Materials Science and AI
The ironies of modern technology abound. The more important, relevant, and impactful a new technology is, the more entrenched the fundamentals of that discipline are to the history of science. For example, it is an amazing fact that, generally, mathematics precedes technology by about 100 years. That is, some mathematician in some university discovered an abstract axiom, complete with proof that no ordinary intelligent human being could even read, but it makes sense in that niche community of intellects who become bored when confronted with Mensa puzzles. Then, decades or centuries later, technology is developed that is of sufficient sophistication that, to understand the inner workings, the technologist applies the aforementioned mathematical abstraction.
Remarkably, the math provides a cogent explanation. It happens all the time. The same point can apply to certain disciplines of hard sciences such as chemistry, biology, and, of course, physics. Often, it’s all together.
By now, the readers of this column should know that they have inadvertently stumbled into a discussion on materials science. Is materials science chemistry? Is it physics? Could it also include biology? To the intrepid young scientist unpolluted by the random barriers imposed by the arbitrary lines within these disciplines, there is little meaning to the subjects outside a high school class. However, these casual definitions apply to every high school and college on the planet, extending all the way to the Nobel committee members when they hand out their valuable and prestigious prizes.
That brings us to the focus of today’s discussion: What is materials science? For the purpose of this dialogue, let’s agree that materials science is the least understood, most unknown, least lucrative of any scientific discipline that a budding graduate student might be willing to spend 8 years deciding upon and composing an acceptable thesis to be awarded the vaunted title of Ph.D. The exalted reward is usually a nontenured position in a midwestern college that allows the scientist to continue living in poverty and befits someone with supreme academic credentials whose colleagues patronizingly refer to as doctor. However, it is an unambiguous axiom that is as misunderstood as abstract mathematics while simultaneously being as clear as plastic that nothing exists in science or its close cousin, modern technology, more pertinent to the future success of civilization than materials science.
It is instructive to consider some examples. The value of alternative energy sources such as solar, wind, and tidal are all pointedly diminished if the instantaneous demand for power fails to match the level of generation. Nuclear- and carbon-based power can be generated and throttled to meet demand, but solar and wind work on their own schedule, not on that of the Department of Energy.
Batteries, those ubiquitous bricks of many shapes and sizes of electrical energy, are very effective. However, the present state of battery technology has limits. Modern batteries are constructed from the rare earth minerals of lithium, cobalt, manganese, zinc, nickel, and other elements that are hard to obtain and expensive. The sum total of the last 50 years of technological development in battery technology has been to make batteries denser, heavier, and even more expensive.
Even the vaunted lithium-ion battery, invented in the 70s and responsible for the electric vehicle industry, is limited in capacity. Oh yes, don’t forget that a crack in an 1800-pound, $15,000 EV battery results in a hazmat emergency of sufficient urgency to shut down an interstate highway for 6 hours during rush hour.
How does science solve problems of this magnitude? I’m sure by now that the reader has guessed the answer: breakthroughs in materials science.
An example from the realm of modern celebrity may help to accentuate the point. If one of the many companies owned and directed by the present technology megastar of the 21st century, Elon Musk, finds new, lighter, cheaper, and safer materials to build better batteries with greater capacity and that can be charged more that 700 times before having to be “recycled,” he becomes the world’s first trillionaire and many times over.
This sphere of thinking applies in every application and industry, even the most seemingly mundane. The reader may be puzzled as to why an article about AI has been awkwardly droning on about batteries for more than half its allotted column space. Well, I’m getting to the point. Of the many problems confronting the ever-accelerating brave new world dominated by the simulated thinking of GPUs, unlimited data, and the algorithms that make up machine learning, the technology of AI consumes more power than any inappropriate analogy or euphemism that I can dream up. Shall we agree that this means “a lot”?
Larry Ellison and OpenAI, financed by Softbank, are collaborating on building a single AI computer the size of the Dallas Cowboys’ stadium that requires 1.2 gigawatts to run.
It may seem like science fiction, but that is equivalent to the output of a dedicated nuclear power plant. At a half-trillion dollars and a half-million GPUs, it’s a bargain at twice the price, but I digress. The mind-boggling price tag or the power requirement or the consideration of where such a Death Star-like structure will exist is a daunting, but not the biggest, challenge. That challenge is how it will be possible to control the heat within what will be the world’s biggest data center.
Readers who have had the pleasure of working in data centers do not need further explanation. They have experienced days with outside summer temperatures of 110° Fahrenheit, as could be the case in Abilene, Texas, when they still had to bring a coat to work because the data center air conditioning was set to 60°F.
Since modern servers are primarily water- or liquid-cooled, the massive waste of energy in the form of dissipated heat should be contemplated. Sometimes, depending on local and state regulations, that water is simply funneled into the adjacent waterways. “They are boiling the alligators” is a glib phrase I once heard from a protestor. While silly and funny, it may also be sometimes true. This is again a problem for the intrepid and impoverished guardians of the niche yet increasingly scientific discipline of materials science.
Submer and Hypertec are amazing companies that both develop or include these incredible new coolants and cooling systems in their AI systems architectures. Mohan Potheri of Hypertec appeared on the DonSullivanShow.com podcast in August 2025 and discussed the various new liquid coolants being studied, developed, and used to cool and capture the heat generated in these immense data centers.
A few concepts and terms that everyone who has read this far will want to become familiar with include “synthetic dielectric coolant” and “immersion cooling.” I’m not going to try and explain this wonderful materials science in the little space I have remaining, so I’ve linked both the general technical explanation from Perplexity and the podcast discussion with Potheri below:
And finally, please support your local impoverished materials scientist.