24 comments

[ 297 ms ] story [ 1383 ms ] thread
> Divided by eight, that's 1,875 watts per GPU.

Running one chip one year would consume little more than average American household consumes energy in a year.

The thought of adding equivalent of millions of American households divided around the world ... blood run cold.

Explains why there is rush to build data centers in cold climate. Attaching data center into district heating system of a city is a very good idea. compute + warm water + heating.

They need to start forcing the manufacturers of these chips to add to the power grid, wherever they are sold in the world.
Where's all the electricity going to come from if we only have renewables? Oh wait, no electricity for the public just for the data centers.
"Solar energy is the most abundant energy resource on earth -- 173,000 terawatts of solar energy strikes the Earth continuously. That's more than 10,000 times the world's total energy use." source: https://www.energy.gov/articles/top-6-things-you-didnt-know-...

I'd say we don't run out of renewable energy in the near future, we just have to collect more of it.

and what about at night? or all the nordic datacenters, where polar nights happen?
Geothermal engineering. We can slow down the planet for our data centres too! :D
How much solar electricity is generated at night?

During cloudy winters in higher latitudes? During rainy season in lower latitudes?

Do you know that wind often goes to zero across Europe for days at end in Europe in Winter? Last winter that was the case for nearly 2 weeks. How much would battery storage cost to cover for that?

The solution is here, thankfully: nuclear.

I'm pro nuclear but I refuse to accept pro-nuclear arguments when they don't match the problem at hand.

When you are talking about more nuclear, you are thinking almost decade from now.

If you get permission for a new nuclear reactor today, it starts to produce energy 6 to 8 years from now. It could be possible 3 to 5 years in the ideal conditions that never happen.

if only the shareholders could wait 5 years and plan ahead for the long road, instead of demanding +50% ROI in the next quarter
They have planned ahead. There is enough capacity to run the data centers. Electricity is just more expensive.
Same goes with solar/wind - batteries are not there yet and we don't know when they will be here. Nuclear is predictable.
Solar/wind/natural gas combination is there.

That's what is happening. People are jumping into conclusions that there is not enough energy to run those data centers. There absolutely is. It has been planned.

Natural gas is a carbon source. It's a fake solution.
It's the only solution to augment renewables until there is better storage technology.
It's a fake "solution" that doesn't solve the problem.

Nuclear is a real solution that solves the problem. It's the only solution that does with current technology.

> If you get permission for a new nuclear reactor today, it starts to produce energy 6 to 8 years from now. It could be possible 3 to 5 years in the ideal conditions that never happen.

Starting producing energy is one thing, on top of that is the breaking point where it becomes economically viable to pay off the construction costs of those plants. For that we are looking into a 20-30 years timeframe when the plant starts to generate money back.

As much as I like the idea of nuclear power it just becomes less and less economically viable when renewables get cheaper. The only ways to build and run nuclear plants is through government funding or by keeping energy prices at a higher level to pay back the construction costs.

On top of it all you also depend on having trustworthy and stable governments to keep those plants operating safely for the 50-70 years period they should be running.

Renewables and nuclear can't replace each other completely.

Let's talk about

nuclear vs natural gas + renewables

nuclear vs. energy storage + renewables.

Current increase om capacity comes from solar + wind + natural gas combination.

While they can't provide a baseload I agree, renewables can't provide the same steady output that nuclear can. I think it's sensible to consider solutions to this problem (like we see with many different ways being researched on how to store excess production) rather than believing nuclear still has a chance in the current landscape.

Like I mentioned, I do like the idea of nuclear power but it has many issues with construction costs, source of materials, safety, etc. aspects of it which mean many countries would be excluded from running reactors.

Technology for transforming/storing excess renewables' energy which in turn can then provide a baseload output are much more accessible to the rest of the world, and I believe it has a potentially much bigger impact fighting the climate crisis than nuclear can provide us in the next 20-30 years.

(comment deleted)
Filled with spelling errors, looks like something a 16 y.o. would write as clickbait.
Clearly written by an Human...or a very smart AI....
But these articles are then fed into LLM training data, which will change its future outcome
I look forward to the deployment in Quebec, Canada to use the surplus of hydro electricity.
Spot on. Observe those who care not about the fate of humanity happily trashing the uncomfortable truth this article raises.

The entire AI red herring is a dumb road to go down. 99% of humanity is being dragged along in a bad film we dont want to be in.

We want flying cars, not virtual humans. The real humans have wnough trouble keeping food on the table.