TECH COLUMN

Chapter 0: Ultra-Future Forecast ― “Why Will the Future Be Decided by Heat?”
0-1. Will AI Be Stopped Not by Intelligence, but by Heat? ― Future Performance Competition May Actually Be a Cooling Competition ―

Heat

release:
Chapter 0: Ultra-Future Forecast ― “Why Will the Future Be Decided by Heat?”<br>0-1. Will AI Be Stopped Not by Intelligence, but by Heat? ― Future Performance Competition May Actually Be a Cooling Competition ―

The evolution of generative AI, led by ChatGPT, has accelerated explosively over the past few years.
Writing text.
Creating images.
Creating videos.
Writing programs.
Conversing with people.
A world that was science fiction not long ago has already become reality.
Behind the scenes, however, a problem has been quietly growing enormous.

That problem is heat.

■ The Evolution of AI Is Actually a Battle Against Heat

The smarter AI becomes, the more computation it performs.

Performing massive computation requires a massive number of GPUs (graphics processing units).

GPUs, in turn, consume a large amount of electricity.

And electricity almost always ends up as heat.

In other words, the structure is:

Evolution of AI

↓

Increase in computation

↓

More GPUs

↓

Higher power consumption

↓

More heat generation

■ Why Do AI Servers Get So Extremely Hot?

In today’s AI servers, there are a growing number of cases in which a single server uses several kW to several tens of kW of power.

This exceeds the power consumption of an ordinary household.

Moreover, much of that energy turns into heat.

In other words, an AI server is a computer and, at the same time, a gigantic heat-generating device.

■ This Is Where Cooling Becomes the Problem

CPUs and GPUs reduce their performance when they exceed a certain temperature.

If the temperature rises further, it can cause failure or runaway.

In other words, if you cannot cool it, you cannot raise performance.

■ Is the Future of AI a Cooling Race Rather Than an Intelligence Race?

In recent years, news about improvements in AI performance has kept increasing.

Behind the scenes, however, companies are working hard on technologies such as

● Liquid cooling

● Heat dissipation materials

● Data center cooling

● Reducing power consumption

● Heat spreading

● Thermal control

In other words, the future is beginning to become a competition not only over how smart an AI you can build,

but over how well you can make heat work.

■ Why Heat Becomes a Constraint on Future Industries

This is not just about AI.

● EVs

● Robots

● Semiconductors

● Medical devices

● The space industry

● Wearable devices

What they all have in common is the demand to become smaller while becoming higher-performance.

However,

Miniaturization

↓

Higher density

↓

Concentrated heat generation

↓

Heat cannot escape: this problem always arises.

In other words, future industries always come back, in the end, to heat problems.

■ The Researcher's Perspective: "How Do We Get Heat Out?"

In the world of research and development, various technologies are being developed, such as

● High thermal conductivity materials

● Graphene

● CNT (carbon nanotubes)

● Next-generation TIM

● Vapor chambers

● Radiative cooling

● Liquid cooling

Especially in AI applications, the limits of conventional air cooling alone are becoming visible,

and the shift to liquid cooling technology has begun to advance rapidly.

■ The Shop-Floor Perspective: Why Things Do Not Cool as Theory Predicts

On the shop floor, however, much messier problems occur.

For example,

● Thermal sheets lift

● Air layers form at the contact surface

● Lamination positions shift

● Materials deform

● Tiny steps increase thermal resistance

● Automated machines cannot supply parts stably

● Performance changes between prototyping and volume production

In other words, on the shop floor, using a good material is not the end of the story.

■ Thermal Management Becomes a Question of Making the Process Work

Thermal management is not determined by material performance alone.

● How to apply it

● With what precision to process it

● How to make contact

● How to mass-produce it

● How to supply it stably

Only when all of these work does the performance come out.

In other words, thermal management is a question of making the process work.

■ What Matters from OTIS's Perspective

From OTIS’s perspective, heat problems are not simply a matter of selecting heat dissipation materials.

What matters is

● Contact

● Processing accuracy

● Microfabrication

● Lamination

● Supply format

● Volume production stability

In particular, the higher the performance, the more sharply the difficulty of the final 1mm of heat increases.

■ What OTIS Can Do

At OTIS, through

● High-precision die-cutting

● Thin-film processing

● Microfabrication

● High-precision lamination

● Processing of thermal management materials

● Stacking of irregular shapes

● Reel supply

● Support for automation

● Stabilizing volume production

we can contribute to bringing thermal management materials closer to a form that can be used in volume production.

■ Outside OTIS's Expertise

On the other hand, OTIS is not a company that specializes in

● GPU design

● Semiconductor design

● Cooling equipment design

● Dedicated CFD analysis

● Development of the materials themselves

However, in the area of making heat solutions work in volume production processes, we may be able to play an important role.

■ Summary

The future may be decided by heat

The evolution of AI is also the evolution of intelligence.

At the same time, however, it is also a battle against heat.

Future performance competition will be decided by whether all of the following can be achieved at the same time:

● Computing performance

● Power consumption

● Cooling

● Thermal control

● Assembly

● Volume production

And behind the scenes, inconspicuous but important technologies that support the final 1mm of heat will be needed.

Heat problems are not a behind-the-scenes issue for future industries.

Rather, they may be the constraints that determine whether the future itself can be realized.

 

Column Supervisor: Koji Kakumoto (President, OTIS Co., Ltd.)

After studying abroad and working in planning and development at a trading company, he joined OTIS Co., Ltd. in 2011. While primarily working in the Corporate Planning Department, he has also served concurrently in manufacturing and technical divisions, and since 2018 has served as President, working to drive business growth and strengthen the organization.

 

This article is a general technical explanation intended for educational purposes and does not refer to any specific company, product, or technology.

Latest Articles

Related Articles

  • 10-10. Proposals Based on the Customer’s Process (Part 3) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    10-10. Proposals Based on the Customer’s Process (Part 3) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    In Part 1, we looked at the perspective of designi... Read More
  • 10-10. Proposals Based on the Customer’s Process (Part 2) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    10-10. Proposals Based on the Customer’s Process (Part 2) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    In Part 1, we looked at the idea of understanding ... Read More
  • 10-10. Proposals Based on the Customer’s Process (Part 1) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    10-10. Proposals Based on the Customer’s Process (Part 1) ― A Component Only Becomes Good When One Looks Beyond the Drawing to “How It Is Used” ―

    Requests to processing companies often start from ... Read More

CONTACT

Okayama

+81-867-42-3690

Tokyo

+813-6810-4830

Contact