NVIDIA: The "Lucky One" in the Semiconductor Industry?

Entrepreneurial Aspirations, Laying the Foundation: Founded in 1993, NVIDIA (pronounced "en-VID-ee-o") is currently at the age of 30, which is known as the "thirty-something" in Chinese culture. Interestingly, Huang Renxun also founded NVIDIA at the age of 30. The founders also included Chris Malachowsky and Curtis Priem, whose vision was to introduce 3D graphics into the gaming and multimedia markets. Since NVIDIA launched its first GPU, the GeForce 256, in 1999, it has laid the foundation for reshaping the computing industry. The GeForce 256 is hailed as "the world's first GPU," quickly dominating the gaming market, which has laid a solid foundation for NVIDIA's development in the semiconductor field.

CUDA Architecture, Opening a New Era: In 2006, NVIDIA introduced the CUDA architecture, opening a new era for the application of GPUs in scientific computing and data processing. CUDA provides a complete set of programming tools and interfaces, enabling researchers and developers to start using GPUs for high-performance computing tasks, such as deep learning, physical simulation, and bioinformatics.

Advertisement

AlexNet, Making GPUs Stand Out: In 2012, the AlexNet convolutional neural network (CNN), supported by NVIDIA GPUs, won the ImageNet challenge with an overwhelming advantage, surpassing the closest competitor by 9.8 percentage points. The success of AlexNet attracted widespread attention in the fields of artificial intelligence and computer vision. It proved the strong potential of deep learning, especially deep convolutional neural networks, in visual tasks, thus triggering a surge in research and application of deep learning. AlexNet's training used two NVIDIA GTX 580 GPUs, demonstrating the potential of GPUs in accelerating deep learning training. This achievement also accelerated the popularization of GPUs in deep learning research and application.

Mining Boom, An Unexpected Surprise: 2017 was also an unexpected year for NVIDIA. At that time, cryptocurrency was very popular, and mining was in vogue. Miner chips became the favorites of millions of investors. Due to the performance and energy efficiency of its chips, NVIDIA's gaming chips (especially the GeForce RTX 30 series) were very popular among cryptocurrency miners. The mining boom brought NVIDIA huge profits, but also some problems. For example, due to the scramble of miners, it became difficult for gamers to obtain high-performance graphics cards. In 2017, NVIDIA's OEM revenue increased by 200%, and part of the growth was mainly due to the increased demand brought by the cryptocurrency mining boom. This attracted the attention of the U.S. Securities and Exchange Commission (SEC). In May 2022, NVIDIA was fined $5.5 million for not fully disclosing the impact of cryptocurrency mining on its gaming business in its financial report for the fiscal year 2018.

It can be said that GPUs have swept through gaming, mining, and AI, becoming the hot spots in the market. Rather than saying that NVIDIA has seized wave after wave of opportunities, it is more accurate to say that the inherent high-speed computing capabilities of GPUs have made it possible.

Artificial Intelligence, Achieving the Pinnacle: In 2023, the large language model ChatGPT became an instant hit, and large models became the "hottest" in the market. NVIDIA's GPUs, once again, have become the preferred choice in the field of deep learning due to their high-speed processing capabilities, large memory, and architecture optimized for AI. NVIDIA's H100 chip has no rivals, once again pushing NVIDIA to the forefront of the era.For Nvidia, 2023 has undoubtedly been a record-breaking year. The surge in artificial intelligence has completely transformed the semiconductor industry's landscape in 2023: its total market value has exceeded one trillion dollars; it has surpassed TSMC, Samsung, and Intel to become the top semiconductor company; and it has set a record for data center GPU chip sales exceeding those of CPUs. Nvidia has been elevated as the new king in the semiconductor field.

Market Value Surpasses the Trillion-Dollar Threshold

On May 30, 2023, Nvidia set a historical high, with its total market value exceeding one trillion dollars for the first time, topping the list of global chip supplier market values and also becoming the world's seventh company with a market value over one trillion dollars. If you think Nvidia is a flash in the pan, then by the close of trading on December 26, 2023, Nvidia's market value had climbed to 1.22 trillion dollars. Meanwhile, Intel's total market value was 212.9 billion dollars, AMD's was 231.6 billion dollars, and TSMC's was 491.1 billion dollars.

Wall Street analysts' forecasts for Nvidia are "inscrutable," with its valuation reaching as high as 1.6 trillion dollars.

Nvidia's market value breaking through the trillion-dollar mark is mainly due to its leading position in the field of graphics processing units (GPUs). GPUs are an important component of computer graphics processing and are widely used in gaming, artificial intelligence, machine learning, and other fields. Nvidia's GPU products are highly competitive in these areas and have therefore been widely welcomed.

Nvidia's market value breaking through the trillion-dollar mark also marks the strong development of the global semiconductor industry. Semiconductors are the foundation of modern information technology and are widely used in emerging fields such as artificial intelligence, the Internet of Things, and autonomous driving. With the rapid development of these fields, the semiconductor industry will usher in new growth opportunities.

Revenue Soars to the Top of the Semiconductor Industry

The third quarter of 2023 was definitely a more historic moment for Nvidia. In that quarter, Nvidia's revenue reached an astonishing 18.12 billion dollars, a year-on-year increase of 206%, surpassing TSMC and Samsung, Intel, and rising to the top of semiconductor revenue.However, that's not all; Nvidia is likely to become the top semiconductor supplier in 2023. Based on the financial reports disclosed by various manufacturers for the first three quarters, Nvidia is estimated to be in the second place in terms of semiconductor manufacturers' revenue. If we only look at the sales revenue of semiconductor products (excluding foundries), Nvidia is likely to be the number one in 2023, surpassing Intel. Affected by the weak storage market, Samsung is estimated to still be in a loss state in the fourth quarter, so Samsung should be ranked fourth among these four companies. What's more eye-catching is that analysts predict Nvidia's revenue will reach 89.2 billion US dollars, about three times that of 2022, with an astonishing increase of 59%.

Looking at the top ten global semiconductor suppliers, if we consider the establishment time of the parent company or the acquired company, TSMC was established on February 21, 1987, and has been around for 36 years. Intel was founded in 1968, which is about 54 years old. Samsung Electronics and SK Hynix, two South Korean companies, have more than 40 years of experience in semiconductor sales, and AMD started producing logic ICs 54 years ago. Broadcom was established 32 years ago, and its parent company Avago entered the semiconductor field 52 years ago. Qualcomm was founded in July 1985 and has been around for 38 years. STMicroelectronics was established in 1987 by the merger of the Italian SGS Microelettronica and the French Thomson Semiconductor. The semiconductor business of SGS and Thomson can be traced back to the 1970s. Texas Instruments (TI) was founded in 1930 and entered the semiconductor business in 1954. Infineon Technologies was originally part of Siemens AG, which was founded in 1847, and Siemens started producing semiconductors in 1953. Infineon was spun off as an independent company in 1999.

Among them, Nvidia, founded in 1993, can be said to be the youngest company, and this year is exactly the "thirty years of age" of "standing firm". When Nvidia entered the industry, Intel had already been famous, and AMD had also become a well-known large semiconductor enterprise.

Big winner in data centers: GPU sales exceed CPU for the first time

For a long time, data centers have been dominated by CPUs, with Intel leading the entire server market with its Xeon central processing unit (CPU) series for a long time. It is the heart of computer servers, just like the CPU is also the brain chip of personal computers. However, this year, the market sales of GPUs in data centers have gradually begun to surpass CPUs.

This can be seen from a series of financial reports of several key chip giants in 2023. It is well known that chip giants such as Intel, Nvidia, and AMD are all competing around data centers.

In the first quarter of 2023, Nvidia's data center business revenue was 4.28 billion US dollars, at which time Nvidia had not yet exceeded the total of Intel and AMD in the data center. By the second quarter, Nvidia's revenue in the data center directly doubled, reaching 10.3 billion US dollars, close to twice the total of Intel and AMD. In the third quarter, Nvidia set a new record for data center revenue, reaching a peak of 14.5 billion US dollars, which is about to be three times as much.

It can be seen that the growth rate of GPU sales is far faster than that of core server CPU chips. This also means that accelerated computing chips represented by GPUs have far surpassed general computing chips like CPUs. According to Mercury Research, which has been tracking the GPU market since 2019, Nvidia's share in the data center GPU system market in the third quarter was as high as 99.7%, excluding any devices used for networking, and the remaining 0.3% was shared by Intel and AMD. It is important to know that AMD spent about four to five years to obtain a 20% share in the data center CPU market.As stated by Nvidia's founder and CEO, Huang Renxun: "A new era of computing has arrived. Enterprises around the world are transitioning from general-purpose computing to accelerated computing and generative AI. Startups with large language models, consumer internet companies, and global cloud service providers have taken the lead, and the next wave is poised to surge. The era of generative artificial intelligence is taking off." Nvidia's GPUs, CPUs, networking, AI foundry services, and AI enterprise software are all engines of rapid growth.

Recently, Intel CEO Pat Gelsinger hinted that if Intel had not terminated one of his favorite projects, Larrabee GPU, about 15 years ago, then Intel, not Nvidia, would be creating this technology in the field of artificial intelligence hardware today. "Thirteen years ago, when I was driven out of Intel, they killed this project that would change the shape of artificial intelligence," Gelsinger said when talking about Larrabee. Nvidia's success in providing GPUs for AI modeling is "very fortunate."

However, at a recent Chinese American Semiconductor Professional Association (CASPA) dinner, when asked, "Do you think the GPU architecture will still dominate in the next few years? Is there still a chance to beat Nvidia?" Huang Renxun jokingly replied: "You have a chance, but not much. In fact, the chances are slim."

It is foreseeable that as Intel and AMD have both launched new AI chips, it is believed that in the coming year 2024, there will be an epic battle between Intel, AMD, and Nvidia in the field of artificial intelligence data center processors. Ultimately, market adoption rates will determine who will win this battle.

In conclusion, "We are always only 30 days away from bankruptcy." It is this vigilance that has allowed Nvidia, a company that has gone through several ups and downs and has been on the verge of bankruptcy several times, to survive again and again, becoming a well-deserved semiconductor giant. To quote Huang Renxun's speech at this year's National Taiwan University graduation ceremony to encourage students: "Whether it's for food or to avoid being seen as food, just run."