1. Market Review
In August, after the pressure from crowded trades was relieved, the U.S. stock market saw a wave of gains, but there was a growing structural divergence.
AI-driven hopefuls like ServiceNow, Meta, and Apple reached or approached new highs, while Microsoft’s performance was relatively moderate. In the sportswear sector, ONON reached a new high, while LULU’s rebound was weak, following a drop of more than half this year. In e-commerce, Latin American leader Mercado Libre hit a new high, whereas PDD gave a poor outlook, leading to a sharp decline. Although NVIDIA (NVDA) initially showed strong growth, information revealed in its earnings report and subtle changes in supply and demand within the industry increased market concerns over sustained demand, resulting in heightened volatility.
Moreover, despite growing market concerns about economic landing issues, Walmart, Costco, and Mastercard, which reflect the overall situation of consumer/essential consumption, hit or neared new highs. We believe that in the absence of systemic risks, this divergence is likely to continue. For us, the priority remains identifying opportunities globally with competitive moats and unique alpha.
AI – Our View Remains Unchanged:
Nvidia currently has very high market expectations. Despite industry investments reaching hundreds of billions of dollars in card purchases, Nvidia’s revenue is only in the tens of billions, yet it is priced with the expectation that the AI market will reach several trillion dollars or even higher. Even if this expectation is realized in the future, for Nvidia to see significant further growth at this level, the forecast for AI’s potential would need to be substantially raised. While possible, this seems overly optimistic. Moreover, if we believe AI applications will eventually become a multi-trillion-dollar industry, there’s no reason why opportunities won’t spread to applications, suggesting that current AI applications are undervalued. So far, signs increasingly indicate that Nvidia is more likely overvalued than AI applications are undervalued.
AI applications have a promising future because human value is increasingly recognized, and AI enhances human efficiency, replaces certain tasks, and can even accomplish things humans cannot. The potential value of AI is immense due to the large population it can serve. However, many high-quality AI application companies have not yet been fully priced for this potential. Currently, AI is still advancing in areas like enterprise services, personal assistants, robotics, and autonomous driving, but it will take time, as none have yet reached the point of crossing the chasm.
Edge AI is most likely to become a personal AI super assistant. However, the ability to deliver such a super assistant at the edge is extremely demanding, and globally, very few companies have the comprehensive ability to achieve this. In B2B enterprise services, there’s Microsoft, in the internet sector, there’s Meta, and for edge AI, there’s Apple.
Last month, we discussed how long AI large model training can continue to burn money. The core conclusion was that with each generational upgrade of large model training, the demand for computing power multiplies. Training a single GPT-4.0 requires several billion dollars in hardware investment, while GPT-5.0 would need tens of billions, and GPT-6.0 would reach hundreds of billions. To maintain Moore's Law-like iteration speed in large models, the next generation would require an investment of hundreds of billions of dollars. Without seeing significant commercialization, even the tech giants would struggle to keep up with this scale of investment. Whether these giants can follow suit, and how many can keep up, will impact Nvidia’s growth expectations in the years to come.
In reality, although it’s too early to disprove Nvidia’s investment logic based on its earnings reports and upstream-downstream industry tracking, some flaws have already begun to emerge:
Revenue Distribution: In Nvidia’s latest earnings report, U.S. revenue showed no quarter-on-quarter growth, with its share falling from over 50% to around 40%, while revenue from China and Singapore grew significantly, now making up a combined 30%. Normally, North America’s hardware demand would be more sustainable and have growth potential, whereas demand in China may contain more one-off purchases, raising concerns about its sustainability. Furthermore, Singapore contributed nearly 20% of Nvidia’s revenue, but it’s hard to imagine that domestic demand in Singapore and surrounding Southeast Asian countries could support such a large amount of purchases.
H100 Chip Supply: According to feedback from server manufacturers, the H100 chip isn’t as scarce as before, and it's relatively easy to purchase. Meanwhile, downstream customer demand has started to weaken. It's still unclear whether this is due to overall demand declining or if customers are waiting for the next-generation B-series GPUs, leading to a lukewarm attitude toward the soon-to-be-obsolete H100.
Product Complexity and Supply Chain: Looking ahead, Nvidia’s product iterations are becoming more complex, and challenges related to design and the supply chain are growing. The risk of on-time delivery is increasing, and the redesign and delayed delivery of the Blackwell series may just be the beginning.
In summary, Nvidia’s sales forecasts were previously based on production capacity due to the persistent shortage of GPUs. However, with demand starting to soften and challenges emerging in design and delivery, earlier expectations may have been overly optimistic. With consensus expectations remaining high, both short-term and medium-term growth assumptions are now being challenged, and current expectation adjustments may not be sufficient.
Consumer Trends – Interesting Observations
One interesting trend is in the sportswear sector. At the end of June, Nike's earnings report gave weak guidance, and Lululemon (LULU) showed a sharp slowdown in growth in early June, causing concerns about its growth potential. At the time, ONON followed Nike and LULU in declining, likely reflecting two concerns:
Skepticism about the growth potential of new sports brands. When LULU reached around $10 billion in revenue, its growth hit a bottleneck. Aside from a few older brands like Nike and Adidas, which have far surpassed this threshold, other brands have struggled. Under Armour, which once seemed promising, stalled at around $5 billion, and Asics, despite being an established brand, has plateaued at roughly $4 billion. This raises concerns that the fast early growth of new brands may be driven by short-term trends, with $3-4 billion in revenue being a difficult hurdle to overcome for most brands. At the time, ONON was priced at around 4.5x PS for 2024, while LULU, which had hit a growth ceiling, was at 3.5x PS (before its decline, LULU’s PS was 6x). This implied that the market believed ONON would hit its ceiling after growing by another 30%, reaching around $3 billion in revenue.
Sportswear is a discretionary consumption category. With the industry's growth already slowing down, all brands, including ONON, were expected to be affected.
However, in Q2, ONON still achieved nearly 30% growth, and in the U.S. market, where concerns were highest, its growth only slowed slightly, maintaining over 25% growth. Next year, the company's overall revenue might reach around $3 billion, and in another 2-3 years, its retail scale in the U.S. could be equivalent to 50% of LULU’s.
The performance of sportswear brands has shown an interesting divergence. It seems the market may have overlooked that ONON’s niche market isn’t as competitive as it appears, while demand in this niche is growing rapidly. ONON is still in the early stages of customer penetration, network expansion, product innovation, and category expansion, driven more by its own alpha than by the overall industry trend.
Another interesting trend is in the energy drink sector. Initially, we assumed that energy drinks and coffee both serve as caffeine-based stimulants, and in the U.S., energy drinks are much cheaper than freshly brewed coffee, especially Starbucks. Americans drink about 10 cups of coffee a month, but less than one bottle (400ml) of energy drink, indicating huge growth potential. Sugar-free and flavor-diverse energy drinks align with the broader trend toward health-conscious and flavorful beverages, suggesting even faster growth. However, it appears that energy drinks are more of a discretionary purchase. Even though the price per bottle is only $2, there are many cheaper alternatives. Meanwhile, despite being much more expensive, coffee drinkers have not been widely converted to energy drinks. As a result, the energy drink industry’s growth has turned negative this year. Celsius, a third brand that was expected to grow by covering more channels, innovating products, and meeting diverse needs, currently holds only a 10% market share, and its growth and market share expansion strategy is now being challenged.
Latin America – Recent Increase in Investment
Taking e-commerce as an example, we’ve found that Latin America is a fascinating market. On one hand, it’s a large market with a vast population and a per capita GDP of around $10,000, offering significant potential. On the other hand, the population is youthful, which favors the rapid penetration of emerging trends like e-commerce. More importantly, the unique characteristics of the Latin American market have created a high barrier for the local "kingpin" Mercado Libre, making it difficult for international giants like Amazon and Chinese cross-border e-commerce companies to compete effectively.
Although e-commerce is a highly competitive industry, even today, the outcome in China’s market is still far from settled. However, in Latin America, which is still in the early to mid-stages of development, the competitive landscape is much clearer. The uniqueness of the Latin American market lies in its long logistical distances and high fulfillment complexity. Despite China’s abundance of affordable and high-quality products, shipping to Latin America takes a long time, and customs clearance is unpredictable, posing significant risks of inventory devaluation for sellers. Even when goods are successfully sold to the Latin American market via cross-border channels, issues such as added taxes and high logistics costs significantly reduce the cost advantage of Chinese goods.
Moreover, if companies attempt local production, it is challenging to replicate the high efficiency and low cost found in China. In terms of last-mile delivery, GMV (Gross Merchandise Value) scale has a significant impact on fulfillment speed and unit logistics costs. This allows local players like Mercado Libre, even with an asset-light approach to building their network, to deliver most goods within 48 hours. In contrast, foreign entrants, whether they build their own logistics or rely on partnerships, face higher unit costs and longer delivery times.
Given these high barriers and the market’s strong growth potential, Mercado Libre is positioned to achieve higher profitability at an earlier stage of the industry’s development.
China Assets – Structural Opportunities Are Dwindling
In the sectors we cover, an increasing number of areas are deteriorating, though a few niches still offer bright spots. Examples include emerging demand sectors such as Pop Mart and Lao Feng Xiang (representing relatively low-penetration demand), as well as standout companies within industries that are otherwise in decline, but still managing to capture market share and improve quality and efficiency, like Moutai in the liquor industry, and Tencent and Meituan in the internet sector.
While both the U.S. and Chinese economies seem to be slowing down, and there are companies with alpha that are outperforming their sectors, the visibility and certainty regarding future fundamentals and pricing differ greatly between the two markets. Overall, it’s becoming increasingly challenging to identify opportunities in Chinese assets.
Here’s a glimpse of mid-year reports from several industries:
Property ManagementValuations for both private and state-owned property management firms have dropped significantly. Greentown Service, a branded and independent property management company, posted relatively solid results in its mid-year report. Revenue growth slowed to 10%, but profit margins remained stable, cash flow was good, buybacks were executed, and regular annual dividends (possibly with special dividends) are expected. However, revenue growth is slow, and visibility for some business segments is low. With single-digit net profit margins, any slight disruption could heavily impact profits. For state-owned property management companies, which were previously expected to grow rapidly, growth has now slowed significantly. Profit margins and cash flow quality have not yet deteriorated, but post-IPO revenue growth and net margins have improved, likely due to support from real estate affiliates—though the extent is unclear, and the future is uncertain. For property management companies associated with distressed real estate firms, such as Country Garden Services, revenue has stagnated. While profit margins improved somewhat in the first half of this year compared to last year’s poor performance in H2, the sustainability of this improvement is uncertain. Apart from basic property services, most business areas remain challenging, and cash flow is still weak.
EducationIn August, the "Opinions on Promoting the High-Quality Development of Service Consumption" document mentioned encouraging education and training services, indicating a more relaxed regulatory environment. However, private schools and education companies are unlikely to significantly increase investments based on these guidelines, as the overall policy tone hasn’t changed, and the rule-making mechanisms remain the same. Although there is currently some regulatory leniency, it’s uncertain when restrictions might tighten again. New Oriental's two consecutive quarters of earnings misses reflect, to some extent, the caution education companies are exercising to avoid appearing overly profitable, which could draw negative social attention. The perception seems to be that companies in livelihood sectors shouldn’t appear excessively profitable.
Liquor & BeveragesThe liquor industry is deteriorating. While Moutai's prices remain stable, the supply-demand dynamics are not as favorable as before. The premium segment has begun to feel the impact—Fenjiu's growth has slowed significantly, with Q2 profits growing only 10%, below expectations of 15-20%. Shede reported a revenue decline and a sharp drop in profits.In the coffee and tea beverage sector, store expansion has slowed, with most new openings from Luckin and a few from Starbucks. However, considering both milk tea and coffee together, store openings still show over 20% year-on-year growth. Luckin is gaining market share in the coffee segment, but when considering the entire milk tea + coffee market, its share is stable. In Q1, the combination of off-season, competition, and demand factors led to poor pricing and profit margins. Q2 showed a recovery, but not to last year’s levels. As new milk tea products launch in the peak season, some improvement is possible, but demand remains uncertain. Drawing from insights in the North American beverage industry, during times of weak demand, consumers don’t substitute drinks—they just drink less.
HotelsThe hotel industry is declining. Room supply has increased by 5-10% compared to the end of last year, while demand has barely grown, leading to negative RevPAR (revenue per available room) growth.Supply is unlikely to clear quickly, so hotels are cutting prices to maintain occupancy. If that doesn’t work, they rebrand to slightly better-performing hotels, and if even that fails, they exit the market. The demand outlook is poor, with business travel, which accounts for over half of demand, facing structural challenges from remote alternatives and broader macroeconomic impacts. While leisure travel demand helps offset some of this, the willingness and ability to pay are different between these two segments, and the supply-demand imbalance shows no signs of changing. Leading hotel chains, though affected by declining RevPAR, are still gaining market share and upgrading products. Their asset-light franchise models help maintain good profit margins, but future prospects are uncertain.
InternetInternet companies are generally experiencing slowing growth, though efforts to improve efficiency have boosted profit margins or reduced losses, and many are increasing dividends and buybacks. However, once efficiency gains are fully realized, there are few new highlights to look forward to, and for most large companies, maintaining stability will be an achievement, as they are not immune to macroeconomic impacts.In terms of dividends and buybacks, the future may become more polarized. Companies like Tencent, which have strong cash flow from overseas, are likely to maintain aggressive payouts. JD.com, which is actively pursuing high dividends and buybacks, faces challenges in generating sufficient USD, as it is currently borrowing to fund its buybacks—raising doubts about sustainability. Pinduoduo, on the other hand, has explicitly stated it will not engage in dividends or buybacks. While many companies are showing an increased willingness to distribute profits, the ability to do so is constrained by capital outflow issues, which affect the sustainability of high dividend payouts and buybacks.
2. Market Outlook & Expectations
Our view on investment opportunities in both the U.S. and Chinese markets remains largely unchanged: we continue to look for structural opportunities. However, opportunities in China are relatively harder to find, while the U.S. presents relatively more. Markets like Latin America and Europe also offer opportunities, so we aim to source as much alpha as possible from different regions.
Regarding the overall U.S. stock market, both favorable and unfavorable factors are expected to coexist in the near term. Favorable factors include:
The crowded trades risk has been largely released after the July sell-off;
A U.S. interest rate cut is becoming more imminent, with the Federal Reserve having ample policy space;
Although the Bank of Japan's rate hikes are a long-term direction, there will still be flexibility in their approach.
Unfavorable factors include:
The core driving force for U.S. tech stocks lies in AI, but AI applications are still awaiting validation, and the sustainability and growth potential of demand for large AI model training, led by Nvidia, is now being questioned;
Economic growth is slowing down, and it is unclear where the bottom lies without signs of improvement or strong external forces;
The upcoming U.S. election adds an additional layer of uncertainty.
Overall, we do not currently see systemic risk. As long as the economic slowdown remains within reasonable limits, it’s still possible to find structural opportunities.
Where are the opportunities?
AI is a key source of long-term alpha. As human value continues to rise, AI, which enhances human efficiency and even replaces human labor, will increase in value accordingly. This is only the beginning, and the potential ceiling is very high.
Lifestyle changes focused on health and greater self-care will continue to create opportunities in the consumer sector.
Increased purchasing power will allow for higher pricing in certain products/services and will enhance the relative competitiveness of high-value products/services.
In a rate-cutting cycle, certain sectors that were previously hard to evaluate due to rising interest rates, but have good competitive dynamics and long-term growth potential, may present opportunities once the short-term interest rate risk is resolved.
Fintech and other emerging investment opportunities.
Opportunities in other markets, such as Latin America and Japan.
精彩评论