Why The Best AI Stocks Could Be Cooling The Chips

When investors look for the best AI stocks, they typically focus on AI data centres or the chip manufacturers behind the AI boom.

But data centres and chips aren’t much use if they can’t keep their cool.

Previously we discussed how there is “no AI without power” and how energy is flowing through AI data centres to turn prompts into output. However, every watt of energy becomes heat and heat puts AI technology at risk.

As a result, almost 40% of the energy usage of AI data centres are used for cooling purposes.

Yet, it is a trend many looking for AI stocks may overlook.

Why do AI data centres run so hot?

Before we get into artificial intelligence, let’s go back to high school science and the first law of thermodynamics.

“Energy cannot be created or destroyed; it can only change from one form to another.”

This is what is happening every day at AI data centres.

AI chips consume energy when they run. But this energy has to go somewhere and that energy gets transformed into heat.

When fully operational, many modern AI chips can reach between 176°F to 185°F (or between 80°C and 85°C). With a typical data centre being made up of as many as 10,000 to 500,000 chips depending on its size, that is a lot of really hot, expensive, technology.

And it could get worse in the future.

A key metric for AI data centres is rack power density, or the total amount of power consumed by all equipment inside a single server rack.

Some estimates put a traditional enterprise server room at around 11kW per rack. However, a modern Nvidia server rack used for artificial intelligence requires up to 142kW, a 12x increase in power requirements.

Goldman Sachs estimates that a typical AI server rack in 2027 could require 50x the power of a 2025 server rack. As a result, it says that data centres could soon require industrial level cooling systems akin to those seen at aluminium smelting plants. Recent estimates from Deloitte believe a single server rack could draw as much energy as 600kW by 2027.

Why air cooling can’t keep up

From the first law of thermodynamics, to the second, principally the part where heat moves spontaneously from hot objects to cooler objects but never the reverse.

AI chips concentrate enormous heat into small areas. And if a chip or server rack gets too hot, its performance can be impacted, if not fail completely.

So, AI data centre operators need to find a way to keep their servers cool.

Historically, data centres cooled their servers by blowing cool air onto the racks but this is no longer cutting it. Air cooling stops being effective at around 50kW and as you read earlier, we’re currently at around 142kW and higher.

Enter liquid cooling methods.

Whether it is the use of liquid coolant moving through metal cold plates and rear door heat exchangers that get help move or exhaust hot air away from server racks or total immersion, where the chips are completely immersed in a specialised liquid, the use of more complex liquid cooling rather than air is increasingly the go-to approach for AI data centres.

Which stocks are cooling the AI boom?

With cooling becoming a critical problem for data centres, it is no surprise that many companies are looking to take a slice of this market.

The Global X AI Infrastructure ETF (ASX:AINF) is a great example of a portfolio being built around finding companies that are seeking to solve AI’s power problem. This includes, on top of other key infrastructure challenges of AI, how to cool the chips needed to run it.

Global X identifies companies like Schneider Electric (ENXTPA:SU), which is one of the portfolios’ five largest positions. Schneider provides, amongst other things, liquid cooling solutions and intelligent airflow optimization technology.

Another large position is Eaton (NYSE:ETN), which provides thermal and liquid cooling systems specifically designed for energy-intensive AI data centres.

What does it mean for investors?

AI can only reach the heights of its most bullish advocates if the technology behind it stays operational and cooling has become one of the most critical challenges facing artificial intelligence.

Yet, many investors focus on the obvious AI names, like the chip manufacturers, the data centre operators and other similar listed companies whose share prices have risen sharply over the last few years.

Cooling isn’t as sexy as chip manufacturing.

But it is no less important in bringing the AI boom to life. In fact, as increasingly more advanced chips are manufactured, the demand for cooling is likely to rise with it. This could make cooling an underappreciated pairing to the rise in AI chips and data centres and a fertile place for investors to look for AI stocks.

Explore the fair value estimate for the Global X AI Infrastructure ETF.

Disclaimer

This article was written independently by the author, without issuer input or approval. Global X Management (AUS) Limited has a marketing services agreement with Simply Wall St. Details on compensation and other important information can be found in the disclosure and disclaimer at the end of this narrative.

Have feedback on this article? Concerned about the content? Get in touch with us directly. Alternatively, email editorial-team@simplywallst.com

Simply Wall St analyst Andrew Legget and Simply Wall St have no position in any of the companies mentioned. This article is general in nature. We provide commentary based on historical data and analyst forecasts only using an unbiased methodology and our articles are not intended to be financial advice. It does not constitute a recommendation to buy or sell any stock and does not take account of your objectives, or your financial situation. We aim to bring you long-term focused analysis driven by fundamental data. Note that our analysis may not factor in the latest price-sensitive company announcements or qualitative material.