It is nothing new that today's data centers, especially those focused on artificial intelligence inference processing require large amounts of electricity and also water supplies to ensure their operations remain at optimal levels.
Recently, it was reported that Sam Altman, CEO of OpenAI said that he wants to develop a dedicated data center for the purpose of processing AI needs that has a power capacity as high as 5 GigaWatt.
If this figure is difficult to understand, it is the same amount of power produced by five nuclear plants at one time. It is understood that this 5GW data center consumes as much power as 100 data centers of the same size used for purposes other than processing artificial intelligence requests.
Another example is that this 5GW of electricity can power the entire city of Miami in Florida, and is enough to power the state of New York twice. In short, that much electrical power for the use of only one data center is very unreasonable. And Sam Altman is seen wanting to develop five to seven such data centers around the world for OpenAI use only.
The cost for the development of such a data center is also very expensive. It is estimated that one such data center, which will house around two million AI processing chips will cost as much as $100 billion. In comparison, Fortune reports that Google will soon spend $3.3 billion to develop two of their own data centers.
Another issue is regarding the power supply required to power this data center. It is not denied that many power supply companies can accommodate the consumption of 5GW, but this is only temporary. For OpenAI needs, it will use 5GW of power supply around the clock to process inferences received from millions of users who will use OpenAI technology.
This is seen as troubling because it shows that the technology sector is increasingly using environmental resources, and many of the promises to make their businesses zero carbon or zero pollution seem to be further away due to the energy and water consumption of artificial intelligence technology.