According to PricewaterhouseCoopers, self-described as a "multinational professional services network" and one of the four largest international accounting firms, climate change indicated by drought will put at risk a third of global chip production by 2035. The copper mining process takes about 86 times as much water by weight to produce one unit of copper, says the accounting firm.
This overlooks the fact that practically all copper that has been put to use in the past is recycled and that substitutions for copper have been found in some industries.
Presently, only the copper mines in Chile, which produce 7% of the world's copper supply, are threatened by drought but the future could be grim for the production of chips needed for cell phones and other consumer products as well as AI data centers.
So what would this mean? In the case of the microscopic copper conductors used in AI data centers basic economics tells us that the price of the chips would increase and that some data centers wouldn't be able to function economically. Or that technology would come up with a work around. Or that droughts wouldn't increase by 2035.
The AI data center drum beat, and the AI concept itself, doesn't seem to have an indicated end point. How many data centers does the country and world really need to function efficiently? How much valuable raw materials like copper and electricity can be devoted to what amounts to an enormous digital filing system?
While practically all Americans have a pocket-size computer that connects them to the internet, it seems to be a fact that the same is now true of AI, at least in its preliminary form. Will there be a choice necessary between copper as used for piping and copper used in AI chips and electricity pumped into data centers as opposed to peoples' home lighting and washing machines? What's most important?
No comments:
Post a Comment