[Report on NVIDIA China Customer Rules] I won't sell even if semiconductor stocks fall! The reason for distinguishing between demand and customer business

Just as semiconductor stocks fell again, some concerning news regarding NVIDIA has emerged.

The report states that NVIDIA is asking overseas cloud companies to sell at least 51% of their services to companies outside of China. It is news that makes one think, “It might not be enough to just not sell directly to China.”

However, I do not intend to sell NVIDIA based on this report alone. I also won’t change my bullish outlook on AI-related semiconductors.

The reason is that the story of people wanting to use AI and the story of it becoming harder to sell to those who want to use it are two different things. Of course, the latter cannot be ignored either. If the business of cloud companies buying GPUs gets clogged, there is a concern that subsequent semiconductor orders will be postponed.

This news becomes much easier to understand if you grasp that distinction.

Distinguish between the continued decline in semiconductor stocks and this report

First, let’s talk about the stock price. In regular U.S. trading on November 2, the semiconductor stock ETF SOXX fell about 3.79% from the previous day’s closing price. NVIDIA’s closing price was $202.06, down about 1.26%.Stock price data

When the sell-off of semiconductors continues, people get anxious, wondering if “AI investment is over.” If you add negative headlines about China to that, I understand the urge to sell even more.

However, the Yahoo Finance article that carried the FT report is dated November 3. The update is displayed as 19:51 Japan time, which is after the previous day’s regular U.S. trading session ended.

Therefore, I would not explain that “this report caused the decline on November 2.” It is reported that communication with customers had been taking place for some time, but it cannot be determined from this alone how that affected the previous day’s stock price.

The stock price has already fallen. On the other hand, what is the new information available now? I want to make sure I don’t mix up this order.

In the morning article, I covered Palantir’s profit growth, but the point of this discussion is different. It’s not about “whether there are companies that can make money with AI,” but rather, who the companies buying NVIDIA chips can sell their services to. This is the crucial point.

I don’t want to rush to switch stocks just because of the size of the decline. On days when semiconductor stocks are being sold off across the board, I want to separate the new facts that have occurred in the business from the movements of the stock price.

Even if you don’t sell directly to China, it affects your customers’ business

Cloud companies buy GPUs and rent out their computing power to users. You can think of it as a service where you can rent as much as you need instead of buying expensive equipment yourself.

For example, suppose I open a rental shop that owns GPUs. I stock the equipment, rent it to customers, and collect payment through usage fees.

What happens if the supplier tells me, “There are also conditions for the composition of your renters”? I have to think not only about whether I can get the equipment, but also whether I can gather customers who meet the conditions.

According to the article, NVIDIA is asking cloud companies to ensure that at least 51% of their sales are to companies outside of China. It is reported that this applies to future new contracts and will not be applied to existing rental agreements.

The important thing is that this is not an announcement that Chinese AI demand has disappeared. It is a story about concerns that even if there are companies that want to use it, it will become harder to expand business due to customer composition conditions.

That said, it is also dangerous to arbitrarily decide the denominator of the 51%. I have no intention of arbitrarily calculating the impact on each company by concluding whether it is “51% of sales,” “51% of GPUs,” or “51% of the number of customers” based solely on the information in the article.

The necessary response will change depending on how the reported conditions are defined in actual contracts. The burden should differ between cloud companies that are skewed toward Chinese users and companies that are already selling widely across the globe.

This is my own assessment, but if they need to develop new customers, the sales effort will increase. There might be situations where they lower usage fees to fill idle GPUs. If that happens, the time it takes to recover the equipment costs will lengthen.

If that happens, it also becomes a reason to push back the next GPU purchase. As an NVIDIA shareholder, what concerns me is how far this trend will spread.

Furthermore, the specific conditions this time are based on reports from interviews with involved parties. NVIDIA’s response published in the article also did not explicitly acknowledge the 51% condition. I am not treating this as a story about the government announcing a new law.

Even company forecasts that exclude China sales cannot be considered safe

Here, I am reminded of NVIDIA’s company forecast. Some people might think, “If they aren’t including China in the first place, doesn’t this story have nothing to do with them?”

In the fiscal year 2027 second quarter earnings released on August 26, NVIDIA forecast revenue for the next quarter at $108 billion, with a range of plus or minus 2%. At that time, they explained that they did not include revenue from computing products for data centers destined for China in the forecast.

But “revenue from computing products for China” and “demand from Chinese companies using overseas clouds” are not the same scope.

Even if the cloud companies buying chips from NVIDIA are not Chinese, their users might include Chinese companies. This report concerns the composition of the end-users.

In other words, even if the portion of chips sold directly to China is excluded from the company forecast, the possibility remains that if overseas customers reduce their capital investment, it will have an impact through a different channel.

“Not including China in the forecast” is different from “having absolutely no China-related impact.” This is what concerned me most this time.

On the other hand, just because there is this difference, it is a leap to assume that all sales to overseas clouds are at risk. Which companies will be affected by the conditions, and how much will they reduce their planned purchases? Unless we know that much, we cannot calculate the revenue loss.

The $108 billion figure was also a forecast as of August, not a figure newly guaranteed in response to this report. I do not think “it’s unscathed because it’s not in the forecast,” nor do I think “they will lose everything destined for overseas.”

What is important is not just where NVIDIA delivered the products, but whether those customers can recover their usage fees and buy the next set of equipment. It might feel like a roundabout way, but the meaning of the news changes when you consider the profitability of the businesses using the GPUs.

Moving beyond GPUs is not a solution either

“In that case, I should just stop holding NVIDIA and buy other semiconductor stocks.” I do not subscribe to this line of thinking either.

Data centers do not run on GPUs alone. Communication equipment and semiconductors that connect large numbers of computers are also necessary.

For example, if a cloud company pushes back its plans to increase new equipment, orders for network equipment used alongside GPUs could also be delayed. This is a point that cannot be ignored when considering related companies like Broadcom.

Of course, not all companies sell to the same customers in the same proportions. Some companies have other businesses like software, and the adoption of their products also differs. Therefore, I am not saying the magnitude of the impact is the same.

But if there is a portion that relies on the capital investment of the same cloud companies, you cannot say “it’s safe because it’s not a GPU.” Just changing the stocks you invest in does not necessarily mean you can escape the risks you were worried about.

I support AI semiconductors, but I think separately about what to hold within that sector. This news alone is not enough evidence to place Broadcom above NVIDIA, and I have no intention of creating a new buying priority list.

Instead, I want to use this as material to think about where customers’ businesses might get stuck. Are they unable to gather enough users? Will usage fees drop? Will the timing of equipment purchases be delayed? Depending on the problem, the companies involved and the impact on profits will change.

Conversely, if equipment is filled by demand outside of China, it is easier to continue investing even under the same conditions. I do not intend to be uniformly pessimistic just because there are conditions; I intend to think about it by linking it to actual changes in orders and sales.

To go a step further, just because a customer moves to a different cloud provider does not necessarily mean there will be additional orders for NVIDIA.

If there are available GPUs at the destination, they can use those first. If it is just a matter of a different company taking on the same work, the demand for computing may remain, but sales of new chips may not increase.

Therefore, I do not consider the amount of AI usage and the new purchase of GPUs to be the same thing. Even if I maintain the view that computing is necessary in the long term, I think separately about whether short-term sales will grow as planned.

If I am thinking about my own holdings, the order is as follows.

• Has the number of companies wanting to use it decreased, or have the usage conditions changed?

• Is existing equipment sufficient, or are new GPUs necessary?

• Is the purchase schedule just being delayed, or are the orders themselves disappearing?

Because of this difference, the timing of when sales come in and the future growth rate will change. If you simply dismiss it by saying ‘AI demand is strong so it’s fine,’ I don’t think you can apply the news you’ve read to your investment decisions.

I will not sell NVIDIA based on this report alone.

So, why won’t I sell? Because I don’t believe the reasons I have for expecting things from NVIDIA have been undermined by this report.

In the financial results announced in August, sales for the quarter ending July 26 were $96.221 billion. This was a 106% increase compared to the same period last year. Data center sales were $89 billion, an increase of 117%. These are the results announced by the company, not my estimates.

Of course, just because it grew significantly in the past does not mean it will continue to grow in the future. Even so, what we learned this time is not that ‘computing to run AI is no longer needed,’ but that there is a possibility of restrictions on customers’ sales destinations.

I maintain the view that work using AI will increase and demand for computing equipment will also increase. If orders shift to companies that can gather enough customers outside of China, there is room for NVIDIA’s products to be chosen even if users and sales channels change.

However, this is my investment judgment, not a guarantee that a recipient for the demand will definitely be found. If customer purchase delays and cancellations spread and even NVIDIA’s sales forecasts are lowered, I will not be able to push through on long-term demand alone.

I will change my mind when the impact on the business becomes concrete like that. It is not because the news headline is scary, but if the orders and profits I expected really decrease, I will reconsider my reasons for holding.

My answer right now is: Continue holding. I will not move to sell based on this report. On the other hand, I will not increase my holdings just because the price has dropped. Because what can be expected in the long term and whether today’s stock price is undervalued are separate judgments.

Being bullish on a day when semiconductor stocks are falling is not about pretending not to see bad news. What is newly understood, and what is still unknown? Based on that, I choose to keep my expectations for AI semiconductors.

---

Portfolio changes in this note post

There were no trades this time. There is no change in the number of shares held.