January Market Review

January ended with significant increases in all markets.

• S&P 500: +2.70%.
• Nasdaq: +2.22%.
• Stoxx Europe: +6.29%.
• All Country World Index EUR: +3.90%.
• Global Fixed Income Index EUR: +0.18%.

In equities, the prize for the month of January goes to Europe with a rise of 6.29%, well above the rest of the indices. Why? I believe the main reasons are as follows:

• Fears about US tariffs were reduced, especially after the easing of Trump’s stance on trade policies.
• The emergence of Chinese artificial intelligence “startup” DeepSeek led to a significant drop in U.S. tech stocks, especially Nvidia, whose market value dropped nearly $600 billion in a single day. To put it in perspective, Spain’s GDP is approximately $1.6 trillion.
• The European Central Bank (ECB) cut interest rates by 0.25 points, leaving them at 2.75%, with the aim of reviving the eurozone’s stagnant economy. ECB President Christine Lagarde hinted at the possibility of further cuts, which further increased optimism in the markets.

What has happened in the American technology sector?

This month we have witnessed the collapse of Nvidia, one of the most prominent companies in recent years, which lost almost $600 billion ($600,000,000,000,000) in a single day due to the arrival of DeepSeek, an artificial intelligence (AI) assistant developed in China. To put the magnitude of this drop in perspective, Spain’s GDP is $1,900 Bn, which means that the loss represents 37% of that GDP. Not only Nvidia was affected, but the rest of the companies linked to AI such as Broadcom (-17.4%) and AMD (-6.37%), which caused a 3% drop in the Nasdaq.

I know I have already talked about Nvidia on other occasions, and I am sorry to repeat myself, but I consider it key to understand what is currently happening. Nvidia dominates the GPU (Graphics Processing Unit) market with a 90% share, while Advanced Micro Devices takes the remaining 10%. In the data center GPU sector, its share amounts to an impressive 98%. Why is this information relevant? Because the most expensive component in data center construction is precisely GPUs, which account for almost 50% of total spending¹.

Nvidia expects sales of $47 billion in 2024 in the Data Center segment alone. Where do these sales come from? According to Bloomberg, Microsoft accounts for 18.9% of the total, Meta (formerly Facebook) accounts for 10%, Alphabet 6%, Amazon 5.5%, and Tesla 2%. In other words, with the exception of Apple, the rest of the “Magnificent 7” account for 42% of Nvidia’s sales. Based on this data, Microsoft would be paying almost 9 billion dollars to Nvidia for its data centers, and so on for the other companies. In short, the market for data centers and GPUs is huge to this day and should continue to grow… or so we believe.

Suddenly, the Chinese startup DeepSeek, which you’ve probably heard of lately, made its appearance. This artificial intelligence assistant has shaken up the global AI industry, generating both excitement and concern.

Until recently, the United States seemed to be at the forefront of the AI race. President Trump had announced “Stargate,” a $500 billion AI infrastructure megaproject backed by key figures such as OpenAI’s Sam Altman and SoftBank’s Masayoshi Son. However, the launch of DeepSeek has drastically altered this picture.

DeepSeek in its “debut” quickly climbed to the top spot in Apple’s App Store, surprising investors and significantly affecting the value of tech stocks. Its appeal lies not only in its performance, comparable to ChatGPT-4, but also in its surprisingly low cost. While OpenAI invests more than $100 million in developing its advanced models, DeepSeek claims to have built its application in just two months with less than $6 million. If this is true, the change in the industry would be monumental: less investment and thus many companies like Nvidia would be severely affected, as reflected in its 17% drop on the day of the announcement. However, as the daughter of a friend of mine says: “Mom, what you buy in a Chinese shop, breaks the next day”. So, beware of betting on the cheap.

Regardless of whether or not DeepSeek will turn out to be a success story, this phenomenon calls into question the valuations of technology companies, especially those linked to artificial intelligence.

There are several multiples we use to assess the valuation of companies. Two of the most common are P/E (price to earnings) and P/S (price to sales), which help us determine whether these companies are overvalued or not. To make this comparison, we contrast them with the multiples at which the U.S. S&P 500 trades.

It seems that these companies are overvalued compared to the rest of the market, considering that the historical average P/E of the S&P 500 over the last 25 years has been 18x and the P/S has been 1.85x.

Is it reasonable for them to trade at these multiples? Some investors are willing to pay these multiples because of the high margins and their great optimism about the future growth of these companies. If they can sustain that pace of expansion and margins, it might make sense, but those multiples imply an extremely optimistic scenario. In fact, Nvidia alone trades at a P/E of 47x and a P/S of 26x.

Thus, the drop in the market for these companies is better understood. There is an excess of optimism, and just a small disruption is enough to trigger a stock market tsunami, a phenomenon that is not new and has been repeated in the past.

Every time I see these valuations, I am reminded of what the CEO of Sun Microsystems said after the collapse of the tech bubble at the turn of the century, referring to his own company: “At a valuation of 10 times earnings, to give you a 10-year payback, I would have to pay you 100% of earnings in dividends for 10 years in a row. That assumes my shareholders would approve. That assumes I have no cost of goods sold, which is very difficult for a hardware and software company. That assumes I have no expenses, which is virtually impossible with 39,000 employees. That assumes I pay no taxes, which is very difficult for a hardware and software company. That assumes I have no expenses, which is practically impossible with 39,000 employees. That assumes I pay no taxes, which is very complicated. And that assumes that you don’t pay taxes on dividends, which is basically illegal. And that assumes that, without investing anything in R&D for the next 10 years, I can maintain the current level of income. Now, having said that, do any of you want to buy my stock at $64? Do you realize how ridiculous those basic assumptions are? You don’t need transparency. You don’t need footnotes. What were we thinking?”

This company, which reached a $200 billion valuation in 2000 ($64 per share), ended up being sold to Oracle for $7.4 billion, representing a -96.3% drop.

An investor’s worst enemy is himself. Biases, both emotional and cognitive, can ruin the best analysis. Daniel Kahneman, Nobel laureate in economics and author of the book Thinking, Fast and Slow, talked about these biases. When I read it, I realized that I fell into several of them, which helped me to be more aware and to try not to get carried away by them. 

In the case at hand, we seem to be witnessing what the author called the “herd effect”. This phenomenon occurs when investors follow the majority without analyzing for themselves whether the decision is correct. It is based on the belief that if everyone is doing it, it must be a good idea.

Are we experiencing a herd effect at the moment? I don’t know for sure, but what is clear to us at Altum is that, at the risk of being wrong, we do not think it is prudent to invest at these multiples.

1. AI’s $600B Question | Sequoia Capital ↩︎