There are a lot of posts here pushing batteries.
Batteries are an expensive solution that doesn't scale well at the grid level. It is useful for grid stability (fast frequency response) but simply a non-starter when you're dealing with national grids.
Batteries are an added cost to the system, without producing more electricity, and as a result prices will go up.
A far cheaper source of flexibility is Demand Side Response. Particularly data centres that are willing to be market actors. Compute can happen anywhere, so it should happen where the wind blows and the sun shines. It is cheaper to transmit bits than Megawatts.
Demand side response drives up costs a lot. You end up with expensive, rapidly depreciating capital equipment sitting idle and not earning any revenue. The same problem applies whether the equipment is a GPU cluster or aluminum smelter. If we're going to have a modern industrial economy then we need to have enormous quantities of cheap electrical power available 24 hours a day.
Long distance high voltage transmission lines can help to an extent but create the same sort of concerns about dependence on unreliable foreign countries as fossil fuel imports.
> Batteries are an expensive solution that doesn't scale well at the grid level.
I'd like to see the reasoning behind why they don't pan out. LoFePO4 have dropped to $60/kWh in China. At 3,000 cycles that means they add about 2¢ to every kWh they store.
We don't get that cheap price where I live of course, but they being installed at a rapid pace now. I think most are being installed "behind the meter", which means they are being installed by people who pay retail. That's happening because paired with solar, they've dropped below the break even point at retail prices. Grid scale needs roughly another factor of 3 price drop to hit the same point. If CATL's $10/kWh sodium batteries that get 10,0000 cycles pan out, it will drive the price down by another factor of 10.
Your "demand" side response arises naturally with batteries. Those who can do without the power simply won't buy one. Or if they can get by with only a little emergency power, they buy a small one.
I experienced that first hand. I owned a 4.8kWh battery a while ago. That is by any definition is small. It costs about the same as a generator at today's prices (it didn't back then). A flood caused power to be cut off for a week. We only fired up the generator once, before discovering we could reduce our usage to what a small battery and a 6.6 kW solar array could cope with, even in the very overcast conditions that accompany a heavy rain event.
Demand side management is a nice concept, but it is neither free nor a cure-all:
It has real costs because it limits the utilization of involved infrastructure and is simply not feasible for a lot of industries. It does not help when residential demand exceeds the available supply either.
The most practical solution will probably be a mix of overprovisioning (especially considering how cheap solar panels have become), battery storage and fuel powered fallback, with the balance shifting as long as batteries and panels get cheaper.
Grid level battery storage is already coming online at scale (e.g. https://www.ess-news.com/2025/08/18/statera-energy-powers-up...).
LiFePo cells are already down to ~$60 for 1kWh (8000 cycles), which is pretty palatable for a lot of applications and prices still trend down.
Demand Side Response can be enabled by batteries and can save money.
For example, the OVO Charge Anytime tariff provides EV charging at just 7p kWh [1]. Average kWh cost is 26.35p/kWh[2]. From the linked case study:
> £7.7m/€9m total customer savings
Once Vehicle-to-Home and Vehicle-to-Grid is more widespread the savings will be even greater [3].
1. https://info.kaluza.com/hubfs/Charge%20Anytime%20EU%202024-0...
2. https://www.nimblefins.co.uk/average-cost-electricity-kwh-uk
3. https://info.kaluza.com/hubfs/What%E2%80%99s%20next%20for%20...
A huge portion of compute is triggered on request, so there isn't that much ability to time shift it. A build was just kicked off because I merged some code. In theory, that could happen overnight. In reality, changing the delay from 20 minutes to 12 hours would be unworkable.
On the other hand, when I commit code at 4PM on the West Coast it can be handled on a server in a place where there is low electricity cost.
There's not really a choice in replacing fossils with renewables&nuclear if we want to mitigate the climate catastrophe.