Even ChatGPT-4o Doesn’t Understand Liquid Cooling!

Data Center Liquid Cooling AI Fail
Warning! This is my attempt to interject a little engineering humor into the otherwise bland subject of data center design liquid cooling.

Warning! This post is my attempt to interject a little engineering humor into the otherwise bland subject of data center design. First off, let’s set aside the irony that only a few years ago introducing water into the data center white space was considered blasphemy.

As the founder and CEO of a company that designs, builds, and maintains data center facilities, I write and present a lot. Lately, I mostly talk about how AI and HPC workloads affect the data center facility -that means liquid cooling at the rack level. But like when “cloud” was new, “liquid cooling” can mean many different things depending on the context.

My goal is to create a series of “explainer” videos that will cover the topic data center liquid cooling in depth, show my expertise in the field (I have it 😉) and help some potential clients. The first step for this project is to do the research. As any good researcher/writer knows, ChatGPT-4o and similar tools are very useful for this.

Anyway, the point of the story is that along the way, I typed the following prompt: “Create for me a series of pictures depicting large scale and small scale direct to chip and rear door liquid cooled AI and/or HPC deployments.”

What follows are the pictures that DALL-E produced…

Data Center Liquid Cooling AI Fail

When I saw this image, I thought, “Wow, these are incredible.” Immediately after, I thought, “Wait a minute! I have no idea what I am talking about if these are real!” That’s right, I thought, “If AI generated this, it must be real!” After a moment or two of thinking, and reason taking over, I began to see the humor.

What I love most about the picture above is the child-like simplicity that went into the idea of “direct to chip liquid cooled AI…” I figure it went something like this, “How can we get liquid directly into the racks? Hmmm. I know liquid comes in kegs, I’ll just put the kegs on the top of the IT racks! But how will they know what they are? I know, I’ll just write DIRECT-CHIP LIQUID CHIP on one of the kegs, put it on some air-ventilated coasters, and place it on the floor beside them!! Genius!”

Data Center Liquid Cooling AI Fail

I am equally amused by DALL-E’s second picture. I love the fact that it opted, in this case, to use a series of “sixtel kegs” in lieu of a full keg. Perhaps trying to represent different beers, I mean liquids?! I am puzzled however, by the need for the “front liquid cooling equipment, and trash can?”

I’ve saved my favorite for last…

Data Center Liquid Cooling AI Fail

There is so much to unpack here!

  • Intra-cabinet sixtels!
  • Cabinets with 2 server fronts (figure it out)
  • Cabinets with no rear access required (liquid cooling is amazing)
  • Frankenstein-like injection vials above a ventilated toroidal transformer??
  • And of course, large, clear glass distilling bottles!!
  • But my favorite is the need for the industrial, presumably high-velocity, axial fan above the two left-hand cabinets.

I need to catch my breath. 😊 Anyway, I hope you enjoyed reading this as much as I enjoyed writing it.

Please look out for my previously mentioned video series. I cannot promise it will be as humorous, but I’ll do my best.

On a more business note, we here at INTUVA, a division of PTS Data Center Solutions, are at the forefront of designing and constructing adaptable data centers that meet the emerging needs across various geographical and industrial landscapes. Among many disciplines, we are in fact specialists at integrating power-dense servers, including liquid-cooled racks, into colocation and edge data centers, ensuring optimal performance for AI and high-performance computing (HPC) applications. Meanwhile, another of PTS’ divisions, GRID7 Energy is pioneering the transformation of power solutions for critical facilities. GRID7’s innovative approach includes the deployment of natural gas linear generators, solid-state fuel cells, and renewable energy sources such as solar, wind, and battery energy storage systems (BESS).

Together, INTUVA and GRID7 are not just anticipating the future needs of data centers but actively shaping them, ensuring our clients are prepared for a landscape where power availability and efficiency are paramount.

For more information on how we can help your organization stay ahead of the curve in data center design and energy solutions, visit INTUVA Solutions and GRID7 Energy.

Share the Post:

Related Posts

Join Our Newsletter