Is Nvidia the next Cisco?
Latest update time：2024-09-08
Reads：
Edited by CoolPlay Lab
First published on WeChat public account:
Cool Lab (ID: coollabs)
"Before the Internet bubble, we also believed that Cisco would continue to grow."
This was how Wall Street investors evaluated Cisco at the time.
Perhaps the PTSD brought by Cisco was too great. After Nvidia released its second-quarter financial data for fiscal year 2025
(natural year ending July 2024)
, even though all businesses were growing, investors still left.
Because Nvidia's growth rate is slowing down significantly.
Nvidia achieved revenue of $30.04 billion this quarter, a year-on-year increase of 122.4%.
This growth rate is already out of reach for many companies, but it is still a significant slowdown compared to Nvidia's year-on-year growth rate of more than 200% in previous quarters.
What may make investors even more pessimistic is Nvidia's guidance for the third quarter.
In the next quarter, Nvidia expects revenue of $32.5 billion.
Although this is higher than market expectations, it is equivalent to a year-on-year growth rate of less than 80% in the next quarter, which makes investors exclaim "Not good, quick profit".
After the financial report was released, Nvidia's stock price fell by
more than 7% before the opening of the next trading day.
In fact, ever since Nvidia's growth rate began to soar, the market has never stopped discussing when Nvidia's bubble would begin. This collective withdrawal of investors also reflects Wall Street's caution towards such upstream technology companies.
However, judging from the current situation, the AI industry is still in its early stages of development. Nvidia's major customers are increasing capital expenditures related to AI chips and cloud deployment. Downstream demand is still very strong. It is normal for investors to leave the market because the growth rate in the short term is lower than expected, but it does not mean that Nvidia will face problems.
Nvidia's real problem may come in the future,
when its peers and downstream companies are trying to break its monopoly and the commercialization of AI enters a period of slowdown.
How Nvidia can maintain its growth momentum will be a real challenge.
01 Growth, but also slowdown
Nvidia delivered an expected financial report.
Revenue continued to grow. This quarter, NVIDIA achieved revenue of
US$30.04 billion, a year-on-year growth rate of 122.4%,
achieving revenue higher than market expectations.
From the perspective of each major business, with the increase in AI demand, the data center business has become the main force contributing to Nvidia's revenue. This quarter, the
data center business
achieved revenue of US$26.3 billion, a year-on-year growth rate of 154%, accounting for 87.55% of the overall revenue. Among them, computing revenue was US$22.6 billion, a year-on-year increase of 162%, and network revenue was US$3.7 billion, mainly due to
InfiniBand and Ethernet revenue.
This quarter,
Nvidia's gaming business
revenue also continued to recover, achieving revenue of US$2.88 billion, a year-on-year increase of 15.8%. The recovery of the gaming business is closely related to the increase in downstream hardware sales. IDC data shows that global PC market shipments increased by 5.4% year-on-year in the second quarter of this year. Coupled with the increase in sales of GeForce RTX 40 series GPUs and SOC game consoles, Nvidia's gaming business has maintained a growth level.
However, from the perspective of overall growth rate, the growth rate of games is still slowing down significantly compared with the year-on-year growth rate of more than 80% in the fourth quarter of 2023.
Dolphin Investment Research
The automotive business also achieved revenue of US$346 million and a year-on-year growth of 36.8% this quarter, which was mainly due to
the AI Cockpit solution.
The automotive business accounts for a very small proportion of the overall revenue, with only less than 2%.
Naturally, Nvidia also places its main growth hopes for the next quarter on its outstanding data center business.
For the next quarter's revenue forecast, Nvidia gave a guidance of US$32.5 billion, which is also higher than market expectations.
Even though everything was growing and higher than market expectations, Nvidia's stock price did not rise after the release of its financial report. Instead, the stock price fell by nearly 7% after the financial report was released.
In fact, the reason is very simple: everyone has higher expectations for NVIDIA.
The 122.4% year-on-year growth rate is indeed far ahead, but compared with Nvidia itself, the growth rate is still slowing down. After all, Nvidia has achieved a year-on-year growth of more than 200% for three consecutive quarters. Moreover, compared with the slowing growth rate this quarter, the expected growth rate of 79.4% in the next quarter makes investors feel even more uneasy. You know, many of Nvidia's investors have experienced the decline of Cisco and the Internet bubble. They have suffered from PTSD, so it is normal for some people to leave the market when seeing the slowdown in growth.
In addition to the slowdown in growth, Nvidia's
gross profit margin performance
was also lower than expected.
This quarter, Nvidia's gross profit margin of 75.1% was lower than the market expectation of 75.4% and lower than Nvidia's own performance in previous quarters.
In response, Nvidia explained that it was due to "the increase in the proportion of new products in the data center business." As we all know, this new product is
Blackwell,
which Nvidia said it would mass-produce and deliver in the fourth quarter.
It is normal for costs to rise before new products are mass-produced.
Investors have given negative expectations for Nvidia's slowing growth and falling gross profit margin. However, this matter must be viewed dialectically.
Judging from the performance of this quarter's financial report, NVIDIA still has a very large competitive advantage. This quarter, NVIDIA's inventory turnover was 1.19 times, which was an increase from 1.01 times in the previous quarter, indicating that the product is still in short supply. In comparison, the inventory turnover of old rival AMD was only 0.62 times. NVIDIA's accounts receivable turnover this quarter was 2.26 times, and it was able to collect the principal more than twice in a quarter of 90 days. This means that it has considerable voice in its upstream and downstream. In comparison, AMD only had 0.62 times.
Based on its very strong upstream and downstream advantages, Nvidia's cash flow performance is also quite strong. This quarter, Nvidia's operating cash inflow was US$14.489 billion, while investment and financing cash flows were both outflows. This means that Nvidia is also very confident about the future and is using the money it earns to expand, repay loans or distribute dividends. In short, it is still a "money printing machine."
In addition, Nvidia's cost control
performed well
despite its gross profit margin falling short of expectations
.
Even though marketing expenses increased by 35%, in terms of percentage, Nvidia's marketing expenses this quarter were US$842 million, accounting for only 2% of total revenue. Even under the premise of promoting Blackwell, marketing expenses still accounted for a very small proportion, which shows that Nvidia's reputation in the industry basically does not need to be maintained.
Dolphin Investment Research
R&D expenses increased by 51.47% year-on-year to US$3.09 billion, basically accounting for 10% of total revenue, which is also a relatively healthy level.
With good cost control, Nvidia achieved a net profit of US$16.6 billion this quarter, a year-on-year increase of 168%, also a significant increase.
Overall, Nvidia delivered a financial report that met expectations, but the reason behind investors' withdrawal is that past growth will not continue, and the ultra-high gross profit margin is also unsustainable. With the revenue growth slowing down and the gross profit margin decreasing, especially in the third quarter when Blackwell has not yet achieved mass production and delivery, investors do not see a big rise in the short term.
02 Still far from a “bubble”
Ever since Nvidia took advantage of the AI boom, the discussion about whether Nvidia is a bubble has never stopped.
The investors' quick "quit" this time also shows how much damage the Internet bubble has caused them - however, Nvidia is not Cisco, at least not yet.
Looking back at the Cisco bubble, it is not difficult to find that the Internet wave began in 1994 when the US government opened up many restrictions on Internet commerce. Online advertising, e-commerce, and various commercial activities have ushered in the Internet era. When Cisco's stock price plummeted in 2000, the entire Internet industry had been developing for 6 years and had reached its growth ceiling.
Looking back at the AI industry, it has been only two years since GPT became popular, and the development of the entire AI large model demand driven by GPT is also in the early stages of the industry. Nvidia's explosive growth lasted less than a year.
Moreover, the commercialization of the AI industry has not even begun. The attitudes of major Internet companies in China and the United States towards large models are "embracing", "laying out" and "exploring". None of them has figured out the path to AI commercialization. As a company that sells shovels to gold diggers, Nvidia has the most say in the early stages of the industry's commercialization.
The development of the industry can also be seen from the capital expenditure budget
of Nvidia's downstream customers
. We can look at the capital expenditure of Internet giants. The combined capital expenditure of the four giants, Meta, Microsoft, Amazon, and Google, reached US$58.2 billion in this quarter, a year-on-year increase of 70%.
Amazon Cloud official website
Among them, Amazon stated that it had spent US$35 billion on capital expenditures such as data centers for its network service cloud department in the first half of this year, and it expects capital expenditures in the second half of the year to be higher than in the first half; META also raised the median capital expenditure for 2024, and expects capital expenditures to be between US$37 billion and US$40 billion.
The ever-increasing spending budgets of downstream companies on cloud business
are precisely Nvidia's most promising source of revenue in the future.
Industry growth does not mean that all industry revenue will go into Nvidia's pocket, so in addition to looking at the industry, we also have to look at the competition.
When the industry stops growing rapidly, fierce internal competition will begin. At this time, even the industry leader who had previously been firmly established will become dangerous due to the emergence of competitors. This is the case with Cisco. Around 2000, Cisco was indeed the industry leader, but the increasingly fierce competition has always challenged Cisco. For example, Juniper Networks launched the enhanced router M-160 in 2000, which directly challenged Cisco's previous almost monopolistic position in the router market. That year, Huawei also released the high-end router NetEngine. Two or three years later, Cisco's leading market share was gradually swallowed up.
However, judging from the current AI chip industry, Nvidia still has no rivals.
AMD, the industry's most direct competitor, has also grown in the past two years, but there is still a large gap between its revenue scale and Nvidia. In the second quarter of 2024, AMD's revenue was US$5.835 billion, which was about 20% of Nvidia's revenue in the same period.
In addition, we have compared data such as inventory turnover rate and revenue turnover rate above, and we can also see that when NVIDIA products are in short supply, AMD's inventory for one quarter actually takes more than one quarter to sell out, and it does not have that much say in downstream buyers.
In terms of products, AMD's high-end MI300 chip is mainly benchmarked against Nvidia's H100, but Nvidia has already reached the next level. The Blackwell chip GB200, which is expected to be mass-produced and delivered in the fourth quarter, includes 2 B200 GPUs + 1 ARM Grace CPU. It has a 30-fold performance improvement over the same number of H100s, and a 25-fold reduction in cost and energy consumption. In terms of performance, no rival can surpass it for the time being.
Nvidia is so powerful thanks to a system called the "walled garden" created by Huang Renxun.
Nvidia's "walled garden" system is similar to Apple's self-developed system, except that Apple is ToC and Nvidia is ToB.
The key behind Nvidia's "walled garden" system is a software platform called
CUDA
, which was launched in 2007. The main problem this platform initially solved was using Nvidia GPUs to run non-graphics software, such as some cryptocurrency mining algorithms and the recently popular AI software.
Later, NVIDIA launched a special code library to respond to the needs of software developers. What is a code library? It is software that has been written by predecessors and has been tested in practice and is very reliable. You can use it directly through CUDA.
By June this year, Nvidia stated at the shareholders' meeting that CUDA now includes more than 300 code libraries and 600 AI models, supporting 3,700 GPU-accelerated applications, and more than 5 million developers from approximately 40,000 companies are using these applications. (Financial Times)
In other words, NVIDIA has built a very strong ecological barrier. Any outsider who challenges it will have to face one problem: how to transfer the code system that these engineers have built on CUDA over the past nearly 15 years to their own system.
Whether from the perspective of market demand or competition, Nvidia is still far from a "bubble" at present.
03 Is it really impossible to surpass it?
People are afraid of becoming famous, and companies are afraid of becoming strong.
The AI pie is too big, and no one can be willing to let Nvidia take it all alone.
Competition among peers is inevitable.
Direct competitor AMD also stated in a recent earnings conference that it will launch MI325X in the fourth quarter of this year, MI350 in 2025, and MI400 in 2026. It also said that MI350 should be very competitive compared with Nvidia's Blackwell; in addition, Intel's Gaudi3 AI accelerator processor and Lunar Lake processor and other products have also performed well in the market.
In addition to
AMD and Intel
, two well-known rivals,
Nvidia's downstream buyers will also become its competitors.
Google, Amazon, and Microsoft are all developing their own chips to deploy and train AI. META and Microsoft, two of Nvidia's major customers, will also purchase AMD's AI chips...
Developing chips on your own is time-consuming and labor-intensive. Why would downstream companies prefer to invest time and money to do this?
It's very simple. Compared with its competitors, downstream customers do not want Nvidia to achieve a monopoly in the industry, because this means that the downstream buyers' voice will become lower and lower. Whether it is in terms of procurement costs or credit time, the downstream will have no voice. Therefore, downstream players are more willing to "watch the fun and not mind the trouble" to purchase AI chips from Nvidia's competitors to encourage competition in disguise, or simply develop their own.
Nvidia’s rivals are not limited to this. On August 15, AI startup Groq completed a new round of financing with a financing amount of US$640 million and a valuation of US$2.8 billion.
Groq official website
What Groq does is build fast AI reasoning technology.
Groq® LPU™ AI inference technology is a hardware and software platform that delivers exceptional AI computing speed, quality, and energy efficiency. (Leidi)
Its investors are even more star-studded, including PE giant BlackRock, industrial fund Samsung, and Cisco.
These industries and VC/PE investors flock to innovative companies on the track, and the logic behind this is that they see through Nvidia's customers and want to break its monopoly.
It is true that people who sell shovels always make more money than people who dig for gold. However, when gold diggers find out that they are "working" for people who sell shovels, they will rebel.
In addition to the current competition, in the longer term,
the commercialization progress of AI
will also affect Nvidia's market position. After all, Internet giants are obsessed with AI and have invested heavily in it. However, if they find that AI cannot significantly increase their revenue in a few years, and Nvidia, as an upstream company, has extremely high pricing power, then it is not impossible for them to reduce their purchases from Nvidia.
These crises do exist, and Nvidia has to think about them all the time, but they are not factors that will affect Nvidia in the short term.
In the short term, the key for NVIDIA is whether it can ensure the timely mass production of Blackwell. In this quarter's financial report, NVIDIA said that it expects Blackwell to bring $1 billion in revenue to the company in the fourth quarter. In addition, NVIDIA said that the market demand for GPUs based on the Hopper framework is also very strong. They said that the GPU capacity available to cloud service providers around the world is basically zero. They do not want to build more general computing infrastructure, but want to accelerate computing. Hopper is still the best choice for enterprises.
Looking at the overall situation, it is a fact that Nvidia is leading, but no matter how great a company is, it is impossible to maintain a very high growth rate all the time. When business inertia slows down and
competitors are besieging it from all sides, can Nvidia still maintain its lead?
This question will have to be answered by time.
Note: The cover image is from the Douban movie "Aquaman" stills
-END-
京公网安备 11010802033920号