summation
- Google, a major customer of NVIDIA, has already developed its own chips since 2020 and is using them in its cloud, and many platform companies such as Amazon and Meta are actively working to reduce data center costs.
- It is expected that there will be additional measures to regulate U.S. exports to China. NVIDIA's dependence on China for sales is quite high at 20-25%.
- NVIDIA's current value assumes continued high growth of 30-40% over the next five years. It's not an impossible goal, but it's not an easy number either.
outline
previous post We looked into the reasons why you should buy NVIDIA based on its incredible performance. NVIDIA's dominance in the data center market is solid, and we believe this will continue for a considerable period of time. However, it is a different story from stock prices. Good performance does not necessarily lead to good stock prices. As is the case with many growth stocks, stock prices that were running well sometimes fall due to concerns about minor contraction, and sometimes such concerns are reflected in performance. This time, we would like to look into the concerns raised against NVIDIA.
Inspection point 1. Current customers are potential competitors.
As the main role in improving data center performance shifted from CPU to GPU, Intel's position as the absolute leader in data centers weakened, and AMD, which was considered a competitor to NVIDIA in the GPU market, was no match for NVIDIA in the data center field. The data center market, which has become even hotter before and after the advent of generative AI, has been a huge success for Nvidia, and its actual market share has also increased rapidly.
The fact that NVIDIA's products seem difficult to replace for a long time in the market triggered by generative AI is a story that can be established assuming that the existing structure does not change. As data center investment costs are increasing rapidly, many platform companies are directly producing AI semiconductors.
Google has already been providing it on its cloud platform using the TPU v4 model since 2020, and is known to be 1.3 to 1.7 times better in terms of performance and power consumption than NVIDIA's existing flagship chip, the A100. A TPU v5 model corresponding to NVIDIA's new lineup, H100, is also being developed.
In addition, MS is testing an AI-specific chip called Athena in OpenAI, as well as Amazon's Graviton and Inferentia. Tesla's D1, Meta's MSVP, and MTIA are developing and utilizing products specialized for each environment. Because the demand for generative AI is enormous, demand for NVIDIA chips is expected to remain solid, but since the initial purpose of platform companies to develop AI semiconductors is to reduce data center investment costs, efforts to expand the application of their own chips are also expected to continue.
Inspection point 2. Export restrictions to China
AI semiconductors are by far the core competitiveness of each country. Now, due to the excessively prolonged US export restrictions to China and Russia, it has become as if it is not a big deal, and although this is partly true, it is still something that must be considered when investing in NVIDIA. This is because about a quarter of Nvidia's sales come from China. On the company's second-quarter earnings call, management expressed confidence that any additional actions related to China would have no immediate impact.
Regarding Chinese regulations, NVIDIA is avoiding regulations by supplying A800 chips with reduced performance exclusively for China. However, China appears to be overcoming the lack of high-performance chips by bundling multiple low-performance chips, so it can be expected that there will be additional sanctions from the United States. Li Yanhung, founder of Baidu, a Chinese platform company, said that the technological level of Chat GPT and Baidu's 'Erniebot' is only 1 to 2 months apart.
Separately from this, China, like American platform companies, is also making efforts to produce its own chips. On the 27th, news broke that the Chinese company Huawei has succeeded in manufacturing a chip equivalent to NVIDIA's A100.. Of course, nothing has been revealed yet, and it is known that China does not have the technology to implement the 7nm-level A100 process, but it may be that the technology was secured through a third country that is not subject to US sanctions with China's full support. The article also introduced an official from the Semiconductor Industry Association (SIA) who said that China has secured two factories and is building three new factories.
A series of concerns related to China may not have as large an impact as Nvidia believes. However, for recent Nvidia investors who must guarantee high growth, even a small impact may come as a burden.
**Inspection point 3. It is true that it is overvalued even considering the high growth.**
In the last article, I mentioned that it was digestible based on the second quarter performance and guidance. However, this is based on the predictions of NVIDIA and analysts and may change at any time. Those who view NVIDIA's current value as reasonable assume that sales and net profit growth over the next five years will be more than 30-40%.
NVIDIA sales, earnings growth and forecasts (Source: StockRow)
The table above shows NVIDIA's current and future sales and net profit growth forecasts. Considering that forecasts are being raised after the second quarter performance guidance, the premise of 30-40% does not seem like an extremely unreasonable goal. Assuming that the above forecast remains unchanged and sets the net profit growth rate for 27-28 at 35%, NVIDIA's PER after 5 years will be approximately 24, which is the current market average. It is difficult to predict the PER after 5 years, but assuming that it is still treated as a growth stock and assuming NVIDIA's average PER of 50 over the past 10 years, it means that the company has roughly double the upside potential over the next 5 years.
However, this is only an assumption. That's just my simple calculation assumption. Can the growth rate, which has fallen since 2024, rebound to 35% in 27-28? If not, will the forecasts continue to be raised in a direction that will slow down the falling growth rate?
Regardless of NVIDIA's future, there is no change in the fact that the current stock price reflects the future value in advance, at least to some extent. Moreover, there is another fact to consider. The big players who own Nvidia have already made a lot of money. If a situation arises where risky assets need to be offloaded, it can be the most actively sold stock.
conclusion
We maintain the judgment from the previous article that it would be a waste to filter out only the value indicators. I believe that the current overvaluation can be sufficiently resolved. However, the area where comfortable investment was possible has passed too far. I think it is necessary to pay continuous attention to corporate performance and forecasts and respond promptly.
>
Related article: NVIDIA (NVDA) 2nd quarter earnings release review (Q2) Earnings)