Nvidia: AI computing is accelerating

NVIDIA held its 2018 fiscal year investor conference, where the company explained its business and strategic progress.

It all started with a vision for the computer graphics industry

As in previous years, Nvidia's investor conference was first attended by its flamboyant boss Huang, who mainly shared two things he thought were the most amazing: his foresight that computer graphics computing was the future of computer computing more than 20 years ago, and his expansion of GPU programmability 15 years ago. It was by doing these two things and persisting in them that Nvidia developed into a giant in artificial intelligence computing.

Based on NVIDIA GPU, NVIDIA's computing business covers games, professional image processing, data centers and autonomous driving. Next, NVIDIA will also enter other industry fields, such as medical and industrial fields.

Data Center: Continued triple-digit growth in fiscal 2018

In fiscal 2018, Nvidia's data center business revenue was nearly $2 billion, a year-on-year increase of 133%. The latest Volta architecture has been adopted by major OEM and ODM server manufacturers and public cloud service providers. The number of customers of Nvidia's inference platform has exceeded 2,000, and its accelerated computing architecture and applications are also widely used by supercomputing platforms.

All this has just begun. Nvidia believes that the potential target market size of the data center business will be as high as US$50 billion, of which the three major target markets of high-performance computing, large-scale/consumer Internet, and cloud computing/industry are US$10 billion, US$20 billion, and US$20 billion respectively.

In the fields of high-performance computing, scientific and technical computing, modeling and simulation, more and more companies and institutions are applying AI computing capabilities to their work. In the large-scale consumer Internet, AI is not only used by large enterprises, but also by more and more small and medium-sized enterprises. In the cloud computing and industrial fields that mainly use high-performance computers, it also provides NVIDIA with huge growth opportunities.

NVIDIA's data center strategy includes developers, NVIDIA platforms, cloud service companies and computer companies.

Developers are an important part of this. NVIDIA provides a platform for developers, and the applications developed by developers can be used by users and engineers of public clouds, private data centers, OEMs, and ODM platforms.

NVIDIA's underlying platform is the CUDA GPU computing architecture, on top of which are a wide variety of data center servers, inference and edge servers, and above that are various software, industry solutions and applications.

NVIDIA's developer ecosystem is already very large. Last year, in the field of high-performance computing alone, NVIDIA had 350,000 developers and more than 500 applications supporting GPUs.

In the field of large-scale consumer Internet applications, the cuDNN (CUDA Deep Neural Network) used for training has been downloaded more than 840,000 times, while the TensorRT neural network inference engine used for reasoning has been downloaded more than 30,000 times.

In the cloud computing and industry fields, 2,800 companies have joined Nvidia's artificial intelligence startup program, and the number of enterprise DGX customers has reached 550.

Deploying NVIDIA's AI computing accelerators in data centers means accelerating applications and reducing the demand for CPU servers, which can significantly save costs. According to NVIDIA's data, after using accelerated computing, in terms of high-performance computing, AI training and AI reasoning, the overall cost can be saved by 79-86% compared to using only CPUs.

High-performance computing is an important part of NVIDIA's data center business. NVIDIA believes that this target market is worth more than $10 billion. NVIDIA believes that every supercomputer needs to be accelerated, and the computing power of every supercomputer in the future will reach the exascale level. It is expected that the world's first exascale computer will be born in 2021-2022.

In the field of large-scale consumer Internet applications, NVIDIA believes that the market size will be $20 billion, of which training and inference may each account for half.

In the field of public cloud and industry applications, the potential market size will reach 20 billion US dollars. Currently, medical care, transportation and industrial manufacturing are the three largest fields, which will affect all industries in the future. The reason why NVIDIA puts public cloud and industry applications together is that industry applications are often built on public clouds or proprietary data centers provided by public cloud service providers.

Take the medical industry as an example. The market size of this industry is as high as 100 billion US dollars, of which 25 billion US dollars are medical devices and 5 billion US dollars are image processing software. It can be seen that the application of deep learning in the medical field is increasing significantly, such as academic papers related to artificial intelligence medical care and the financing of start-up artificial intelligence medical companies.

The same is true in the field of autonomous driving, where data centers have broad application prospects.

In short, for NVIDIA, the AI-driven data center business is currently its most important growth point, and its growth rate is ahead of other business segments. With a potential target market size of 50 billion, compared with the current revenue of only 2 billion US dollars, the market imagination of the data center is still not small.

Self-driving cars: It’s just the beginning

Nvidia's business in the automotive field is shifting from car infotainment systems to autonomous driving. Nvidia believes that every car will have autonomous driving capabilities, which is a revolution in the $10 trillion market. Autonomous driving can reduce traffic accident casualties, and also help improve transportation efficiency and reduce logistics costs.

For Nvidia, this is a potential target market with a size of $60 billion.

Nvidia's current revenue in the automotive sector is 558 million yuan, with nearly 400 partners. The potential target market size for self-driving cars is 100 times its current revenue.

However, automakers face many challenges in participating in autonomous driving, such as massive data processing, energy consumption, massive driving tests and verification, etc.

NVIDIA's solution is based on artificial intelligence, unified architecture, end-to-end solution, and open platform.

For example, the existing ADAS assisted driving is mainly composed of four decentralized computing units.

Nvidia's solution only requires one computer, which will bring lower costs and greater computing power.

In order to reduce the cost and time of field testing required for autonomous driving training, NVIDIA released the autonomous driving simulator Drive Constellation.

This is a simulator that can use VR technology to simulate and test the autonomous driving system. It can rely on the powerful performance of NVIDIA GPU products to create a completely virtual world and simulate and test the algorithm of the autonomous driving system 24 hours a day.

"If 10,000 Drive Constellation systems are deployed, it will only take one year to reach a road test level of nearly 3 billion miles. Drive Constellation is a tool of great significance to the entire autonomous driving industry." Huang Renxun said that Drive Constellation will be officially launched on the market in the third quarter of this year, and by then many of NVIDIA's partners will be able to enjoy its "acceleration" in autonomous driving research and development.

In summary, NVIDIA believes that every car in the future will be self-driving, which means there will be opportunities in the market size of up to 60 billion. NVIDIA's key strategies in this field are AI, Xavier, end-to-end solutions, and open platforms. Currently, NVIDIA has launched the test and verification platform Drive Constellation for autonomous driving. NVIDIA has more than 370 partners in the field of autonomous driving.

Gaming business: E-sports and social networking drive growth in the market

In fiscal year 2018, the revenue of the gaming business continued to grow, increasing by 36% year-on-year to $5.5 billion. In 2017 (fiscal year 2018), NVIDIA updated the gaming graphics card product line Pascal: 1080 Ti, 1070 Ti, TITAN Xp; more and more manufacturers adopted the Max-Q thin and light gaming notebook design; gaming notebooks sold well; Nintendo's Switch, which uses NVIDIA graphics cards, was a great success; the number of GFE customers exceeded 100 million; the GFN beta version was released; and cryptocurrency brought demand for mining graphics cards.

The gaming market is huge and continues to grow. Today, almost everyone is a gamer. Games are related to social interaction and competition, and are replacing traditional media and entertainment.

In 10 years, the revenue of the game software business has reached 100 billion US dollars, the number of game players worldwide has exceeded 2 billion, and game users cover all age groups.

Although users are increasingly obsessed with smartphone games, the PC gaming market is still viable. Nvidia believes that as an open gaming platform, it can cover mid-range, high-end and low-end games and is still the hottest place for innovation in gaming content. (At present, Nvidia's graphics card chips have a very low market share in the smartphone market)

Nvidia's GeForce platform is the world's largest gaming platform, of which Pascal graphics cards account for 70%.

E-sports games continue to drive the growth of the gaming industry, with an audience of more than 100 million people and prize money exceeding $100 million. Overwatch and League of Legends are the leaders in the field of PC e-sports games.

The social appeal of games has expanded their influence. The number of users watching live games on Twitch continues to grow, and word of mouth spreads among users. NVIDIA also provides users with the function of recording and sharing game clips in graphics cards.

The GeForce gaming experience center platform has 100 million customers worldwide and has 800 million game videos.

Users' demand for cinematic gaming quality will continue to drive graphics cards forward.

In short, NVIDIA believes that the gaming industry is huge and still growing, GeForce is the world's number one gaming platform, e-sports and social networking drive gaming growth, and movie-level games require more GPU computing, so the future of the gaming industry remains exciting.

Financial data: Revenue increased by 41%, operating profit increased by 63%

In fiscal 2018, Nvidia's revenue increased by 41% year-on-year to US$9.7 billion, gross profit margin increased by 100 basis points to 60.2%, operating profit increased by 63% year-on-year to US$3.6 billion, and earnings per share increased from US$3.06 last year to US$4.92 this year.

By business, the gaming business grew 36% year-on-year to US$5.5 billion, the professional image processing business grew 12% year-on-year to US$34 million, the data center business grew 133% year-on-year to US$1.932 billion, the fastest growing of the four major business segments, and the automotive business grew 15% year-on-year to US$558 million.

The company has formed a relatively diversified business structure and maintained rapid growth. The number of GFE users has exceeded 100 million, emerging professional image applications account for more than 30%, the revenue of the seven major players in the data center business increased by 155%, and the number of autonomous driving partners exceeded 320.

Despite the impact of the expiration of Intel's patent license, Nvidia's gross profit margin still increased by 1 percentage point to 60.2%.

The gross profit margins and gross profit contributions of different businesses vary. For example, the gaming business is currently the largest source of gross profit, with a gross profit margin in the middle, while the data center has a better gross profit margin, but its gross profit contribution is lower than that of the gaming business. The gross profit margins and gross profits of the OEM and automotive businesses are still low.

Operating expenses increased by 19% year-on-year to US$2.2 billion, but the proportion of operating expenses in revenue has dropped to 23%. Operating expenses are mainly invested in games, artificial intelligence and autonomous driving.

As of this fiscal year, the company's cumulative investment in research and development has reached 15 billion US dollars.

Currently, software business revenue has increased compared to the past in the company's business, and software accounts for approximately 40%.

Thanks to the reuse of the company's GPU computing platform in different businesses, the operating profit margin increased to 37%.

Operating cash flow increased 109% year-over-year and net cash increased 28%.

Since 2013, the company has returned a total of $5 billion to shareholders through repurchases, dividends, etc.

The company's cash use mainly includes four forms: operating expenses, capital expenditures, investment and mergers and acquisitions, repurchases and dividends.

via: 199IT Financial Report Data Center