Video Cards (Global Market)

Russian video card market

Main article: Video cards (Russian market)

2024: Nvidia's global video card market share hits record 88%

By the end of the second quarter of 2024, Nvidia's share of the global GPU-based accelerator (GPU) market reached a record 88%. Overall, Nvidia has dominated the graphics card segment for more than 20 years. This is stated in the materials of the 3DCenter, which TAdviser got acquainted with at the end of September 2024.

It is estimated that Nvidia has been an almost permanent leader in the global GPU accelerator market since 2002. AMD was only able to take a four-quarter lead in 2004-2005. However, since then AMD has failed to take the palm from its competitor. It is also noted that in the mining booms in 2017-2018 and in 2021-2022, AMD indicators began to grow, but in these cases the lag behind Nvidia was very significant.

Nvidia won thanks to the rapid development of artificial intelligence. Nvidia accelerators are used by operators of the largest data centers and hyperscalers. According to the analytical company Tech Insights, in 2023, Nvidia delivered 3.76 million GPU-based accelerators for data centers: this is about 1 million more compared to 2022. Taking into account AMD and Intel products, shipments of GPU accelerators for data centers in 2023 were at the level of 3.85 million units against about 2.67 million units in 2022. Thus, Nvidia captured 98% of the accelerator segment for data centers.

According to 3DCenter, a significant increase in sales of Nvidia GPU solutions was observed in 2015-2016, at the end of 2018, and also at the end of 2020. At the same time, peak values ​ ​ were recorded in the first and second quarters of 2024, which is due to the rapid expansion of the global data center infrastructure and the emergence of new AI models with a huge number of parameters, the training of which requires large computing power.^[1]

2023: Rise in demand for video cards due to AI boom

Artificial intelligence (AI) models require a lot of processing power to process a huge amount of data and execute complex algorithms. Traditionally, CPUs have been used to process data in computers. However, video cards (GPUs) have become the preferred choice for AI tasks for several reasons:

1. Parallel computing: Video cards have thousands of cores that allow parallel computing. This means they can handle multiple tasks at the same time, which is especially useful for AI models that work with large amounts of data.

2. High performance: Video cards are specially designed for graphics processing, making them very efficient in performing operations on matrices and vectors, which are the basis for many machine learning algorithms.

3. CUDA and OpenCL: CUDA (Compute Unified Device Architecture) and OpenCL (Open Computing Language) are software platforms that allow developers to use the power of a video card to process data. They provide an API and a set of tools for programming a video card and optimizing AI tasks.

4. Low cost: Video cards typically have a better price/performance ratio compared to CPUs. They are available in a variety of price categories, making them more accessible to AI researchers and developers.

5. Specialization in graphics tasks: video cards were developed primarily for graphics processing, so they have specific capabilities such as hardware texture acceleration, filters and shaders. These capabilities can be used to optimize the performance of AI tasks, especially in the field of computer vision.

In general, the use of video cards for AI models allows you to significantly speed up calculations and increase the efficiency of algorithms.

2020: 3m deliveries rise during cryptocurrency boom

The volume of world discrete video cards at the end of 2020 increased by about 3 million units compared to 2019 and amounted to 41.5 million units, which was partly facilitated by the demand for equipment for the extraction of cryptocurrencies. This calculation was shared by the research company Jon Peddie Research (JPR).

According to experts, in 2020 the video card market returned to growth after the disastrous 2019, when product shipments were record low - 38.56 million units.

Discrete Graphics Card Market Size Change Dynamics, Jon Peddie Research Data

The result registered in 2020 was lower than the previous years and is much inferior to the volume of sales in 1999, when about 114 million video adapters were sold. Revenue in the market under consideration in 2020 amounted to $14.8 billion, and the average cost of the sold video card was $360.

The report says that Nvidia has been leading the discrete GPU market since the early 2000s, and in 2020 the company's market share reached its maximum value in history - 83% in physical terms. At the same time, AMD's presence indicator, on the contrary, fell to a record low of 17%.

According to experts, in recent years, including 2020, Nvidia has been about four times ahead of AMD in sales of discrete graphics. In addition, Nvidia has long dominated the laptop standalone GPU segment thanks to gaming solutions. Historically, this market has been small, but it has grown sevenfold in seven years, according to Nvidia. JPR highlighted Nvidia's strong positions in the mobile video card market - in 2020, the company delivered about 50 million discrete GeForce GPUs for gaming laptops.^[2]

Now laptops have become the fastest growing gaming platform, "said Nvidia CFO Colette Kress at one of the analytical conferences in March 2021. - The fourth quarter was our 12th consecutive quarter, when the overall year-on-year growth of laptops with our graphics accelerators was in the double digits. The release of laptops with GeForce RTX 30 graphics became one of the largest in the history of sales - more than 70 different devices were released... Now we have 50 million gamers playing on laptops with GeForce graphics.

Unlike AMD and Intel, Nvidia only supplies discrete graphics, and its quantitative product shipments have decreased due to an acute shortage of desktop products. Despite these problems, Nvidia's leading position in the discrete GPU segment has not been affected.

Dynamics of the change in Nvidia and AMD shares in the discrete GPU market, data from Jon Peddie Research

At the end of 2020, there was an increase in the cost of cryptocurrencies, which was one of the catalysts for the rise in demand for video cards. This, in turn, led to higher prices for these computer components. Thus, the MSI (Emesai Computer) Radeon RX 6900XT graphics adapter  has risen in price by more than 100%. Great demand for video cards was observed both in the world and in the Russian market.

According to experts, the rush demand of miners for these components aggravated the previous situation: due to the COVID-19 pandemic and the downtime of the production facilities where the cards are assembled, interruptions in their supplies occurred back in the second quarter of 2020. To this was added a high demand for cards for gaming computers.

John Peddy of JPR, who has been watching the graphics market for more than three decades, talks about a new version of Ethereum. In his opinion, this option of cryptocurrency will make it unprofitable to use video cards.

Ethereum is now very convenient for mining video cards, but after the introduction of version 2.0, this will remain in the past, he said.

Notes