Is the AI's "home" custom-made? What is the "AI-enabled data center" that will support the future?
Hello, I'm John, an AI technology commentary writer!
Recently, AI has become more and more familiar in our lives, such as chatting with people on ChatGPT or having AI draw beautiful illustrations for us. But have you ever wondered, "Where on earth do these smart AIs live?"
In fact, there are “homes” and “workplaces” where AIs learn and answer our questions."Data Center"Now that AI is evolving at a tremendous speed, it has become necessary to upgrade this data center to a special one just for AI."AI-enabled data center"What is it?
Today, I will explain in an easy-to-understand way all about AI-enabled data centers, the "unsung heroes" that support this AI era behind the scenes!
What is a "data center" anyway?
Let’s start with the basics. In a word, what is a data center?"A huge, ultra-high-tech facility packed with lots of computers (servers)"You read it right!
Think of it like a giant library, except instead of books lining the shelves are a bunch of servers, which store and process all kinds of digital information, including photos and videos from the social media we use every day, website information, and important corporate data.
These data centers enable us to use the Internet comfortably anytime and anywhere, and they can truly be called the heart of modern society.
Why does AI need "special" data centers?
So why do ordinary data centers not work and why do we need something "special" for AI? Because AI requires a lot more "power" than ordinary computer work.
Training an AI, especially a large-scale AI model, is like a human reading millions of books, understanding them all, and creating new knowledge. This requires three main special skills:
1. Tremendous computing power
AI training requires a huge amount of calculations, the kind of calculations that would take years to complete using just the CPU (the brain) of a regular computer.
That's where the action is"GPU"This is a component. It was originally developed to smoothly display the beautiful graphics of computer games, but it is very good at "performing many calculations at once." We need to gather thousands or tens of thousands of these GPUs to train AI all at once.
2. The ability to move large amounts of data quickly
AI learns by "eating" huge amounts of data, which serve as textbooks. The amount of data can be as much as tens of thousands of movies. This huge amount of data must be transported smoothly from the storage location to the GPU that performs the calculations without any blockages.
To put it in perspective, it's like there would be huge traffic jams on normal roads (conventional networks), so we would need super expressways (high-speed networks) with 10 lanes on each side.
3. Cooling power to keep everything cool
A GPU running at full power generates a tremendous amount of heat. A server with many GPUs is a mass of heat. If left unchecked, the heat will destroy it.
Therefore, AI-enabled data centers need to have cooling that is far more powerful and efficient than a home air conditioner.cooling systemBy keeping the entire facility at an optimal temperature, the AI can continue to work stably 24 hours a day, 365 days a year.
Summary of the main features of "AI-enabled data centers"
To summarize what we have discussed so far, data centers that can maximize AI performance have the following characteristics:
- Powerful GPU Servers: There are many powerful computers that specialize in performing AI calculations.
- Ultra-fast network: A communications network for instantly exchanging huge amounts of data between servers.
- Advanced Cooling System: A system for efficiently cooling the large amounts of heat generated by servers. A technology known as "liquid cooling," which uses liquid to directly cool the servers, is also garnering attention.
- Huge Storage: A place to store huge amounts of data for AI learning.
- Stable power supply: All of this requires a huge amount of power, so a stable, uninterrupted power source is essential.
Only when all these elements are in place can a data center be called an "AI-enabled data center."
How does it relate to our future?
You might be thinking, "That sounds complicated, but how does it relate to our daily lives?"
The development of this AI-enabled data center will further advance AI technology. For example,
- The arrival of AI assistants that can have more natural conversations
- Development of new medicines and improvement of early detection technology for diseases
- Further improving safety with autonomous driving technology
- Improving the accuracy of climate change predictions and disaster simulations
This data center is supporting the future where AI will play an active role in all fields, from medicine to entertainment and social infrastructure. The amazing new services and technologies that we will see in the future may well be born from here.
When we ask AI questions or generate interesting images on our smartphones, it's kind of exciting to think that such large-scale equipment is running at full speed behind the scenes. It really makes you realize that the remarkable evolution of AI is firmly supported by technological innovations in these unseen places.
This article is based on the following original articles and is summarized from the author's perspective:
Download the 'AI-ready data centers' spotlight
report