[yostalex]

Let's evaluate some basic constants.

Replacing fossil fuels with renewables is a shift in the vulnerability vector. The issue isn't even that China controls the production of solar panels and batteries. Production can be launched domestically. China controls 70-90% of refining - the processing chains of critical minerals (rare earth metals, polysilicon, lithium).

Renewables work perfectly for low-density consumers (residential sector, regular commerce). For heavy infrastructure, this won't work.

For example, let's look at AI data centers. AI data centers consume gigawatts of dense energy. Renewables are low-density energy. The problem comes down to spatial energy density (Watts per square meter — W/m²). A server rack for AI training consumes up to 40-100 kW. Solar and wind energy are diffuse (scattered) sources. Their density is about 5-20 W/m². A hyperscaler data center is a concentration of colossal energy in a minimal area (hundreds of megawatts per building).

Training LLM models cannot be interrupted when the sun goes down or the wind dies down. AI requires 24/7 baseload (base generation). The capacity factor of solar is 15-25%, wind — 30-45%. Batteries can smooth out peaks for 2-4 hours, but cannot provide seasonal or multi-day baseload.

Where do we plan to build solar and wind parks? - In deserts and offshore zones. This will require a radical expansion of the grids. We will run into a copper deficit (and things aren't smooth with aluminum either).

Long-term structural capital will go into nuclear energy, gas generation (as a backup), and copper/uranium mining.

[frm88]

https://www.reuters.com/world/europe/iaea-says-drone-targete...

May 3 (Reuters) - The Russian-held Zaporizhzhia nuclear power plant in southeastern Ukraine informed the International Atomic Energy Agency (IAEA) on Sunday that a drone had targeted its external radiation control laboratory.

This is not the first event targeting nuclear plants/laboratory. Given the current geopolitical situation, nuclear plants are in danger of being targeted on purpose and the damage would be considerable.

[yostalex]

If we look at the situation through the lens of news headlines and tail risks in an active war zone — I completely agree. Drones over the Zaporizhzhia NPP do indeed create a powerful media picture.

But let's look at how Big Tech assesses the risks. Hyperscalers employ top-tier analysts; they perfectly see the difference between media noise and actual damage.

Attacking an NPP is fundamentally unprofitable. It’s an asset worth tens of billions of dollars, and any rational capital (just like a state) will protect it. What we’re seeing right now — and I’d like to point out this isn't the first time — are strikes on non-critical external nodes. It’s a tool for informational pressure, not an attempt to trigger a second Chernobyl.

In reality, a direct strike on the reactor's containment is a taboo. In military doctrine, it’s equated to the use of radiological weapons, which instantly triggers protocols for uncontrolled escalation. None of the key actors want that.

Has Big Tech already poured money into nuclear? Fact? Fact. Not because they’re blind to geopolitics, but because the math of deterrence works: the risk of deliberately destroying an NPP equals the risk of starting World War III, at which point nobody is going to care about the fate of AI investments anyway.

Let's also take this fact into account: the risk of an attack on an NPP in an active war zone (Ukraine) is simply not comparable to the risk for an NPP in Pennsylvania or Texas.

Personally, I have nothing against green energy, but economics and physics are harsh realities.

[frm88]

In reality, a direct strike on the reactor's containment is a taboo. In military doctrine, it’s equated to the use of radiological weapons, which instantly triggers protocols for uncontrolled escalation. None of the key actors want that.

I'd love to trust this, I really do, but there are political leaders publicly threatening to erase civilisations. I live in Europe. It is still forbidden here to harvest wild mushrooms. Dead animals are burned, not eaten. To us, this is a possible reality

[dalyons]

> Long-term structural capital will go into nuclear energy

nearly zero capital will go into nuclear energy, unless it can be made a lot more affordable. It is structurally completely uncompetitive, and uninvestable without massive state backing.