Decrypting NVIDIA: 6 key points to help you understand, the secret of the AI king’s stock price rising 240% (Part 2)

Table of contents

Review of the previous article

NVIDIA founding background:
1990s: Focus on the computer game industry
2000s: In addition to games, we also need to make Mars simulations and automotive chips
2010s: GPUs are perfect for training artificial intelligence!
2020s: 30 years of GPU hard work, harvesting fruitful results in artificial intelligence
NVIDIA’s super business model:
At the NVIDIA Investor Relations Conference in April 2016, Jensen Huang mentioned,
NVIDIA's business model is mainly driven by two core elements:
"Platform and Ecosystem" and "Leverage and Scale Effect".

NVIDIA’s financial backer dad:
Customers such asAWS (Amazon Web Service), Meta, Microsoft, Google, etc.Just bring it to NVIDIA 40% revenue!

Click here to review the previous article👉Decrypting NVIDIA: 6 key points to help you understand the secret of the AI king’s stock price soaring 240% (Part 1)

Preface

Last week’s article mentioned NVIDIA’s founding background, business model, and major customers. Today’s article continues to share NVIDIA’s competitors, potential risks, and future risks.At the same time, the three major future trends mentioned by CEO Jensen Huang in the 6/2 Keynote Speech were also highlighted., let’s watch it together!

3 Takeaways to Take Away If You Only Have 1 Minute

1. NVIDIA’s competitors

NVIDIA's main competitors are AMD and Intel. according toData for 2023 Q4, NVIDIA's market share in the GPU market reached 80%, occupying a leading position in the GPU market in the fields of AI and gaming.
But AMD’sMore affordable small chip designIt is also gradually increasing its market share, making it a long-term threat to NVIDIA.

2. What potential risks does NVIDIA face?

The risks NVIDIA faces are not primarily in the product itself;Rather, the market environment fluctuates.
For example, changes in the cryptocurrency market and falling virtual currency prices will also reduce mining demand, thereby affecting GPU sales.
at the same time,Sino-US trade warThe regional sales restrictions brought about may also affect NVIDIA's performance in the Chinese market.
Of course, NVIDIA's biggest customers were mentioned last week: technology giants such as Microsoft, AWS, Meta and Google , they alsoCurrently developing its own AI chip, the purpose is to reduce dependence on NVIDIA GPU.
In the future, NVIDIA will also need to face greater price pressure, or come up with solutions for chip customization that can make technology giants pay for it.

3. How will NVIDIA develop in the future?

6/2 Jensen Huang’s Keynotes mainly mentioned three major directions in the future:

Digital twin:
NVIDIA plans to use digital twin technology toSimulate the earth environment, creating models that simulate the earth's environmental conditions can help predict and reduce the impact of diseases and climate change, especially in Taiwan with a variable climate, which is of great help to the fields of environmental protection and public health.

Physical reality AI robot:
NVIDIA will integrate AI and the physical world and train AI to run in simulated physical environments through PhysicalAI, making it smarter in the real world.
For example: Hon Hai can simulate the computer room environment to train AI robots, and simulate various situations in real time, allowing the robots to operate in an environment consistent with the real physical world.Simulate possible operational errors in advance, improve the accuracy and efficiency of operations.

GPU beyond Moore’s Law:
NVIDIA CEO Jensen Huang said that GPU performance and energy efficiency have significantly improved, and costs have dropped, making large language models possible. He pointed out that using Blackwell to train a model with 2 trillion parameters like GPT-4,Requires 1/350th the power of 2016 Pascal GPUs.
Huang Renxun emphasized,The computing power of NVIDIA's GPU server products has increased by 1,000 times in the past 8 years. In comparison, Moore's Law can only increase by 40 to 60 times during the same period.
These technological breakthroughs allow AI models to run at lower costs and with higher efficiency, thereby creating more value for various industries.

NVIDIAcompetitors

Don’t NVIDIA’s GPUs have any competition?
Is there no GPU from other companies that I can buy?

1. The legendary NVIDIA replacement?
NVIDIA’s GPU competitors

NVIDIA's main competitors are AMD (Advanced Microelectronics) and Intel.
according to Data for 2023 Q4, GPU market share: NVIDIA 80%, AMD 19% and Intel 1%

2. Competitor comparison: AMD Super Micro

Next, let’s make a simple comparison between NVIDIA and AMD:

	NVIDIA (NVDA)	AMD (AMD)
market status	GPU market leader, especially in AI and gaming	The second largest player in the GPU market
Product strengths	GeForce series GPU (game, AI, Data Center)	Radeon series GPU, Ryzen series CPU
ecosystem	Mature product ecosystem and powerful CUDA software moat	Small die design with high CP value
market share	The market share of GPU AI applications is as high as 95% (NVIDIA currently has no competitors in GPU for AI applications)	The entire GPU market share is approximately 19%
2024 Q1 gross profit margin	64.6 %	47%
Pricing Strategy	Targeting high-end consumers	Targeted at consumers with limited budget but who want a good graphics card
Advantage	Strong hardware performance and CUDA software advantages form a strong ecosystem	Small chip design can drive down costs
Disadvantages	New to the CPU market, the product is not yet mature and the market share is low.	Small scale and relatively weak R&D capabilities
2023-Q2 GPU market share	81%	19%

NVIDIA potential risks

Although NVIDIA's market share, revenue, stock price, and products are all great, there are still potential risks.

A balanced report on NVIDIA's potential risks: What challenges will NVIDIA face in the future?

2 big risks for NVIDIA

1. Changes in market environment

Cryptocurrency Market Volatility:
When cryptocurrency prices fall, mining demand decreases and GPU sales also decline.
Sino-US trade war, US sales restrictions on China:

The United States' restrictions on high-end technology products sold to China will also affect NVIDIA's sales, because China is also one of NVIDIA's important markets.

2. The threat of customized AI chips

Tech giants begin developing their own AI chips:
Companies such as Microsoft, AWS, Meta, and Google are certainly not fuel-efficient, and spending money all the time on NVIDIA’s AI GPUs is no solution either!
They have begun to develop their own customized AI chips, because internal development costs are lower and they can reduce dependence on NVIDIA GPUs. This trend puts NVIDIA's already high-priced GPUs under greater price pressure.

In fact, the above risks do not come from NVIDIA's products themselves, but more from how to respond to changes in the external market environment and consumer demands in real time, so that NVIDIA can continue to gain a foothold in the GPU market.

NVIDIA future development

2024 Q1: 2 platforms, 1 new technology, 1 software

After reading this, if you think this article is good,
Welcome to subscribe to my newsletter [Roxanne's Tech Talk]
I will share more interesting science and technology stories on it! 🥳
Join 500 people and learn the latest technology and knowledge together Subscribe Roxanne's Tech Talk

According to NVIDIA’s 2024 Q1 financial report just announced on May 22, CEO Huang Jenxun also revealed at the meeting NVIDIA’s next actions. In simple terms, it can be divided into: 2 platforms, 1 new technology, and 1 software.
These four new trends are all closely related to AI and generative AI (GenAI).

1. 2 platforms: Hopper, Blackwell

Hopper platform:
The Hopper platform is specially used for AI training to improve the application efficiency of AI models in various industries. The Hopper platform provides AI with powerful computing capabilities, helping AI process instructions faster and more accurately.
Blackwell platform:
The Blackwell platform is a new platform that supports large-scale generative AI and is designed for the training and operation of ultra-large-scale AI models.Such as ChatGPT, which helps these models perform better when handling complex tasks.

2. New data transmission technology: Spectrum-X

Spectrum-X is a new data transmission technology that effectively improves the data transmission speed in Ethernet data centers.Allow large-scale AI models to run more smoothly in the data center.

3. Generative AI software: NVIDIA NIM

NVIDIA NIM is enterprise-level generative AI software that can run on the cloud, local data centers, and RTX AI PCs (personal computers equipped with NVIDIA RTX series GPUs) to help enterprises conduct accurate data analysis and make decisions.

Huang Renxun said that the next industrial revolution has arrived, and AI will bring explosive growth to productivity in all walks of life.He estimates that there are currently about 15,000 to 20,000 generative AI start-ups in the market, all looking forward to training on NVIDIA chips.

While the market is expanding rapidly, NVIDIA has also become a solution provider.

For example, NVIDIA provides a complete AI training solution, from the data center technology Spectrum-X required for training, to the training platform Hopper, to the operation platform Blackwell, and sells Total Solutions to enterprises, NVIDIA is no longer the NVIDIA that only sold GPUs in the past!

Also featured in 6/2 Keynotes: The three major directions mentioned by Jen-Hsun Huang

1. Saving Taiwan from climate disasters – Digital twin:
To put it simply, a digital twin uses AI artificial intelligence technology to create aEarth Twin.
On this simulated Earth, NVIDIA expects to predictFuture overall environmental changes, disease reduction and climate change impacts.
Huang Renxun specifically mentioned the application scenario in Taiwan, which can help predict the trend of typhoons, prevent possible natural disasters and take precautions in advance. It is really practical for Taiwan with a changeable climate.

2. AI robots that can adapt to the physical world:
NVIDIA will integrate AI and the physical world, and use PhysicalAI to train AI to run in simulated physical environments, so that these robots can better adapt to the physical rules of the real world and achieve more precise operations.
For example, companies such as Hon Hai have begun toSimulate computer room environment to train AI robots, conduct training on movements of hardware assembly and integration.
By simulating various situations in real time, you can firstAvoid losses caused by improper robot operation, and at the same time, it also allows real-life robots to "learn and imitate" the actions and behaviors of these simulated world robots, realizing what Huang Renxun said."Let robots learn how to be robots."

3. GPU computing power growth will exceed Moore’s Law:
Huang Renxun said that the significant improvement in GPU performance and energy efficiency, while driving down costs, has made the training of large language models possible.
In his speech, he mentioned that using the Blackwell architecture (AI super chip launched by NVIDIA) to train a model with 2 trillion parameters like GPT4,The power required is only 1/350 of the 2016 Pascal GPU.
Huang Renxun emphasized,NVIDIA's GPU server products have increased computing power by 1,000 times in the past 8 years.This is much higher than the 40 to 60 times predicted by Moore's Law within 8 years. The chart on Keynotes (below) also confirms a paper published by NVIDIA last year. Huang's LawNo joke.

3 Takeaways from this article

1. NVIDIA’s current competitors in the market

NVIDIA's main competitors are AMD and Intel.
According to data from 2023 Q4,NVIDIA’s market share in the GPU market reaches 80%, occupies a leading position in the GPU market in the fields of AI and gaming, but AMD’sMore affordable small chip designIt is also gradually increasing its market share, making it a long-term threat to NVIDIA.

2. What potential risks does NVIDIA face?

The risks NVIDIA faces are not primarily in the product itself;Fluctuating market environment.
For example, movements in the cryptocurrency market,When virtual currency prices fall, mining demand will also decrease, which in turn affects GPU sales.
at the same time,Regional sales restrictions brought about by the US-China trade warIt may also affect NVIDIA's performance in the Chinese market.
Of course, NVIDIA's biggest customers were mentioned last week: technology giants such as Microsoft, AWS, Meta and Google, who are also working onDevelop your own AI chip, the purpose is to reduce dependence on NVIDIA GPU. In the future, NVIDIA will also need to face greater price pressure, or come up with solutions for chip customization that can make technology giants pay for it.

3. How will NVIDIA develop in the future?

6/2 Jensen Huang’s Keynotes mainly mentioned three major directions in the future:

Digital twin:
NVIDIA plans to use digital twin technology toSimulate the earth environment, creating models that simulate the earth's environmental conditions can help predict and reduce the impact of diseases and climate change, especially in Taiwan with a variable climate, which is of great help to the fields of environmental protection and public health.

Physical reality AI robot:
NVIDIA will integrate AI and the physical world and train AI to run in simulated physical environments through PhysicalAI, making it smarter in the real world.
For example: Hon Hai can simulate the computer room environment to train AI robots, and simulate various situations in real time, allowing the robots to operate in an environment consistent with the real physical world.Simulate possible operational errors in advance, improve the accuracy and efficiency of operations.

GPU beyond Moore’s Law:
NVIDIA CEO Jensen Huang said that GPU performance and energy efficiency have significantly improved, and costs have dropped, making large language models possible. He pointed out that using Blackwell to train a model with 2 trillion parameters like GPT-4,Requires 1/350th the power of 2016 Pascal GPUs.
Huang Renxun emphasized,The computing power of NVIDIA's GPU server products has increased by 1,000 times in the past 8 years. In comparison, Moore's Law can only increase by 40 to 60 times during the same period.
These technological breakthroughs allow AI models to run at lower costs and with higher efficiency, thereby creating more value for various industries.

Thank you for reading this article!
If you feel that you still have more to learn, please subscribe to my newsletter [Roxanne's Tech Talk]
I will share more interesting science and technology stories on it! 🥳

Join 500 people and learn the latest technology and knowledge together Subscribe Roxanne's Tech Talk
You are also welcome to communicate through Linkedin👩🏻‍💻 Roxanne Chen