The Artificial Intelligence energy paradox to data centers

Energy consumption from data centers continues to rise, and with artificial intelligence (AI), energy demand will increase even more. However, AI also holds the potential to monitor and manage energy usage much more efficiently.

As such, the role of AI in data centers therefore presents a paradox: energy consumption is significantly increased, AI also offers solutions for better energy management in data centers.

To fully comprehend this paradox with drawbacks and opportunities, here are some key elements to consider:

Machine Learning, High-Performance Computing and similar AI based algorithms are data and energy intensive and require high computational power capability.
Data storage and data management across systems becomes paramount to AI and is also energy intensive.
Data centers require reliable operations 24/7, and many AI applications require continuous and real time processing, which also leads to continuous energy usage.
With AI, it will also be possible to foresee equipment failure and take action for preventive maintenance.
Utilizing AI, it is possible to analyze energy consumption patterns in real-time and predict peak-time usage so that inefficiencies and potential areas for improvement can be identified and energy distribution can be optimized.

Subsequently, as AI advances, it poses both challenges and opportunities. One inevitable consequence that will impact datacenters is that heat generated by higher power density to perform AI related task will change cooling needs and require advanced and efficient cooling systems.¹

Liquid cooling for the future

Data centers are estimated to account for 3-4% of global power consumption by the end of this decade.² While AI can help predict heat generation patterns and adjust cooling mechanisms in real-time, their high-TDP chips generate more heat and require more efficient cooling systems. TDP (Thermal design power) refers to the maximum amount of heat a CPU or GPU is expected to generate under general usage and traditional air-cooled systems are becoming inadequate as the power increases. The data center liquid cooling market size is expected to see exponential growth in the next few years. It will grow to $10.61 billion in 2028 at a compound annual growth rate (CAGR) of 25.8%.³

Liquid cooling holds a promise for the future because of its superior advantage in cooling capacity. Water has more than four times the cooling capacity when compared to air and outperforms air at heat transfer conductivity by 25 times. Removing excess heat with water requires 50 times less energy than with air. It’s clear that liquid cooling becomes part of the answer as we move towards more power dense data centers.

Liquid cooling holds a promise for the future because of its superior advantage in cooling capacity (Source SWEP)

One of the main components of a liquid cooled system is the brazed plate heat exchanger, also known as BPHE. BPHEs are the most energy efficient heat exchangers on the market and are used in these systems to quickly and effectively transfer heat energy away from the processors.

Bphe in data center cooling

Pledge for climate neutrality accelerates shift

Besides the inherent challenge that AI will increase the energy usage of Data centers there are also other challenges. New energy goals and regulations are driving the push towards climate neutral data centers. One of the main goals in the EU is to make Europe climate neutral by 2050 and data centers climate neutral by 2030.⁴ This goal is shared by the CNDCP, CISPE and EUDCA. This ambitious goal is forcing data center owners to explore new ways of cooling data centers and capturing and reusing excess energy whenever possible.

The CNDCP is the Climate Neutral Data Centre Pact, a consortium of over 100 data center operators and trade associations, including Google, AWS, IBM, Microsoft, Meta, among others, that are accelerating the shift to climate neutral data centers.

The five pledges of the Climate Neutral Data Centre Pact include:

Purchasing 100% carbon-free energy
Prioritizing water consumption and conservation
Reusing and repairing servers
Proving energy efficiency with measurable targets
Looking for ways to recycle heat.

While AI inevitably will drive-up energy consumption, there is no doubt that AI will also accelerate a shift for solutions that can contribute to fulfill the ambitious climate neutral goals for data centers. With superior cooling capacity and heat transfer conductivity, liquid cooling will certainly play its part as we move towards more power dense data centers. The goal for carbon neutral data centers by 2030, while ambitious, seems within reach.

More about Data Center cooling here!

References: