laitimes

Nvidia, is the stock price going to leak?

author:Finance

As Nvidia's share price soared from $108.07 to $419.34, with a market capitalization of more than $1 trillion, its rally began to weaken.

Nvidia, is the stock price going to leak?

Some long-term Nvidia investors began selling stocks and took profits:

The Rothschilds reduced their holdings in Nvidia. Benjamin Melman, global chief investment officer at Edmond de Rothschild, revealed that the company has been overweighting Nvidia since late 2020, but has now partially taken profits and now holds a position that is "much smaller." The Geneva-based asset management institution is owned by the Rothschild family, a veteran European financial family with more than 200 years of history, and currently manages assets of CHF 79 billion.

Cathy Wood's ARKK, the flagship fund of Ark Investment Management, liquidated its Nvidia stake as early as January; On Thursday, Wood's other four funds sold more than 65,600 shares of Nvidia stock, worth nearly $25 million at closing prices.

Damo Dalan, which has held NVIDIA since 2017 and has a reputation as a "valuation master", also sold NVIDIA. "The market value has increased by $300 billion in a week, which is challenging the absolute limit of sustainable value," he said. ”

On June 7, according to the US Securities and Exchange Commission (SEC), NVIDIA director Harvey Jones reduced his holdings of about 70,000 shares on June 2, cashing out 28.433 million US dollars, which is equivalent to more than 200 million yuan.

Management also began to reduce its holdings, which was interpreted by the market as a signal that Nvidia's short-term stock price peaked.

However, there are also awesome characters who come out to sing a lot and support Yingweida.

On Wednesday, Stan Druckenmiller, the billionaire founder of the Duquesne family office, told him that he plans to hold NVDA.USNvidia for years, not months, and said: "If Nvidia is really as big as I think, we will hold it for at least 2 to 3 years, if not longer, not 10 months."

From the performance point of view, NVIDIA achieved operating income of 26.974 billion US dollars, gross margin of 56.9%, operating profit of 4.224 billion US dollars, net profit of 4.368 billion US dollars, and net cash inflow from operating activities of 5.641 billion US dollars in fiscal year 2023 (Fiscal 2023, February 1, 2022 to January 31, 2023). At the end of fiscal 2023, the company's total assets were 41.182 billion US dollars, and the asset-liability ratio was 46.33%. In fiscal year 2024 Q1 (early February 2023 to the end of April 2023), operating income was US$7.192 billion, gross margin was 64.63%, net profit was US$2.043 billion (almost half of FY2023 full-year), and net cash inflow from operating activities was US$2.911 billion. Among the four core sectors, namely data center business, gaming business, professional visualization business and automotive business, the data center business and automotive business are growing rapidly, which is basically in line with the current popularity of artificial intelligence and smart cars.

In fiscal 2023, data center revenue reached $15.005 billion, an increase of 41% over fiscal 2022. Growth is fast, but when the stock price rises to 200 times P/E and nearly 40 times P/B, valuations are completely decoupled from the above performance data - not based on current performance, but based on the market's imagination of its future performance.

So is Nvidia expensive or cheaper?

Will its future performance be able to afford this 4x increase in share price?

How big is the market space driven by the AI "arms race"?

No one can predict accurately.

This makes NVIDIA's valuation almost "metaphysical".

The "valuation guru" Dharma Daran gave a vague prediction:

Nvidia currently has an 80 percent share of the $25 billion AI chip market, he said, and the most optimistic forecast is that the market will reach $350 billion in 10 years. Even assuming that Nvidia reaches 100% market share in the future, its reasonable valuation is 20% lower than its recent share price.

He is skeptical that NVIDIA can secure the trillion club.

The logic is that other members of the "trillion dollar club" (Apple, Amazon, Google, Microsoft) use software to attract large numbers of end users into their ecosystems, which creates what statisticians call a "long tail distribution" of business opportunities — multiple ways to profit from new products and services.

Nvidia, which is essentially a hardware company, selling chips to other companies, faces a natural limitation: the tail tends to be more cautious. Then the benefits are not as great as a consumer-based company with an ecosystem of billions of users.

The problem is that what Damodharan calls the "most optimistic forecast" is a 14-fold increase in market size in 10 years (from $25 billion to $350 billion).

14 times in 10 years, although it sounds awesome and has pushed the limits of our current growth rate for the general industry market, but is it not the "most optimistic forecast"?

Not necessarily.

At least in the history of semiconductor development, not the most optimistic forecast.

In 1958, Texas Instruments engineer Jack Kilby fabricated multiple electronic components on a single semiconductor material, the world's first "integrated circuit", for which he later won the Nobel Prize.

In 1959, Robert Noyce, co-founder of Fairchild Semiconductor (and later the main founder of Intel), also created an integrated circuit. At this time, Noyce did not know Kilby, and unlike Kilby who made a countertop transistor on a germanium substrate and connected it with bonding wires, his version was to cover the transistor with a layer of silicon dioxide insulation through a planar process, and realize the connection between transistors by depositing metal wires on it, and placing the "wires" directly on the ghost chip. That is, Noyce's version has no bonding wires.

The American Apollo program decided to use Fairchild chips to reduce the space occupied by the computer heavyweight that Apollo 11 brought to the moon. Because of this opportunity, chip sales to NASA's Apollo program transformed Fairchild from a small startup into a company with 1,000 employees, and its sales soared from $500,000 in 1958 to $21 million in 1960. 42 times in two years!

Similarly, the U.S. military's Minuteman II missile program revolutionized Texas Instruments' chip business. In 1959, Texas Instruments' integrated circuit sales were measured in the number of dozens, but very many were measured in thousands. By the end of 1964, Texas Instruments had supplied 100,000 integrated circuits to the Minuteman missile program. In less than 5 years, how many times is this?

As we all know, AI applications rely on the dynamic analysis and processing of massive amounts of data, which requires the computer with which it partners with huge memory and ultra-high-speed computing performance. In this regard, NVIDIA has a number of core technologies and core competitiveness, its chips and software can meet the compute-intensive needs of generative AI, known as the "king of computing power" of AI, and become the "largest arms dealer" in AI wars. Its newly released superchip GH200Grace Hopper for AI and supercomputer DGX GH200 have improved amazingly in terms of computing power, bandwidth, and energy consumption indicators.

Market analysts pointed out that although NVIDIA dominates the computing power, it also has wolves and tigers in the back - AMD, Intel and other old rivals are pressing step by step, especially AMD CEO Su Zifeng at CES 2023 to show a strong will to advance. AMD's product Instinct M210 was quoted at about $17,000 at the time of release, while NVIDIA's main H100 exceeded $40,000, with specific performance on Twitter ranging from 3.9-46,000 yuan, and the cost performance was obviously behind. Compared with the mainstream GPU solutions in the market, AMD's heterogeneous combination of CPU and GPU can achieve higher computing performance and energy efficiency to meet more complex computing needs, and may be able to achieve the effect of "overcoming rigidity with softness" in the AI computing power competition. In addition, Google, Microsoft and other partners are also developing their own chips.

When ChatGPT matures, AIGC will no longer be a game for high-end players, but will have to meet the needs of the mass market. At the same time, more segmented application scenarios also require more customized chip solutions. Therefore, the AIGC computing power base is likely to gradually tilt from the situation of NVIDIA alone to the division war of the head enterprise "hundred schools of thought", thereby compressing NVIDIA's profit space in this field.

This is one of the main reasons for the recent bearish bearish NVIDIA - this statement makes logical sense.

However, we can see from the history of the birth of integrated circuits that despite the competition, because the market is large enough, Texas Instruments, Fairchild and other companies have ushered in explosive growth, not contention but together.

The focus of NVIDIA's future performance and stock price may not be on the competitive landscape, but on what kind of development state AIGC represented by ChatGPT will be.

On May 18, Ma Huateng said at the 2023 shareholders' meeting, "At first, I thought that AI was an opportunity that the Internet would not encounter in ten years, but the more I thought about it, the more I felt that this was an opportunity similar to the industrial revolution that invented electricity that has not been encountered in hundreds of years." ”

If it is really an opportunity that "does not happen in hundreds of years" as Ma Huateng said, or an opportunity for explosive applications such as the Apollo and missile programs in history, NVIDIA's recent stock price correction will be just a small wave in a vast chapter in the long run.

This article is from Zhitong Finance APP