When you hear the term artificial intelligence, or AI, does it cause you excitement, or dread? AI is a term that regularly pops up in the news, including headlines predicting an AI ”takeover” that would leave humanity at the mercy of the machines, but this is of course only a prediction. Guesses as to the future of AI’s influence on our world are many and varied, but when I think about it, I don’t even know what effect AI has on my world today. As an everyday GenXer, AI seems like such a distant concept to me that I feel it is not yet part of my life, but have I already been caught in the AI web? Have you? As I have learned while writing this blog, the answer is very likely yes. To answer this question, I went on a research journey to better understand the connection between AI and the data centers that house it, and how all of this influences me as a Hoosier. Please join me as I dive into the world of AI and data centers in this 5-part Hoosier data center industry blog.
Part I: Web Searches and the Data Center
I am not a student, journalist, or author, and I have no need to use OpenAI or ChatGTP. I write from scratch everything that I produce, and I am not in a job that requires using AI. I do, however, love to surf the web: the best new Mediterranean restaurant in Indy, how to know if the cut on my finger requires medical attention, will the comet or northern lights be visible tonight, the identification of a plant I’ve never seen before, etc. So I’m not using AI, right? Wrong! After asking my young adult children to teach me the difference between a standard web search (non-AI) and a contemporary one (AI-assisted), I came up with this library analogy.
A non-AI everyday web search (before approx. 2023) looked for existing material in the world wide web using key words to produce a list of single source webpages that match that key word. This is like a person going to the library to look for information in a variety of limited sources on a particular topic, which is then marked with sticky notes. This might be accomplished in a full afternoon by one person. On the other hand, an AI-assisted search (common from approx. 2023) results in an original product created by AI searching for requested information throughout the web, and compiling relevant information garnered from a multitude of sources into one summary. This is like a person going to the library to read all the material there to find information on a particular topic, marking and copying the relevant information, and finally writing a summary of that topic from every source in the library related to that topic. This is something that might only be accomplished over the lifetime of a person, if that’s even possible. Although it is just an analogy, we can conceptually compare the energy expended by a person doing research using these two different methods with the energy expended by a data center to complete the two different research methods. It is easy to see that a standard web search would require minimal energy for the data center to complete, while an AI-assisted search requires an exponentially larger amount of energy to produce a result. A person working a lifetime to complete a research project would require a mind-boggling supply of resources (sustenance, housing, income, health care, etc.), as a data center working to complete the requirements of AI likewise require seemingly infinite resources (electricity, water, acreage, building materials, maintenance, etc.).
By now, everyone reading this likely already knows that the use of AI requires the building of larger and larger data centers, and hears that data centers are “taking over the country.” But what does that really mean? What are data centers, and why do people say they are bad? Wait, I will answer that for you, please don’t google it!
Data centers, most simply put, are support centers for your laptop, computer, cell phone, and all other processing devices that use the internet. They are like the Union Station for computer lines. Let me bring in some Hoosier history to make another analogy here. Union Station in Indianapolis was the first central train station in the world, built in 1848. (Yup, although the current Union Station was built in 1886.) Trains spread quickly in Indiana, eventually developing into the interurban rail system at the beginning of the 20th Century, the country’s second-most extensive commuter rail system connecting nearly all communities in Indiana with rail lines. (Yup, that’s true too.) Smaller communities connected to larger ones like spokes on a wheel, and those larger communities connected with each other, which then all connected in Indianapolis at Union Station. Data centers are like this, the center of a wheel, connected to processing devices like spokes on that wheel which allows for connections (like a train) and powers the data exchange.
Indiana was truly the railroad crossroads of America 100 years ago. Will it become the data center crossroads of America a century later? That is yet to be seen, but as of this writing, Indiana has 72 data centers, with many more in the works. What does that mean for Indiana, or for me? I wasn’t sure myself, so I looked it up, again.
Data centers are not all the same, and are categorized by size: the smaller ones are called Tier 1, with the largest being Tier 4, or hyperscale data center. To generalize broadly, Tier 1 data centers support one company or entity only, and look more like a giant computer room, and are quite common in Indiana. The larger the data center, the more interconnected it is likely to be. The Tier 4 data centers are more like cell towers. A cell tower is built by a telecommunications company, such as Verizon, to receive and send data for its customers, as well as other smaller telecom companies that lease the facility. Hyperscale data centers are built by tech companies, such as Microsoft, to receive and send data over the internet, but also to create AI-generated data as requested by both its customers, as well as other tech companies’ customers. Hyperscale data centers are interconnected much like the interurban rail line system was. A rail customer in Madison, Indiana to the south could have traveled to New Carlisle, Indiana in the north by train, transferring once or twice at transfer stations, and could have returned to Madison by a different route should one path have happened to be closed. Data moves much the same way on the internet. A person doing an internet search in Madison could be making a request that follows the internet from Madison to the new Amazon Rainier Project data center in New Carlisle (which opened today, October 29, as I write this blog, at a cost of $11 BILLION, Amazon’s LARGEST data center to date), but takes a different track back if the Lifeline Data Center in Indianapolis is too congested.
So, large data centers move data along the internet and also create data. Data is produced by computer coding, so wouldn’t that mean that after the land is purchased, the data center building is constructed and the computers installed, not much more is needed to produce data? The intangible product of data ought to require only intangible effort, one might think, but this is unfortunately not true. Although the product of a data center is only data, massive amounts of electricity and water are needed to support the data center computers.
In the second post in the blog series I will explore the electricity needed to run a data center. Look out for part 2 next week!
Jennifer Ehara
Winding Waters Group Executive Committee
and Hoosier Chapter Sierra Club Communications Team