Do you know what data centers and LLMs are?
Why are so many huge data centers being built and why do they use so much energy?
See images HERE (iStock, Getty Images)
DATA CENTERS are humongous buildings that take up a lot of previously beautiful open space.
They are big because they require both massive amounts of power and lots of cooling to manage the massive computation required.
See images HERE (same images, just in case you didn't look the first time).
LLMs are "large language models" that do our thinking for us superfast. They are called Large Language Models because their massive size enables them to understand and generate language in ways that smaller, earlier AI systems could not. (AI means "artificial intelligence.")
NVIDIA is a company that sells "GPUs" — graphics processing units — that power the large language model services that are behind the whole AI boom, either through "inference" (the process of creating an output from an AI model) or "training" (feeding data into the model to make its outputs better).
GPUs are great for parallel processing — essentially spreading a task across multiple (thousands of) processor cores at the same time — which means that certain tasks run faster than they would on, say, a CPU (central processing unit). While not every task benefits from parallel processing, or from having several thousand cores available at the same time, the kind of math that underpins LLMs requires it.
The central processing unit (CPU) is the primary functional component — the brain — of a computer, the invisible manager inside the computer where data input is transformed into information output (what the human brain can understand).
Like the human brain, the CPU can multitask. It runs a computer's operating system, simultaneously regulating the computer's interactions. It stores and executes program instructions and apps (applications) through its vast networks of circuitry. It manages a variety of other computer operations.
With Blackwell — the third generation of AI-specialized GPUs — came a problem, in that these things were so much more power-hungry that previous generations and required entirely new ways of building data centers, along with different cooling and servers to put in them, much of which was sold by NVIDIA.
While you could kind of build around your current data centers to put A100s and H100s into production, Blackwell was less cooperative and also ran much hotter. NVIDIA's third-generation Blackwell GPUs require entirely new servers — and if you want to run lots of them, an entirely new data center, because they require so much more power and cooling. (I know this is a terrible paragraph and welcome editing from someone who knows what they're talking about.)
NVIDIA also makes the consumer graphics cards you can find in a gaming PC or gaming console, but 90% of NVIDIA's revenue now comes from selling either GPUs for LLMs, or the associated software and hardware to make it all run.
CUDA (Compute Unified Device Architecture) is a proprietary parallel computing platform and application programming interface (from API) that allows software to use certain types of graphics processing units (GPUs) for accelerated general-purpose processing, significantly broadening their utility in scientific and high-performance computing. CUDA was created by Nvidia starting in 2004 and was officially released in 2007. When it was first introduced, the name was an acronym for Compute Unified Device Architecture, but Nvidia later dropped the full name and now rarely uses it. CUDA is proprietary to NVIDIA, and while there are alternatives (both closed- and open-source), none of them have the same maturity and breadth.
Pair that with the fact that Nvidia's been focused on the data center market for longer than, say, AMD, its main competitor, and it's easy to understand why it makes so much money. There really isn't anyone who can do the same thing as NVIDIA, both in terms of software and hardware, and certainly not at the scale necessary to feed the hungry tech firms that demand these GPUs.
A FEW TERMS EXLAINED:
Like the human brain, the CPU experiences both short-term memory and long-term memory.
With a computer, memory usually takes the form of short-term storage for the files most often accessed during recent computer use. When a piece of data first enters an operating system (OS), it's placed within that OS's random-access memory (RAM). A CPU's standard operating memory stores only RAM data "in the moment," similar to a person's short-term memory, before periodically purging it from the computer's cache memory.
"Permanent storage involves read-only memory (ROM), which means data can be accessed but can't be acted upon or altered."
"Secondary storage is akin to long-term memory in humans and involves the permanent or long-term retention of data by archiving it on secondary storage devices, such as hard drives. Output devices like hard drives offer permanent storage."
Open Source means a particular code is public, allowing anyone to view, modify, and distribute it.
Closed Source means the code is private and proprietary, with access restricted to the company that created it.
RESOURCES LEANED ON OR LIFTED FROM, ABOVE:
• Phil Powell, IBM. What is a central processing unit (CPU)?
• The Hater's Guide to Nvidia (Edward Zitron, Where's Your Ed At, 11-24-25)
• Large Language Models (LLMs) From Dummies (YouTube, Pablo Cingolani, Part 1: "Attention")
• LLMs EXPLAINED in 60 seconds (YouTube, @ShawhinTalebi) If you've heard of ChatGPT, you've heard of large language models. But LLMs are not necessarily chatbots. LLMs are word predictors.
What key resources are missing? Please let me know.
Feel free to comment and to link to articles that explain how all this works and what the main problems and suggested solutions are.
Should we let data centers take up so much visible space?
Can we stop their encroachment on our national green space?
I've tried here taking a stab at identifying major problems, the big picture, with these data centers.
I am no expert. I just know this is a looming, escalating problem and that we need to be informed about it.
I welcome links to excellent explanations and information that help clarify what's going on and how to wrestle with it rationally and nationally.
-- Pat McNees, Writers and Editors website