LOGO

In this issue, we explore NVIDIA, an American semiconductor manufacturer that has been in the news since its market capitalization briefly reached the $1 trillion mark at the end of May this year.

About GPUs manufactured by NVIDIA

NVIDIA manufactures GPUs (image processing semiconductors), which are chips designed for display functions such as video, image, and animation display, and have been used in gaming PCs to display images smoothly. In recent years, GPUs have come into the limelight as the bearers of advanced arithmetic processing in automated driving technology and cryptographic asset mining operations.

Now, demand for these GPUs is growing. The catalysts for this are “data centers” and “generative AI” represented by ChatGPT.

GPU, generative AI and data center

Until now, it was common for data centers to be equipped with only CPUs (Central Processing Units), but with the spread of AI, data centers are increasingly being equipped with GPUs in addition to CPUs. However, only about 10-20% of data centers are equipped with GPUs.

However, the situation will change with the spread of generative AI.

Generative AI, such as image generation and natural language generation, requires a process called “inference,” in which an AI model created through training is run to reach a conclusion. The inference process requires more computation than the learning process. Therefore, it is necessary to have a GPU that is also suitable for a large amount of computation.

It is clear that generative AI will be the primary information-generating task in most of the world’s data centers in the future, and that within another decade, most of the world’s data centers will be equipped with GPUs.

In NVIDIA’s most recent quarterly results (May-July), sales in the data center division more than doubled in just three months, even though shipments are not keeping up with demand due to a severe supply shortage. Analysts expect the division’s revenues to exceed $60 billion in the next fiscal year (ending January 31, 2025), more than four times last fiscal year’s (ending January 31, 2023).

Why does Nvidia have such a strong lead?

Background to NVIDIA’s near monopoly on the GPU market

NVIDIA was positioned to advance AI very early on. In 2006, NVIDIA announced CUDA, a programming language for developers to write applications for GPUs. CUDA became an important component for subsequent AI projects.

CUDA eventually grew to include 250 software libraries used by AI developers, and this breadth effectively made NVIDIA the go-to platform for AI developers.

CUDA protects NVIDIA as a competitive “dig” that rivals can never overcome. In a July conference call hosted by Bernstein Research, former NVIDIA Vice President Michael Douglas noted that software is the key to NVIDIA’s ability to pull away from its competitors. He predicted that much of the performance improvement of Envidia’s systems over the next few years “will be software-driven, not hardware-driven.”

The key to Nvidia’s monopoly was software development.

For the time being, NVIDIA remains strong in the market.

For the time being, the market for GPUs for data centers is expected to be almost exclusively dominated by NVIDIA.

Nevertheless, competition is likely to intensify. In addition to competition with semiconductor manufacturers such as Intel and AMD (Advanced Micro Devices) that already handle GPUs, giant IT companies such as Google, Amazon, and Meta are also beginning to develop their own AI semiconductors.

Along with the further evolution of generative AI and NVIDIA’s developments, we will also be keeping a close eye on other companies dealing with GPUs.