1/ Reading more financial reports (annual reports) of listed companies will help break your own wrong prejudices and establish more accurate intuition, richer connections and imagination. We must work hard at this and cannot be lazy.
2/ In fact, if you spend an average of one or two hours a day looking at a company's financial report, especially comparing the data changes in the past ten years, you will often gain new insights, and you will immediately realize what you got into during your previous analysis and research. Such misconceptions.
3/ By taking the time to read financial reports, you can restrain your urge to do various short-term transactions (with negative long-term returns). And because the brain has established a more refined cognitive picture of hundreds of investment objects, it will raise its own opportunity cost higher and higher, and it will naturally be indifferent to most so-called "opportunities"
4/ Studying financial reports, especially financial reports in the U.S. capital market, is essentially a study of Game of Thrones. Because it contains numbers that people with real money have gambled in the market, the data is large and detailed, and the authenticity is relatively high.
5/ Financial reports are essentially a detailed map of the power structure of the capital world. By looking at the changes in the same company's financial report data over decades, comparing the financial data of competitors in the industry, and comparing the financial data of partners in the industry chain, you can better understand the laws of power changes and become more sensitive to the future direction of power.
6/ Even if it is financial fraud, the fraud itself has costs. The fraud will eventually be exposed. Before the fraud is exposed, there are various clues in the financial report that can be sensed in advance.
7/ Instead of carefully studying financial reports, reading various media every day that lack figures, lack or distort context, quote out of context, talk nonsense, brag without consequences, and become a fool without knowing it, is a natural outcome. .
8/ Cisco (stock code: CSCO) was the leader in the field of communication equipment during the Internet boom in the late 1990s. Its fiscal year ends on July 31 of each year. In the annual financial report at the end of July 1999, annual revenue was US$12.1 billion and profit per share EPS was US$0.29. In August of that year, the stock price was around US$30, and the PE ratio was 100. (General financial data will be released within one month after the end of a quarter. out, so the share price quoted is August.)
9/ A year later, in the annual financial report at the end of July 2000, its revenue was US$18.9 billion, EPS was US$0.36, the stock price in August was 65, and the PE ratio was 180. Just five months ago, in March 2000, Cisco's stock price Reaching a record high of $80, it has yet to be surpassed.
10/ Cisco's revenue in 2001 was US$22.2 billion. However, due to severance pay due to large-scale layoffs, the reduction in the book value of acquired assets, and the reduction in the value of a large amount of excess inventory, Cisco suffered a loss in fiscal year 2001.
11/ In the financial report at the end of July 2002, EPS shrank to US$0.26, the stock price in August was 13 (84% lower than the highest point), and the PE ratio was 50.
12/ 21 years later, when Cisco's financial report at the end of July 2023 came out, the revenue was US$57 billion, eps $3.07, the stock price in August was 55, and the PE ratio was 18. One reason for the low PE ratio is slow growth. Revenue in 2023 is only 15% higher than three years ago. Therefore, the market's expectations for its future growth are also low.
13/ In other words, in the 23 years from August 2000 to August 2023, Cisco’s revenue increased by 200% and earnings per share increased by 750%, but its stock price fell by 15%. (Cisco’s annual growth rate since 2011 Cash dividends are paid to shareholders, averaging about 1 – 3% per year, which is temporarily ignored in this discussion.) If you go back to August 2000, when Cisco’s stock price was still $65, you would tell people of that era that although Cisco was It's a very good company. Although Cisco's profit per share will increase more than seven times in 23 years, the stock price will fall by 15%. Will those people think you are crazy and tell you to get out immediately?
14/ This string of data is worth thinking about for investors who “only look at fundamentals and don’t think about valuation logic”.
15/ NVIDIA (stock code: NVDA) is the undoubted leader in the field of artificial intelligence chips. It is a very good company, but a careful analysis of its valuation shows that investors who currently enter the market at US$417 per share will have investment returns in the next five to ten years. The rate (compared to the S&P 500) may not be very high. There are two fundamental reasons:
Revenue growth may be lower than already very optimistic expectations,
Slowly and eventually competing alternatives will emerge.
16/ In the most recent quarter ending at the end of July 2023, NVDA’s revenue was US$13.5 billion, (of which revenue from mainland China, Hong Kong and Taiwan accounted for 41% of its total revenue; revenue from artificial intelligence applications used in data centers was US$10.3 billion US dollars, the revenue of this segment is expected to exceed US$50 billion in the next year), approximately double that of the same period last year. Net profit was US$2.48 per share, 9.5 times that of the same period last year. This is the main driving force behind the surge in stock prices starting at the end of May. Management expects revenue for the next quarter to be approximately $16 billion (up 18.5% from the current quarter). The market expects NVDA's profit per share to be US$10.7 in the fiscal year ending in January 2024; based on the current stock price of US$417, the P/E ratio is 39. The market expects NVDA's monopoly on the AI market to be unbreakable.
17/ NVDA customers spend huge sums of money to buy GPUs, and ultimately need to make profits at the application level to be sustainable. The explosive growth in purchase demand last quarter originated from the emergence of ChatGPT at the end of 2022. It was a panic-buying behavior regardless of the cost. The panic buying caused the gross margin of NVDA chips last quarter to drop from the usual 65 % increased to more than 70%.
18/ David Cahn of Red Shirt Capital recently published this analysis: every dollar spent on GPU corresponds to one dollar of energy consumption cost for the data center (cloud service provider). If NVIDIA sells $50 billion in GPUs for AI applications to data centers every year, that corresponds to a cost of $100 billion in data centers. Companies that provide artificial intelligence applications, whether it is ChatGPT, Microsoft's Copilot, Tesla, Xai, Midjourney, Stable Diffusion, etc., must ultimately make a profit based on this $100 billion cost. Assuming their gross profit margin is 50%, This means that they must earn US$200 billion in revenue from end users every year to be sustainable. Cloud service providers that provide services to third-party application developers also need to make profits. For example, AWS's gross profit margin is about 30%, which will not be considered here.
19/ Where does $200 billion in revenue come from? Microsoft said it will earn US$10 billion a year from Copilot. Assume that Google, Apple, and Facebook will also earn US$10 billion a year from artificial intelligence, which is US$40 billion. OpenAI says it can now generate $1 billion in revenue a year. Assume that Oracle, Tesla, Tencent, Alibaba, Twitter, ByteDance, and other companies can add $5 billion in annual revenue from AI applications. The total adds up to US$71 billion, which is still close to US$130 billion away from US$200 billion in revenue. What should I do if the income gap cannot be bridged? First, you can tell stories to draw a BTC and collect money from the capital market to fill this gap. There is really no money to make up for it. The third-party application service provider has gone bankrupt and the data center has reduced its GPU purchase budget. This is the only way out.
20/ Competition at the AI application level is like a gold rush. At the beginning, everyone rushed in regardless of the cost and raised the price of shovels. Analysts had overly optimistic expectations about the future demand for shovels and the gross profit margin of selling shovels. After all, there are only a few winners in the end. Most of the unprofitable application developers have withdrawn from the market. The growth rate of demand for shovels (GPU) has slowed down and there is too much inventory. The price and gross profit margin of shovels will inevitably decline.
21/ The peak of publicity about ChatGPT was probably around May 2023, when its website had as many as 1.8 billion visits in a single month; the number of active users dropped by 10% month by month in July and August, although some observers attributed this to Students are on summer vacation, but this at least shows that the current market application has temporarily entered a bottleneck period. There is still a long way to go to make money on AI applications.
22/ Midjourney, the once extremely popular AI drawing application, had more than 40 million website visits in April 2023, but it had shrunk to 27 million in July. The user interface of Midjourney is not user-friendly, forcing users to generate images through messy Discord. Recently, ChatGPT demonstrated the latest interface that can generate high-quality images without going through Discord, which is undoubtedly ringing the death knell of Midjourney.
23/ Another potential problem with NVDA is that analysts’ future valuation of it assumes that it can continue to maintain a market share of more than 90% and a gross profit margin of more than 70% in the AI chip market in the next few years. NVDA's monopoly mainly comes from the software ecosystem called CUDA (Compute Unified Device Architecture). This NVDA GPU-based software platform launched in 2007 has accumulated thousands of software instructions specifically optimized for NVDA GPU performance over the past decade, allowing developers to get started quickly and allowing developers to rely on NVDA GPUs. In a short period of time, it will be an insurmountable barrier for potential competitors.
24/ However, several of NVDA’s largest customers, such as Microsoft, Google, Amazon, Facebook, etc., are clearly developing their own exclusive artificial intelligence chips. Google has TPU, Microsoft and AMD are also cooperating, Amazon has graviton in the cloud, and so on. The largest customer is also developing its own competing products. This is a relatively subtle and special situation. Is Apple’s biggest customer making its own mobile phones? Is Tesla's biggest customer building its own electric car? Is Oracle's biggest customer developing its own database? The most recent quarterly financial report shows that a certain cloud service provider accounted for as much as 22% of Nvidia's single-quarter revenue. These large customers will not be willing to continue to contribute more than 70% of gross profits to NVIDIA throughout their lives and do nothing.
25/ Although the CUDA software ecosystem is relatively friendly to developers and is far ahead of its competitors for the time being, the gap with potential alternatives cannot remain so big for five or ten years. In addition to dedicated AI chips developed by large companies, there are also AMD's ROCm, the PyTorch Foundation spun off from Facebook, and OpenAI's Triton, all of which are trying to get rid of their dependence on the CUDA software platform, ultimately eroding NVDA's market share and gross profit margin.
26/ Even if NVDA’s market share drops below 80% and its gross profit margin drops to 65%, it still has a strong position in the AI chip market, but its profit per share, profit growth expectations, and valuation calculations have to be readjusted . Based on Cisco's experience, an optimistic estimate assumes that NVDA's profits in 2046 (23 years later) increase by 750% compared to 2023, which is US$91 per share, but the PE ratio drops to 18, corresponding to a stock price of US$1,638. Compared with the current 417, the annualized return is approximately 6.1%. Not too bad, but not much better than the 5.5% interest rate on short-term Treasury bills.
27/ And what about the pessimistic scenario? You can see the trend of Cisco's stock for more than ten years from March 2000 to mid-2011, which is enough to completely destroy the confidence of any investor who "only looks at fundamentals and does not study valuation logic."
——–
About the author: Wang Chuan, investor, currently lives in Silicon Valley, California. WeChat ID 9935070, Twitter account "Svwang1", Sina Weibo "Silicon Valley Wangchuan", website chuan.us. All articles express the author's personal opinions for reference only and do not constitute investment advice for the assets mentioned. Investments are risky, and entry into the market is subject to cautious.
Click the link below to see the link directory of more than 200 original articles in this public account in the past:
List of articles on Wang Chuan’s public account in Silicon Valley (July 2023)
How to brainstorm and make things easier on the road to investment and entrepreneurship? Welcome to Wang Chuan’s club, a high-end paid community that gathers elites from all walks of life from all over the world with independent thinking and unique perspectives. For details, please click on the article link below
RTFM – Instructions for Investment Club (Second Edition)
If you are interested in applying for membership, please contact Wang Chuan (WeChat ID: 9935070 or Twitter "svwang1" private message) directly.
