They will also want reliable power, which seems like justification enough to keep their options open. A “phased, modular approach” sounds like it would give them options. Maybe Microsoft could install solar and/or batteries later and use their natural gas turbines for backup?

With enough batteries they might get through the night, but seasonal shortages are much harder to handle that way.

I bet a carbon tax on data centers would be popular if the Democrats get back in.

▲

Retric a day ago | parent | next [-]

You can generally solve seasonal shortfalls cheaply by simply adding more solar / wind until you no longer have a seasonal shortfall just a massive surplus in a different part of the year.

Put another way if you want to store power say 1GWh for the worst month, well a solar panel provides a lot of power over even the worst month. 20MW of solar panel averaging even 40MWh/day * 30 days = 1.2GWh and cost way less than 1GWh of batteries.

Near the arctic circle Wind fill the same niche.

▲

buzer a day ago | parent | next [-]

> Near the arctic circle Wind fill the same niche.

What do you mean? Long periods of calm weather during cold temperatures is not unheard of in Finland and does cause issues at times due to amount of wind energy that has been built in recent times.

▲

Retric a day ago | parent [-]

Long periods of low output isn’t 0 energy output over a ~3 month long winter.

Beyond that wind for each location is somewhat seasonal so if you’re doing this for winter energy you pick areas with better wind in winter.

▲

buzer a day ago | parent [-]

Not 3 month period, but looking at e.g. capacity factor over 5 day period can be under 3% and even 10 day period can be under 5%. Capacity factor for whole year is around 30% (22TWh produced with 9433MW in 2025, rounded it up since some capacity came online during 2025). That's a lot of extra power or storage.

	▲	Retric a day ago \| parent [-]
		We’re optimizing for cost here, nothing else matters. You can always turn off a wind turbine. If you’re proposing seasonal storage as a viable option then cutting total storage needs by 90% (to handle a 10 day lull vs 90 day gap) is inherently cheaper. Hell traditional hydro is well suited for such situations as annual output and storage is limited but nontrivial. Doubling output for 10 days isn’t nearly as problematic as trying to cover a 90 day shortfall.

▲

_heimdall 15 hours ago | parent | prev | next [-]

You can't replace batteries with more solar, additional solar only increases the energy produced while the sun is out.

That can help on overcast days, or the middle of winter when days are shorter, but you still need battery capacity to run for stretches when there is no sun producing energy.

For residential its often recommended to have 3 days of storage capacity + the option for a backup generator. I'm not sure what the recommendation is for industrial projects, but I expect they would still need a backup generator. If they already have to put in a generator to handle the full load of the server farm, and they have cheap local gas, why bother with the solar? At a minimum you have both, but the generator (or power from a contracted gap generator) is a given.

▲

Retric 13 hours ago | parent [-]

Nightly storage is cost effective because you use it 365 times a year. Seasonal storage is solving a different problem.

▲

_heimdall 12 hours ago | parent [-]

Nightly storage depends on being able to recharge to full capacity every day. That isn't a risk they could possibly accept, they would have to have an alternative source that could sustain when the sun doesn't come out or equipment fails.

▲

Retric 11 hours ago | parent [-]

> to full capacity every day.

A common misconception you don’t need to hit 100%, you need enough energy to make it either from that day or from prior days.

At grid scale daily production is never 0 and it’s never 100%. You can guarantee a surplus all 365 days a year, it’s just a cost vs benefit function, that’s however different from always fully charging your energy storage. If hypothetically the minimum was 99.9% of nighttime needs the odds you cover that gap the next day is extremely high to the point where a little extra storage makes sense vs aiming for 100% every day.

So now you’re just trying to optimize something for minimum cost. Utilities do this all the time with traditional generation you have random equipment failures and shifting seasonal demand. Thus they optimize maintenance schedules around seasonal demand etc.

▲

_heimdall 4 hours ago | parent [-]

Not a misconception, just a discrepancy in what we were describing. I well understand that you shouldn't hit 0% for battery health and that the last 10-20% charge takes much longer than the middle of the charge cycle. I was responding to the idea of only having capacity for a night and using it 365 days per year, I took that to mean a full charge/discharge cycle every day since otherwise you have more capacity than a single night and don't use it fully every day.

Regardless, that buffer doesn't make a big difference in the topic here. My panel regularly show 0Wh on rainy days, PR effectively 0 as they may be getting only a few percent of real capacity.

How do you propose industrial scale would allow a series of arrays could both not be oversized on sunny days and cover usage on cludy days?

If I'm misunderstanding your argument I apologize. I just don't see how a data center could possibly be okay with having only a single day worth of storage and generation without having backup power scaled to cover their full usage demands.

	▲	Retric 4 hours ago \| parent [-]
		Edit: Grid scale solar is its own thing for many reasons, one of which is not sticking things in a roof tends to dramatically lower risks and costs. But another is they are significantly more optimized. > could both not be oversized on sunny days and cover usage on cludy days? That isn’t a problem that’s worth trying to solve. Grid solar power is so cheap you can waste 70% over a year (even more on sunny days) and in many areas still be paying less than coal per kWh. Ultimately the grid doesn’t care about being oversized for specific days. There’s a wild difference in demand on sunny day in the summer vs a day that’s 70f. When it was 70f in the 1970’s they didn’t go out and destroy nuclear reactors or coal plants because they are unnecessary in May. That’s valuable infrastructure even when not in current use. The only metric grid operators care about is does this grid design need the demand and what does it cost. Peaking power plants exist to fit edge case demands even if not used for months at a time, “peaking solar” isn’t a term anyone uses but solar + batteries can fit that niche.

▲

skybrian a day ago | parent | prev [-]

I don’t believe the “cheaply” part. You end up building storage that only gets used for one cycle a year, so the capital cost is very high per watt.

	▲	Retric a day ago \| parent [-]
		I think you misunderstood, I am saying 20MW of solar power is less expensive than 1GWh of energy storage. Nothing in my post suggests 1GWh of storage is cheap.

▲

cratermoon a day ago | parent | prev [-]

> They will also want reliable power

Given the state of the power grid in Texas, this could be the most important consideration. Why? Texas is not connected to the national power grid, and only electricity from plants operating in Texas is available. The last winter and summer, demands on the grid have severely stressed, as reported in many places. In 2021 there as a state-wide crisis and almost a complete failure. https://en.wikipedia.org/wiki/2021_Texas_power_crisis

▲

chasd00 a day ago | parent [-]

These power gen facilities wouldn’t be connected to the grid, only the datacenter.

	▲	cratermoon 12 hours ago \| parent [-]
		Exactly. The data center gets it's own power, isolated from the flaky grid, and never has to deal with statewide crises.