Electricity Demand = Natural Gas Demand

[dan_s]

I just heard about this six weeks ago. It is a getting a lot of coverage.
Natural stupidity
IRINA SLAV < Very sharp lady that writes for OilPrice.com
MAR 27

“Insatiable”. “Staggering”. “Voracious”. “Obscene”. “Ravenous”. “Shocking”. These are not adjectives describing EU leaders’ appetite for more central government, although they could just as well be used to that end. No, these are adjectives describing something else: AI’s energy consumption.

Reporters have been comparing this consumption to that of small countries. One researcher calculated it at a minimum of 85.4 TWh annually — based on annual NVIDIA deliveries of 1.5 million AI server units.

I won’t even try to pretend I know what “AI server units” means. Software for me is a closed, locked, and hidden in a dungeon book — and I take good care to keep it there.

But when I see alarms blaring about electricity consumption and where are we going to get all that electricity to power our new artificially intelligent toys that are going to make every industry so much more efficient, read profitable, I start paying attention.

Especially when the transitioners show signs of hating it. Because, apparently, powering AI will interfere with the transition. Because, apparently, there’s nowhere near enough non-hydrocarbon energy to meet that demand. It sounds like a perfect example of a perfect storm. It’s delicious.

In October last year, the researcher who calculated AI’s energy consumption gave an interview to Scientific American. In it, Alex de Vries, a data scientist from the Dutch Vrije Universiteit, made some observations that even an anti-nerd like me could easily understand.

He said, for instance, that if Google turned its search engine into an AI-powered machine its electricity consumption would jump to a level comparable to the consumption level of Ireland. Now, Ireland may be a small country geographically but in terms of electricity consumption, Ireland sets the needle at 30.63 TWh.

The example is, thankfully, theoretical, as De Vries explains: “Now, it’s not going to happen like that because Google would also have to invest $100 billion in hardware to make that possible. And even if [the company] had the money to invest, the supply chain couldn’t deliver all those servers right away.” Definitely a burden off my shoulders, this.

The point that the researcher makes, however, stands. Adding AI to everything you can just because you can will be a drain on electricity supply — and add to the already pretty impressive levels of additional demand expected to come from the electrification of everything, even as EV drivers regain their composure from swooning after realising their tyres wear out amazingly fast.

Equally amazingly, no one appears to be disputing the fact there is no physical way to respond to that surge in electricity demand by building more wind and solar, and, incidentally, did you hear about the hailstorm that destroyed a massive solar farm in Texas? This is precisely why environmentalists are going to start hating AI very soon while the oil and gas industry has good reason to love it.

“We’re not going to build 100 gigawatts of new renewables in a few years. You’re kind of stuck,” former Energy Secretary Ernest Monitz, who is no fan of natural gas, told the WSJ last week. He also said that because this is impossible, power utilities would need to expand their nuclear, gas, and coal generation capacity. A nightmare scenario for the those averse to hydrocarbon molecules. < IMO natural gas as the "clean hydrocarbon" will be the only acceptable near-term solution.

The reaction, so far, is underwhelming, featuring weak attempts at irony by citing Conoco’s CEO “licking his chops at the thought of all the lovely profits his company expects to make by supplying fossil fuels to thermal generators.”

The word lovely appears to be the only one the author of the same reaction piece in CleanTechnica knows as an irony tool as he uses it again to refer to “all that lovely electricity” that “might hasten the collapse of the Earth’s environment due to ever increasing average global temperatures.” Lovely.

My outtake from this is that electricity and total electrification are only truly lovely when coming from the right sources. When they’re not, electricity is no longer lovely. They’ve even interviewed all-time transition star Bill McKibben as spreading wisdom and if you think I’ll spare you the quote, you are gravely mistaken. Here it is:

“In a rational world, faced with an emergency, we would put off scaling AI for now. The irony, of course, is that’s it’s often been touted as a tool to help solve climate change. But we have the tools we need — plain old intelligence gave us cheap solar panels.” Guffaw at will. Because AI is accelerating the energy transition, say industry leaders.

Here’s how, according to one senior Shell exec, as quoted by the FT: “AI will optimise the efficiency of energy systems by “reducing the amount of power needed to be generated”, he adds. It could also create “entirely new low carbon-footprint energy systems” and enable suppliers to monitor greenhouse gas sinks.”

So, AI is helping reduce specific corporate emissions by “reducing the amount of power needed to be generated” but it will raise overall global emissions by consuming so much electricity that it would need to be generated from hydrocarbons. How’s that for a wonderful transition riddle? It gets better, too.

Apparently, AI could help grid operators “plan supply more efficiently” by enabling “a more strategic, whole-grid view of supply”. This whole-grid view, the FT article tells us, would improve balancing the “use of different sources of low carbon generation in a dynamic environment”.

The way I translate the above is consumption management. There is no viable way of “planning” supply from wind and solar with any approximation of reliable accuracy, so it would be down to consumption to provide that “balance” of different sources of low carbon generation in a dynamic environment.

In support of this interpretation, This Is Money reported this week that smart meters will be used to manage people’s electricity consumption in such a way as to penalize use during peak demand periods with higher prices while rewarding off-peak consumption with lower prices. The goal: using more wind and solar when they are available.

That’s about the only way to make the transition work, really, with “alerts to tell people they can turn on household appliances if there is an abundance of cheap wind and solar energy.” No doubt the issuers of these alerts would be happy to use AI to make it all more accurate, efficient and whatever else hype word gets slapped on to AI these days, despite the higher energy cost. A dynamic environment, indeed.

Dynamic, in case anyone’s missed it, has become a euphemism for “utterly chaotic and unmanageable”. But AI will introduce order to the chaos — at a not insignificant energy cost, which will probably add to the dynamism of the environment. It would also add gas-fired generation capacity. For backup — or not just.

Per the WSJ: “Toby Rice, CEO of giant U.S. natural-gas producer EQT, said tech firms building data centers are inquiring about buying gas from EQT. Rice said he got the same two questions at the conference: “How fast can you guys move? How much gas can we get?””

The rush is on to insert AI everywhere it can be inserted. The official arguments are along the lines that it would improve everything it touches. Far be it for me to dispute this given my lack of any in-depth knowledge about AI but instinct whispers that AI is very much like the new toy that everyone wants to have and play with.

The excitement is so massive — and I am unironically sure it is at least partially justified — that considerations about the cost of that play are taking a back seat. Or they are being dismissed with the argument that AI will be good for the transition, despite its massive cost in terms of electricity consumption.

So, on the one hand, we have the AI fan camp, for whom the benefits of using the technology outweigh the costs in terms of energy consumption and the inevitable need to use more hydrocarbons to produce that energy. For this camp, it’s an end-justifies-the-means situation, or, as I put it earlier, “Some emissions are more equal than others”.

On the other hand, we have the climate catastrophism camp. It’s not impressed by AI. It doesn’t like its carbon footprint. It will go to war with the fan camp pretty soon, if McKibben’s reaction is any indication and it probably is. The tireless transition advocate and connoisseur of the aesthetic of wind turbines and solar panels has suggested we say no to AI “at least until the emergency subsides.”

This call, I predict, will be taken up by many once they process the fact that AI, which promises to reduce emissions, will actually generate many more emissions. This would pit the activists against their biggest fans and, not to put too fine a point on it, financial backers from Big Tech. Now this would be a show worth watching.