in early June this year, our team did a lot of reflection, and we noticed the cyclical characteristics of the technology market rally back then.
Why do we say it is cyclical?
Because we observed an extremely macro and striking data: The capital expenditure of global cloud vendors has suddenly soared from nearly 0% to almost 1% of global GDP.
In history, the only comparable case is probably the manufacturing investment pulse after China joined the WTO from 2006 to 2007. This is equivalent to a huge national-level demand emerging out of thin air in the world, and demand of this scale will inevitably trigger inflation in the global market.
Because massive investment requires huge credit support, which will also generate enormous debt, accompanied by extreme dependence on cash flow.
After the upstream price rises, it will also involve the issue of profit redistribution across all links of the industrial chain.
Recall those cyclical stocks in the past — coal, steel, shipping. You may think they are all old, traditional industries now, but from 2006 to 2007, they were the absolute "darlings of the market" with excellent growth and huge imagination space.
What aspects of core assets did the market care about back then?
Back then, the level of market attention to global shipping space and the area of the Pacific Ocean is essentially the same as people's current calculation of how many white-collar workers in the US will be replaced by AI and how large the computing power demand will be.
Back then, the growth of these assets was not a problem, but when did their cyclical nature emerge and lead to price declines?
We can observe that when China's coal stocks turned downward in October 2007, the fundamental supply and demand of the industry did not show any danger signals, and supply still fell short of demand. The only signal was that the demand growth rate in the middle and downstream was slowing down, and at the same time, the RMB credit level began to turn downward.
Combined with these historical experiences, we have proposed several core indicators to observe the current technology market rally:
The first is the sensitivity of the current inflation level to the technology sector;
The second is the ratio of future revenue generated by existing input to debt, to see if it is rising.
The Squeezing Effect of K-shaped Divergence
In the process of rapid AI development, we have also seen severe K-shaped divergence in the economy, which is not only reflected in the stock market, but also in macroeconomic data.
For example, China's export data has hit a record high, but the stock prices of many non-AI export enterprises have not performed well, and their Q3 performance has even come under pressure. Why is that?
First of all, if AI hardware (such as optical modules, etc.) is excluded, the overall exports in non-AI fields have been declining since June this year (except for a slight rebound in September).
Secondly, non-AI fields are also affected by another factor: the exchange rate. When global capital flows to the United States, Chinese enterprises will face a scenario where both the RMB and the US dollar appreciate when they settle foreign exchange. This leads to obvious exchange gains and losses for non-AI export enterprises.
Why is the RMB still appreciating when the US dollar is strong?
Because for those enterprises that earn high profits from AI hardware, the appreciation of the RMB has little impact on their earnings. Their profit buffer is very thick, and they can fully absorb exchange rate fluctuations. But in turn, this squeezes the profits of other traditional export enterprises.
This is the squeezing effect of AI. When the upper end of the K-shape (AI hardware) is discussing when the market will peak, the lower end (traditional manufacturing, export enterprises) may not have hit the bottom yet.
A Reference Case Worth Noting Now: The 2020 Bank Stock Rally
Several domestic and overseas tech giants that are the leaders in the AI field recently released their Q3 financial reports, confirming one fact:
Tech giants are still sparing no expense — even if hardware prices rise — to purchase computing power. This is the absolute advantage of the hardware industrial chain in profit distribution. The year-on-year growth rate of capital expenditure of several major cloud vendors is above 100%, they cannot turn back and can only move forward.
But what really needs vigilance is not these deep-pocketed cloud vendors, but the financing capacity of computing power lessees (such as start-ups that purchase or lease GPUs).
Currently, a lot of Token consumption is generated by these start-ups, as well as the financial institutions or venture capital funds that provide financing for them. Before 2024, the stock prices of these enterprises once significantly outperformed the S&P 500.
But since the beginning of this year, the stock prices of these companies have basically fallen back to the levels at the start of the year, indicating severe divergence in the financing market.
This is a bit like the A-share bank stocks before 2020. Back then, bank stocks kept seeing valuation compression. Although building materials and liquor sectors were still rising, the decline of bank stocks reflected the market's deep concerns about real estate and urban investment debt, which sounded the alarm for the real estate cycle in advance.
Now the stock prices of these US financial companies that provide financing for AI start-ups are also sounding the alarm for investors.
The core collateral that determines the future trend is the rental price of various computing power cards (GPUs). As long as the rental price remains firm, the balance sheets of start-ups are healthy, and there is no problem on the financing side. But if the rental price starts to turn downward, we need to be extremely careful. Just like the subprime mortgage crisis back then, the price decline of underlying assets (houses / computing power cards) will lead to a chain reaction of default and sell-off.
Therefore, the rental price of computing power is a very important observation indicator. At the same time, we can also refer to the Token consumption volume and price trend as auxiliary indicators. At present, the overall inference price of Token is going down, and the key is to see whether the growth in volume can offset the price decline.
Hidden Worries of the US Economy: Internal Division Under Investment Pulse
Led by AI, the US stock market has performed very well, but the fundamentals of the US macroeconomy seem to have hidden worries all along. The current state of the US stock market is very interesting: if you only look at the performance of middle-class consumer stocks, it does not look good; but if you look at companies like luxury cruise operators or Nucor Steel, their performance exceeds expectations.
From the third quarter of last year to today, after excluding the information technology and real estate sectors, other areas of the US economy are slowly moving upward. This is because the pulling effect of AI's investment pulse, which accounts for more than 2.5% of US GDP, is spreading to other industries.
For example, construction workers in Manhattan have gone to data centers to do cooling projects, and steel mills have received a large number of orders from data center construction.
From a static perspective, the total US economic volume is growing, and AI investment is driving global capital to flow to the United States. However, there is serious divergence in the internal structure, and conflicts between different classes are intensifying.
The popularization of AI is essentially the pursuit of extreme efficiency, but this may conflict with the benefit distribution of the general public. If the distribution relationship between people is not handled properly in this tech wave, and the middle and lower classes do not enjoy AI dividends but face the threat of unemployment instead, the mid-term political game may bring variables, and the economy may face a slowdown. In the medium term, the US economy also needs a process of K-shaped convergence.
The Future of AI Applications: Waiting for Cost "Reset"
We have spent a lot of time talking about the "darkest hour" and "last leg of the rally" of AI hardware, so when will the downstream AI application market rally arrive?
I actually watched an AI-generated time-travel short drama yesterday. The intervention of AI can indeed help screenwriters make up for many shortcomings in historical knowledge, avoiding factual errors and bugs. But returning to the industrial logic,
I believe the real explosion of AI applications requires an absolute prerequisite: the cost of upstream computing power must see a systematic decline.
For example, before, the analysts in our team all paid out of their own pockets for mainstream large models. Although they are more user-friendly than free domestic large models, their popularity is limited, and everyone uses them very cautiously. Later, the company uniformly equipped everyone with an enterprise office assistant, which is equivalent to providing it for free. Facing free tools, even if people have to tolerate a little AI hallucination, they are willing to have more conversations and tests, and the usage frequency of the application exploded instantly.
This brings us back to the 2006-2007 logic. Back then, it was certain that China joined the WTO and the manufacturing industry supplied the whole world, but that did not stop the cyclical decline of upstream coal stocks and steel stocks. It was precisely because of the subsequent overall surplus of basic production capacity in upstream energy, midstream chemical and steel sectors, and the sharp drop in costs, that the subsequent full prosperity of downstream manufacturing and application end was nurtured.
Disruptive innovation requires extremely low factor costs, so as to provide innovation space for those young people who test and make mistakes in their garages. Only when the cost of hardware and computing power drops significantly, can various applications truly blossom everywhere. Once the applications are proven viable, they will in turn form continuous demand pull for upstream computing power, thus completing the great transformation of the entire AI industry from "capital expenditure driven" to "revenue driven".
Cost Reduction Will Stimulate the Explosion of AI Applications
Currently, the computing power cost of Token is not cheap enough. Although the performance of each new generation of Nvidia's architecture is improving and the unit cost is decreasing, it is still far from the critical point that can stimulate unlimited innovation.
Now we are in the transition and ambiguous period of the great development of AI applications. The sprouts have emerged, but it has not yet reached the level that can drive upstream computing power in reverse.
Before discussing which downstream sub-sectors are expected to stand out, I would like to share a famous Wall Street quote: "Consumers are not willing to pay for a large number of applications." This is the underlying logic.
Few individuals will actively spend money on productivity tools to "do their jobs better" (except for some extremely outstanding people who will actively purchase services). Most people are more willing to pay for entertainment, such as watching short dramas and playing games, and they are more accustomed to the free model. In the mobile Internet era, the final payer is often not the direct consumer, but those who have made the advertising model and platform commission model work.
Therefore, we can divide AI applications into different tiers: the first type, the model that tries to charge C-end users directly for improved work efficiency, will most likely hardly work; the second type, the model that can seamlessly integrate AI into entertainment and social scenarios and get paid by B-end users or realize value through traffic monetization, is more likely to become the real leader in the future.
Investment Suggestions for the Current Market
Finally, in terms of the current practical investment operations, I have three suggestions:
First, as mentioned earlier, keep a certain cash position, and look for mispriced bottom-fishing opportunities during extreme market volatility.
Second, for the widely recognized dividend assets, I believe the market rally has not ended yet, but some sectors may have already overdrawn their annualized compound return space, so moderate allocation is needed.
Apart from the AI sector, there are very few industries with exceptionally prosperous fundamentals in the A-share market at present. Judging from the Q3 reports, the overall profitability is at a downward inflection point. But we are picking the best from the worst, and we see that sub-sectors such as textile and apparel, medical devices, refining and trading, and coal energy (which also belongs to dividend assets but has stronger fundamentals) have started to stabilize or see distress reversal. Almost all the investable sectors now are secondary or tertiary sub-sectors, and it is no longer the era when four or five primary industries moved forward in tandem. The granularity of investment needs to be finer, and the difficulty has indeed increased a lot.
Third, if you still hold tech stocks, after such a large correction, you should at least not cut your positions and leave the market now. Before the fundamentals see a major inflection point, the market will definitely see repeated fluctuations. After the rebound and recovery, you can re-examine whether these companies have competitive advantages and premium rights in the industrial chain. Referring to the cyclical stocks in 2007, many stocks with real core competitiveness doubled when they rebounded from the bottom in 2008 and 2009.
In the darkest hour of the market, emotional venting often overshadows rational logic. But as long as we see clearly the cycle of capital expenditure, the collateral price of computing power rental, and the K-shaped divergence of global capital flows, we can stabilize