Let’s say that AI is wildly successful, and raises overall electricity demand 10%. That number is made up, and way above what k see anybody even proposes, as far as I can see.
How does that impact the energy transition? It means we need tk deploy new electricity generation faster. What’s the cheapest new energy? Renewable energy. What happens when we produce more renewable infrastructure? Future infrastructure gets even cheaper.
The faster and more we spend on new renewable electricity generation, the faster we accelerate the energy transition. And the faster that new renewable generation gets than the O&M of paid-off fossil generation assets, the sooner those assets will get retired.
Too many people think about the energy transition in reverse. We need to build build build as fast as we possibly can, and the faster we do it, the more money we save, the cheaper electricity gets, and the cleaner our electricity mix gets. We need a radical rebuild of everything.
If AI pours fuel on the fire of the energy interchange, Godspeed. Let’s develop those new electricity uses ASAP.
Man, I didn’t see the sub name, just the first bit of the headline. I thought we already had some AI supervillain.
No , LLM to me is power hungry, but people and ai may develop better algorithms and chips over time that are less power hungry. Short term it is nuts.
Long term, I think less power hungry AI will emerge like Verses Ai, and if successful in utility, may take a bigger hold of the market.
“Insatiable thirst for power” – Simple economics, companies will always select sites based on power costs. Cheapest form of power is solar. Global scale companies will always have data centers in the sun somewhere.
No, quite the opposite. It can actually be used as a Demand Side Response (DSR) tool to increase total installed variable (but predictable) renewables like wind turbines and solar pv and its profitability in a “build it and they will come” fashion. Opportunistic energy usage like server farms running AI models are a floor on energy prices as demand for AI can be queued for processing and run faster / slower according to energy prices and the marke price of “AI solutions”
Counter arguments: “but there’s more demand for “AI solutions” during night time” counter: fine, build it across the globe where the sun is shining and cheapest to build (Australia? South America? Africa?), the question and solution travels at near lightspeed on the internet.
5 comments
No.
Let’s say that AI is wildly successful, and raises overall electricity demand 10%. That number is made up, and way above what k see anybody even proposes, as far as I can see.
How does that impact the energy transition? It means we need tk deploy new electricity generation faster. What’s the cheapest new energy? Renewable energy. What happens when we produce more renewable infrastructure? Future infrastructure gets even cheaper.
The faster and more we spend on new renewable electricity generation, the faster we accelerate the energy transition. And the faster that new renewable generation gets than the O&M of paid-off fossil generation assets, the sooner those assets will get retired.
Too many people think about the energy transition in reverse. We need to build build build as fast as we possibly can, and the faster we do it, the more money we save, the cheaper electricity gets, and the cleaner our electricity mix gets. We need a radical rebuild of everything.
If AI pours fuel on the fire of the energy interchange, Godspeed. Let’s develop those new electricity uses ASAP.
Man, I didn’t see the sub name, just the first bit of the headline. I thought we already had some AI supervillain.
No , LLM to me is power hungry, but people and ai may develop better algorithms and chips over time that are less power hungry. Short term it is nuts.
Long term, I think less power hungry AI will emerge like Verses Ai, and if successful in utility, may take a bigger hold of the market.
“Insatiable thirst for power” – Simple economics, companies will always select sites based on power costs. Cheapest form of power is solar. Global scale companies will always have data centers in the sun somewhere.
No, quite the opposite. It can actually be used as a Demand Side Response (DSR) tool to increase total installed variable (but predictable) renewables like wind turbines and solar pv and its profitability in a “build it and they will come” fashion. Opportunistic energy usage like server farms running AI models are a floor on energy prices as demand for AI can be queued for processing and run faster / slower according to energy prices and the marke price of “AI solutions”
Counter arguments: “but there’s more demand for “AI solutions” during night time” counter: fine, build it across the globe where the sun is shining and cheapest to build (Australia? South America? Africa?), the question and solution travels at near lightspeed on the internet.