We will attempt to examine data centers, those physical entities which are composed principally computer servers. I have heard some TV financial news reader prognosticate on date centers. It was clear that he was totally clueless. He was eschewing the New York ban. Of course whatever he said had no basis in reality. He was saying there would be tens of thousands of data centers, of millions of square feet each. Just a side note: if there are tens of thousands of data centers each using 10 GW of power that would be 100,000 GW needed. Now as a note, the US today uses 1,400 GW of power[1]. We would need nearly 100 times more than what the US uses this very day! And this guy is some alleged financial genius!
1 Definition
If one seeks to address the issue of data centers one must first define what they are. A typical data center is a massive physical facility which principally is composed of thousands of computer servers. In addition there are inputs and outputs of fiber optic data lines. In the servers or adjuncts to them are massive amounts of memory.
The data center hardware demands massive amount of electrical power, often well into the Giga watt levels. The power may be externally provided or self-provided. The power must have significant redundancy, so that if one source fails an instant switch to a secondary source can be made. Maintenance is a critical factor for with thousands of servers and millions of terabits of date anticipated failures demanding human interface as well as replacements must be available.
The data center also demands massive cooling. The GW input creates and similar amount of heat. Thus cooling is critical and may be accomplished by some balance of water cooling and air conditioning. The latter demands even more power and it can become a deadly circle.
The data center demands massive areas of land. Often in millions of square feet. Inside these massive plants, besides power and computers is billions of mile of interconnection.
Needless to say they also consume computers and memory. More of both than is contemplated in today’s manufacturing world.
2 Disadvantages
Water: Water is used for cooling. Leaving aside the overall thermodynamic issues, massive amounts of water are needed. This is often from the aquifer. It takes drinking water away from humans. Worse, the massive areas covers by the centers means rainfall does not get to percolate down to refill the aquifers! That means a death spiral for water table stability.
Power: From whence do these entities get the power demanded. They can build their own systems which consume massive amounts of hydrocarbons, with the ensuing sequellae, or they purchase from the grid, which is neither prepared for such distribution nor priced so as to not over price consumer services. The devil is in the details, and as best as possible, they have failed to examine these.
Noise: Running computers and environment systems, not to mention power systems creates massive amounts of noise. Noise is both aggravating but it can cause significant harm to individuals. No one has yet to measure these effects but alas they may be the dominant factor in harm.
Pollution: All of the above result in increased pollution.
Traffic: These facilities demand human interaction. Thus significant increases in traffic and resulting negative effects will ensue.
Abandonment: Finally, as history has shown, all too often the initial exuberance and build outs are excessive resulting in bankruptcies and abandonment of facilities. This means buildings and equipment left to rot on viable land, pollution resulting from the toxic materials used in the facilities, and loss of value to the land and local environment.
3 Needs
It is worth considering the actual needs for such facilities and systems. Currently we have an almost faddish use by many. No real economic need has been demonstrated. If one were to compare this to the deployment of PCs or of the Internet, in those cases immediate economic value was provided. However with AI to the consumer, there are lots of gimmicks but no clear value. Would one trust AI to be my medical provider? Hardly. Would I trust it to be my accountant? No way. Who shows up to the IRS, my laptop? Won’t work. Perhaps as a customer services interface, they can’t get any worse than what they are now!
4 Recommendations
My recommendations are simply:
1. Locations should be distant from any human locations and any areas that would cause negative impacts on natural plants and animals.
2. Power should be self-sourced and penalties for excess carbon emissions.
3. Water usage should be at a minimum.
4. Noise levels should be regulated.
5. Funds should be held in escrow in the case of abandonment so as to remedy potential toxic effects.
