Operation Review: Slightly increased investment in U.S. consumer stocks, maintaining a high position
The changes in investment in February were not significant, with a slight increase in investment in U.S. consumer stocks. The main reason is that our understanding of certain consumer stocks became clearer and their investment logic was further verified, leading to a positive change in IRR (Internal Rate of Return) and certainty. The proportion of investment in AI is between 35-40%, mainly focusing on B2B applications, chip manufacturing, and design tools. Other holdings focus on cloud computing, the internet, and consumer goods. Currently, the proportion of overseas assets is still around 80%.
Market Review: Further cooling of U.S. interest rate cut expectations, the economy shows resilience, with many structural highlights
The expectations for interest rate cuts have cooled further, with the mainstream expectation for the timing of the first rate cut moving from May to June, and the mainstream expectation for the number of cuts also reduced to three, totaling 75 basis points. At the same time, data continues to reflect stronger-than-expected U.S. employment, indicating that consumer spending this year will still be relatively resilient. In China, the risk premium is currently at a relatively high level, while the risk-free rate still has room to decrease. Therefore, the probability of a systematic decline in valuations is significantly reduced. The subsequent main issue is whether the adjustments in fundamentals are in place. From this perspective, due to slower growth and intensified competition, the adjustment of fundamentals in most areas is probably still insufficient, but from a bottom-up perspective, there may be opportunities to explore.
What does the financial report of the AI hardware leader reveal? One of the most important points in Nvidia's financial report is that 40% of its data center revenue in 2023 came from inference, with the demand for chips used in traditional recommendation algorithms shifting from CPUs to GPUs. This shift is significant for pricing; if the 40% inference ratio can be maintained, it would have a very positive impact. Training is akin to preparing shovels before mining for gold; there's a limit to how much demand for these shovels can grow, and eventually, it will plateau. If a large portion of revenue comes from training, then the current high growth rate is not sustainable and might even decline at some stage. However, if a significant portion comes from inference, then the sustainability of growth would be much better.
At a time when AI applications were not widely visible, the company indicated that a significant portion of its revenue comes from inference. However, this inference is not yet supported by AI applications but by traditional recommendation algorithms. Recommendation algorithms have already generated substantial revenue for major internet companies, implying that the demand for chips will continue. This undoubtedly provides significant relief to market concerns.
Through our research and analysis, we believe that the demand for GPUs for traditional recommendation algorithms can explain part of the situation, but the demand is more driven by training needs and customer hoarding. Regarding the sustainability of future demand growth, before AI applications complete a commercial loop, there might still be some room for growth, but it may not be significant:
Training demand primarily depends on large corporations' early investments. There's a bit more room for increased capital expenditure (capex), but more likely from structural adjustments.
The demand for chips for recommendation algorithms replacing CPUs with GPUs has limited space, as the total market size for the latter is only about $20 billion, and it's unlikely to be completely replaced in the short term.
The demand from enterprises and channels hoarding/double booking could also contribute significantly to growth, but unless AI applications complete the commercial loop, the sustainability of this demand is questionable.
Currently, the more important issue for pricing is the long-term market size for AI and NVIDIA's share in the entire value chain. If we reference the ratio of Intel and Microsoft's operating systems to the internet industry in the last internet era, NVIDIA's current market value of $2 trillion might already reflect an AI market size prospect of $10-20 trillion. If the market needs to grow significantly from its current $2 trillion valuation, then the market size would need to be much larger. However, if a $10-20 trillion market size is accepted, are AI application companies undervalued to some extent?
If AI applications are currently undervalued simply because there hasn't been significant progress, but the long-term potential is "set in stone," then our research should focus on companies in AI applications and AI hardware where the pricing is relatively insufficient.
Market Outlook & Strategy
On a macro level, domestically we are relatively cautious, while overseas we are relatively optimistic. However, we will maintain an open mindset and adjust our views according to the actual situation.
There has been no significant change in our macro outlook. Domestically, concerns about further overall valuation decreases have lessened, focusing more on whether adjustments to fundamental expectations are in place. The domestic focus is on exploring overseas opportunities.
Overseas, interest rate cuts are no longer a major concern; the valuation system is at least relatively stable. The main focus is on fundamentals. As long as the economy remains resilient, there are many opportunities.
Where are the opportunities?
AI is an important source of medium to long-term alpha. The value of human capabilities is increasingly recognized, and as AI enhances or even replaces human effort, its value correspondingly increases. This is just the beginning, and the potential is extremely high.
The pursuit of health and a lifestyle that places more emphasis on self-care will continuously present opportunities in consumption.
Post-inflation, the increase in purchasing power may lead to increased pricing space for some products/services, enhancing the competitive edge of products/services that offer high value for money.
If we enter a cycle of interest rate cuts, sectors that were previously hard to assess for interest rate-related risks due to continuously rising rates might see opportunities. Companies with good competitive structures and promising long-term prospects could find opportunities once short-term interest rate risks are addressed.
Comments