【NVIDIA】Get to know Nvidia in one breath, why is he the new king of chips?

foreword

video link

• On May 24, Nvidia released its first quarter financial report
  • The stock jumped 30% on the day
  • Market value breaks trillions of dollars
• Nvidia shines in the data center space
  • Become a monopoly in the global AI training market
  • Market share up to 95%
• Most AI models are trained using Nvidia graphics cards
• Nvidia is an active participant in cloud computing, encryption, metaverse artificial intelligence and other fields
• Nvidia graphics cards become the standard to measure the company's computing power
• Founder Huang Renxun is known as the godfather of AI
• Huang Renxun's growth experience
  • In 1963, Huang Renxun was born in Tainan, Taiwan
  • Went to the United States to study at the age of nine
  • After graduation, engaged in chip design work in two semiconductor companies, including AMD
  • In 1993, NVIDIA was established with two other technicians, dedicated to the research and development of graphics computing chips
  • Reached long-term strategic cooperation with TSMC and Microsoft
  • Successfully launched a series of excellent graphics card products. Among them, GTX 256 is considered to be the world's first GPU, creating a new chapter in consumer graphics cards

graphics card market

• Microsoft develops Xbox, Nvidia wins $200 million order
• Nvidia Takes on Sony PlayStation
• Nvidia's revenue grows rapidly
  • doubling every year
• Nvidia's continuous annexation of rivals
  • swallowed 3Dfx
• ATI was acquired by AMD
• Two Big Fish Left in the Market: Nvidia and AMD
  • Nvidia cannibalizes AMD's market share
    • 2010: 60%
    • 2022: 80%
  • Nvidia Becomes the Global Overlord of GPUs
• Intel has a 71% share
• Nvidia has a 17% market share
• AMD has a 12% market share

Intel vs Nvidia Market Share Comparison

• Intel has a high market share and mainly sells integrated graphics.
  • The integrated graphics card and CPU are packaged together and share memory.
  • Intel occupies the integrated graphics market share by virtue of its CPU monopoly.
  • Unlike Nvidia's discrete graphics market.
• Nvidia holds more than 80% of the discrete graphics card market.
  • Discrete graphics cards have parallel computing features.
  • GPUs are designed for parallel computing.
  • GPU has the ability to calculate with thousands of boxes.
  • NVIDIA graphics cards are highly programmable.
  • NVIDIA graphics cards support a variety of libraries, expanding the application field.

Challenging Nvidia's monopoly and hardware-software integration

• Other companies have tried similar software to challenge Nvidia's hardware monopoly.
• The author hopes that the hardware graphics card and software expansion will cooperate tacitly to form a strong moat.
• "Ecosystem" is a vocabulary refined by the business world, and companies such as Apple and Microsoft have strong ecosystems.

Application of graphics card in the field of artificial intelligence

• Early graphics cards had limited application scenarios and were only used in a few areas that required a lot of computing.
• The Bitcoin mining phenomenon has made graphics cards in short supply, bringing business opportunities for Nvidia.
• Artificial intelligence has been an area that has really benefited Nvidia in recent years.
  • The parallel computing characteristics of GPU are suitable for deep learning and machine learning.
  • It is better to use GPU to train large language models.

The context in which the transformation occurred

• In 2012, the computer competition "ImageNet" sparked a shift
• The top-ranked AlexNet team leads the second place with an error rate of 16.4%, using the neural network training model and Nvidia's graphics card

Neural Networks and the Critical Role of Nvidia Graphics

• The neural network theory has existed for a long time, but it lacks the computing power support for its implementation
• Using Nvidia Graphics Cards to Realize Neural Network Computing Causes a Sensation
• Huang Renxun also attaches great importance to GPU and fully puts it into application in the field of artificial intelligence
• Nvidia Graphics Provides Simple and Convenient Solution for Accelerated Computing

Consensus in the field of artificial intelligence

• Google, Amazon, Microsoft, Baidu, etc. cannot do without NVIDIA graphics cards for large model training
• Nvidia A100 graphics card becomes the standard for training large models
• Nvidia launched an upgraded version of the graphics card H100, with a performance improvement of 4~6 times that of the A100

Nvidia

Fields covered by core technologies

• chip
• supercomputer

Computing power leasing service

• Provide computing power rental service

AI cloud business

• Provide AI cloud services for single customers
• Launched Prisso software to improve the performance of inverse lithography technology

Hardware and Software Services

• Involving hardware, software services and other fields
• Become the biggest winner in the AI ​​wave

Chinese market

• U.S. sanctions limit sales of certain products to China
• Launched special graphics card for China, bypassing sanctions

business division

• game
• Data center (accounting for 56%, an increase of 18% in the first quarter)
• Automotive Professional Graphics Processing
• Omniverse

growth and share price

• The stock price has increased by more than 1000 times
• The sixth largest company in the world
• Growth potential has not yet been fully released

Characteristics of the chip industry

• Moore's Law
  • Points out that the number of transistors that can fit on an integrated circuit doubles every 18 months
  • Has an important impact on the development of the chip industry
• Huang's law
  • GPUs are advancing faster than Moore's Law predicts
  • Capable of more than tripling performance approximately every 2 years
  • Known in the chip industry as Huang's Law
• High cost investment and fast update iteration speed
  • Requires huge initial cost investment and a lot of manpower and equipment
  • The update iteration speed is very fast, and it needs to run continuously
• Building a moat is difficult
  • The industry updates fast, and it is difficult to rely on a product or technology to build a moat that lasts for more than 2 years
  • Patent protection is limited, and the competitive advantage can only be maintained through continuous R&D and rapid iteration
• Talent reserves and R&D capabilities are moats
  • R&D capabilities, talent support, organizational structure, etc. are the key to forming a company's moat
  • Chip companies need to run the fastest, have the most abundant talent reserves, and have strong financial resources to profit in the fierce competition
• acquisition common
  • Intense competition often occurs in the chip industry, and many competitors choose to end the competition through acquisitions after competition
  • Chip research and development is difficult, the economy is unsustainable, and human capital is sufficient
• The Importance of the AI ​​Field
  • The field of AI is one of the important battlefields of the future
  • Graphics cards play an important role in this field
  • Internet giants began to actively carry out self-developed chip research and development to reduce dependence on companies such as Nvidia
• Nvidia's advantage
  • Nvidia has achieved a monopoly in the graphics card industry
  • Foresight and Endurance
  • Whether it can break through the siege again in the new round of melee is still uncertain

In general, the chip industry is faced with the characteristics of high cost investment, rapid iteration, and fierce competition. R&D capabilities and talent support are the key to forming a company's moat. At the same time, with the development of the AI ​​field, the chip industry is facing new opportunities and challenges.