[countvonbalzac]

Same - take lightbulbs for example. Thanks to LEDs, the amount of energy you need to generate X amount of light has reduced considerably, but we still have as much if not more lighting than ever. And it's not like our GDP is suffering due to a lack of sufficient lighting, at a certain point there's no gain to productivity gained from having another lightbulb. Same thing can be said about cars, CPUs etc.

[pseudosavant]

It is a curious assertion for sure, but the next wave of AI will be throttled by a few factors: chip fab capacity, water supply to cool data centers, and electricity to power the chips/servers. Look at what is happening with xAI in Memphis where they are illegally running a dozen turbine generators to power their new AI data center since they can't get enough supply off the grid.

[skybrian]

Costs are going down a lot due to algorithmic improvements. But maybe that just results in increased usage (Jevons paradox)? When there are conflicting trends, the future is pretty hard to predict.

Are any AI companies making money now? Losses can’t go on forever.

[svnt]

It may be throttled by energy supply and water supply at a desirable site, but not by country-level energy costs and water costs, which is all this blog post is looking at.

[bobthepanda]

Depending on where you are this is a national level problem; Ireland and the Netherlands are not approving new data centers

[Havoc]

Think it’s meant more as a broad generalization than something that is always true.

Many physical things take pretty fixed amounts of energy. Eg heating a liter of water.

[morepork]

Even with water heating moving to a heat pump lets you do it using less input energy because you are taking heat from the environment

[nicoburns]

> Think it’s meant more as a broad generalization than something that is always true.

I think you're right, but also that the majority of the problems with the worlds economies (in the richer nations) are because of similar generalizations, and as such I think it important to rebuke them.

Having more cheap energy available is good (all else being equal), but optimising for higher energy usage is absurd.

[bee_rider]

The economy stopped getting better because I started taking shorter, colder showers.

[HideousKojima]

Or refining aluminum, which uses something like 1.5% of all US electricity generation.

[beachy]

The Tiwai Point aluminum smelter uses 13% of New Zealand's electricity [0]

It's overseas owners are constantly playing hardball with the country over the price they pay. Feels like every year they threaten to shut the smelter down unless they get better electricity rates.

[0] https://en.m.wikipedia.org/wiki/Tiwai_Point_Aluminium_Smelte...

[kragen]

it would be much more, but the us imports 130% of the aluminum it consumes from places with lower energy costs: https://pubs.usgs.gov/periodicals/mcs2022/mcs2022-aluminum.p...

[bobthepanda]

Bahrain and UAE being big aluminum smelters is a bit surprising.

[cameldrv]

Exporting aluminum is basically exporting electricity, except aluminum is easy to ship, costs very little to store, and has an indefinite shelf life. For places with a lot of natural gas and no pipelines to export it, it’s often easier to export aluminum than liquified gas.

[kragen]

if you're burning it in aluminum-air fuel cells, it can be literally exporting electricity. right now that isn't a commercial-scale activity, but possibly it will become profitable in the coming years for places with a lot of solar power and no hvdc lines to export it

[HPsquared]

Iceland too. It takes a few workers and a lot of energy. Very sensitive to market conditions though. Peaky.

[kragen]

for some economic activities, energy is not a limiting input; you are implicitly referring to economic production enabled by electric lighting, such as office work, and indeed energy has not been a limiting input for that for at least a century. reducing the cost of energy will not result in more gdp in those sectors

for other economic activities, such as solar panel production, aluminum production, and neural network training, energy is a limiting input. reducing the cost of energy will result in more gdp in those sectors

[Retric]

Dropping energy costs lowers costs in every sector but rarely by that much.

> aluminum production

Dropping energy costs an by 75% only drops smelting prices by about 30% and finished goods by even less.

[kragen]

there's usually a long lag between a drop in the price of an input and the eventual impact on the price of the outputs, because part of the effect is mediated by the adoption of innovations that use more of the newly-cheaper inputs and less of the still-expensive inputs

to take one example, the last time we got access to a major new source of energy was something like watt's steam-engine in 01776. one of the effects of this was the widespread replacement of steel cans (which hadn't been invented in 01776) and glass bottles with aluminum cans in the 01970s, 200 years later. another was the replacement of travel by ship with travel by air, also about 200 years later. the delay is because many intermediating innovations were required, for example, in the aluminum-can case:

- the discovery of electrolysis;

- the discovery of aluminum;

- the discovery of canning;

- the hall–héroult process;

- improved aluminum alloys that permitted the use of 100μm-thick cans;

- the invention of deep drawing;

- epoxy liners that made aluminum cans chemically stable to acidic contents such as coca-cola;

- long-distance trucking which increased the cost imposed by heavier glass bottles.

[njarboe]

We started to access nuclear power as a new source, but then stopped for quite a while. It looks like things are starting up again. We'll see.

[kragen]

the issue with nuclear power is that the humans don't yet have the technology to exploit it economically; at their current primitive level it's uncompetitive with other sources of energy. like printing 1000 years ago or heron's aeolipile

[chickenbig]

> humans don't yet have the technology to exploit it economically

$2,500/kW of capacity isn't too expensive, given the alternatives.

https://www.latitudemedia.com/news/catalyst-the-cost-of-nucl...

The recent batch of 11 reactors authorised by China are perhaps 2.8B USD each for 1.1GW plant (plus a high temperature gas reactor).