Artificial Intelligence requires a phenomenal amount of power. The more humans rely on AI tools, to avoid any actual thinking, the more it becomes an issue. According to one expert, the U.S. electrical grid “is not prepared to meet the demand placed on it by the growth of artificial intelligence in work, social, and daily life.” Your monthly electrical bill is about to increase eight times.
More power, Mr. Gates
The robots are demanding power. They’re starving for more megawatts. It’s getting to the point where our national supply of dilithium crystals has melted away to nearly nothing.
Former top federal energy regulators are warning that the human race is in for trouble if something isn’t done about it now.
Researchers and experts have been whining about it for months. Sounding “the alarm over the amount of energy data centers focused on AI use to house and process data swiftly.” The end consumer always ends up paying the cost but casual AI users would faint if they saw the power bill.
“These facilities are home to tens of thousands of computer servers consuming massive amounts of energy.” They don’t leave lights on when they’re not in the room, that’s for sure.
The teeny-tiny facilities “may only use between 1 and 5 megawatts of power” but the big ones need their very own nuclear reactor. That’s not an exaggeration. They’re pulling Three Mile Island out of mothballs to keep the processors running for Microsoft.
Large AI data centers “have been known to consume over 100 megawatts.” With just one of those megawatts you can run up to 900 homes for a full year. They use 100 times the electricity every month.
It’s not happening
“Oh, gosh no,” Tony Clark gasped when he heard the question. He had just been asked if America’s power grid could handle devoting nine percent of production to the data centers. It’s going to be a big issue because that’s the projected need by 2030.
That may seem like a long time from now but getting electricity generation online requires years of planning, construction and regulatory approval.
Along with Microsoft, NVIDIA and Meta are fighting it out “to lead the way in the AI revolution.” To do that, they all need to develop bigger and better data centers. That means a lot more power to make them work.
Experts are nervous it’s going to crash the grid. As Clark points out, “that amount of load coming online that rapidly would be very, very difficult to serve in the short term.”
Someone needs to figure it out. Pronto. Marc Spitzer agrees that the grid can’t handle the strain. “Not in its current state. But if we plan now, I think we can be going forward.” PG&E customers saw rates raised by 17% at the start of this year to pay for more new equipment. Not only that, there’s an even bigger shocker.
“PJM Interconnection, the largest regional transmission organization monitoring a multistate grid in the United States, is increasing its costs from $28.92 per megawatt-day to $269.92 per megawatt-day starting next year.” Now that you’re back in your chair, after falling out of it, yes that number is correct. “This is a more than 800% increase that is expected to trickle down on consumers’ energy bills.” Just remember how helpful that last ChatGPT session was when you pay your power bill.