07/25/2026
Experiences and thoughts regarding "water consumption" by "data centers" or more triggeringly referred to as "AI data centers". Really the same thing, just different naming and characterizations.
A) Data centers perform a number of computing services, including but not limited to web and/or e-mail hosting servers, other automation servers, server farms for larger web sites and shopping, data drop sites/boxes, instant/direct messaging services, other modern, and traditional communication systems (phone, etc.), off-site backup systems/services and a myriad of other services and, yes, AI. Basically, a data or colo center simply provides housing and resources for computer services and networking, usually along with security physical and otherwise.
The moniker "AI data center" comes across to me as scare mongering given that many folks are scared and/or concerned (I am the latter, a topic for another post) regarding AI in general. Data centers often/usually house multiple services unless owned by a single company to house a server farm for a very specific purpose. Thus, said moniker seems a gratuitous characterization created and used to start the conversation with a scary subliminal message. The discussion is much better supported and continued by using more appropriate and fitting terminology.
B) Looking at a number of varied sources, the USA already has between 4,000 and 5,000 data centers with most accounts trending toward the higher number. Conversely, according to "Search Assist", there are around 1,725 planned data center projects in the USA. Some sources report less than half that amount. Those numbers are food for thought regarding existing impact vs. planned impact. Based on those sampled numbers, that is a projected growth of maybe between 20% and 36%. Given that a number of any type of build projects, often are scrapped, often PRIOR to even breaking ground due to financial and permitting processing issues (and that is without any of the current anti-"AI-data-center" hype involved). There is some pretty fuzzy math involved as the sources vary so widely but possibly useful for discussion.
C) One laughable assertion regards the projected immense amounts of water will be consumed to "cool the chips" in the computers in data centers. I have worked in a number of data centers or colo centers. None of them had any water plumbing. I don't recall the fire suppression systems for all of them but for those I do recall, they were "dry fire suppression" systems, as in zero water or other fluids. The first colo center that I worked on and in, I was involved in the design phase and that dry fire suppression system was a key component that came up many times during both the design and construction phases. Further, pains were taken regarding any external water sources invading the center such as water run as the center was below ground in a parking garage. During the first near flood event, that water drainage system was inspected for effectiveness and any concerns were addressed immediately.
The only cooling system related fluids I have seen were in contained systems within the HVAC systems used to cool the entire room. All computers and other electronic devices only used heat syncing or fully contained cooling systems. The only "water consumption" I have witnessed was condensate having formed on post heat exchanging portions of HVAC units that was evaporating back into the very air (or adjacent air) that it was condensed from in the first place, thus, net zero consumption!
D) My only real concerns would be regarding power consumption and needed grid upgrades. However,
i) Said consumption, while it might increase demand initially, seems to result in excess supply that more than balances that out regarding the pricing impacts of supply & demand. I have seen that in a non-data center project that needed power wherefore, excess power generation was acquired that was going to over supply the local community dropping power costs (until local goverment chased away the targeted industrial customer along with hundreds of jobs and other negative related economic impacts still being felt years later).
ii) Much of the USA electrical grid requires upgrades just to rejuvenate the badly aging grid in many states. A significant number of past wild fires in California in the last decades were from failed electrical equipment and plant that was waiting on permitting and other process issues. The State took PG&E (possibly other grid companies) to court but lost when evidence indicated the bulk of the delays in upgrades and repairs was due to state and local regulations and even broken processes. That backlog of replacements and improvements just to deal with the aging grid in that state and many others is very much a problem as it stands. If data centers prompt a more rapid grid improvement or updating of aging systems, that may be considered a very positive outcome.
E) The USA's leveraging of technology over the last several decades is credited as a significant part of why USA labor-force productivity had surpassed the rest of the world and remains in the top 10. And labor force productivity increases arguably increase quality of life, economic vitality and better for the environment (again arguably).
Increased technical power should continue to improve productivity. Initially some jobs are lost. Past being prolog indicates that a rebound of new jobs to support new technology as well as other jobs to support economic activities that leverage technology also results in new jobs in less technical areas.
For instance, due to advances in technology in recent years USA's workforce as well as other workforces have become much more able to work remotely. This results in considerable reduction in the net consumption of energy of all types including fossil fuels. The commute, for many, has become, roll out of bed, and a short walk to a home office, or sometimes put on clothes and walk to a "remote office" building shared by other remote workers in the community whose work may be nearby or across the globe. Personally, I work remotely a considerable amount of my time with work locally and as much as thousands of miles away. My remote access is dependent on data centers that house/host the servers and software that I use for my remote work as well as daily functioning of work and processes for many of our clients.
If you "Zoom" or use "Teams" you are using a data center "in the cloud". If you web browse for shopping, researching, social networking, e-mailing and other messaging, you are most likely using multiple data centers!
The question, as always, should be, "What is gained vs. what may be lost or consumed (or for that matter, what consumption might be avoided) and estimate if the balance results in a net loss, or, indeed, a net gain?"
My feeling is that data centers are a net gain even for water. Don't forget that commuting requires water in ICE vehicles and while they use a closed system, it is not as closed a system as say an HVAC system. Further, if data centers help remove some of the need for moving people around for shopping, working, whatever, then that seems to net a reduction in energy consumption, too!
As someone who has for decades and continues to work in, around and with data centers, now on a daily basis, most or nearly all the arguments against more data centers has come across as uneducated (?by design?) and emotional hype and scare-mongering.
