Three Mile Island was the only major commercial nuclear accident in U.S. history. Nuclear power in the U.S. has generated

24,196,167 GWh between 1971-2015. At an

average price of 12 cents/kWh, that's $2.90354 trillion. So the

approx $3.4 billion in cleanup and lossses from TMI is 0.117% of that. Or in other words, at a retail price of 12 cents/kWh, the historical cost of cleaning up nuclear accidents in the U.S. is 0.014 cents per kWh.

In contrast,

subsidies for different energy sources are 23.1 cents/kWh for solar, 3.5 cents/kWh for wind, and 0.2 cents/kWh for nuclear. (Tables ES4 and ES4. Solar received $4.393 billion in subsidies while generating 19,000 GWh. Wind received $5.936 billion while generating 5,936 GWh, and nuclear received $1.66 billion while generating 789,000 GWh.) That's right. The subsidy for solar is 1650x more expensive than cleaning up nuclear accidents. The subsidy for wind is 250x more expensive.

Nuclear decommissioning costs are already paid for by the

NRC's Financial Assurance fund. A portion of the revenue from electricity sales are placed into this fund.

The problem with insuring nuclear plants is just a quirk of statistics. The more times you roll the dice, the narrower the bell curve becomes and the more predictable the average outcome. e.g. A 1d100 has an equal chance to produce any result between 1 and 100 - the probability distribution function is a straight line. 2d50 produces a triangular PDF, with the values in the middle tending to be more likely. 10d10 produces an even more compact PDF - a narrow normal curve with results in the middle much more likely than the extremes. And 100d0.5 will always produce 50 - its PDF is just a single peak in the middle.

This is a problem for insuring nuclear plants - because they produce so much energy you don't need very many of them. Whereas there are thousands of coal plants, and (potentially) millions of solar installations, there are only operating 100 nuclear plants in the U.S. So insuring a nuclear plant represents a greater risk for the insurer. Even though the mean outcome will be that there is 1 accident every 30 years, the chance of a 2nd or 3rd accident is still significant and the amount the insurer has to pay out may easily surpass how much they've collected in premiums if they assume the statistically most likely outcome of a single accident.

The insurance company's response is to increase the premium to also cover that 2nd or 3rd event even though they're unlikely. In contrast, with thousands of coal plants they can be much more confident that there will be (say) only 10 accidents every 30 years, and 20 or 30 accidents is extraordinarily unlikely. So the premiums can be lower, even if the average risk (mean) is exactly the same. If there were some way to build thousands of small-scale nuclear plants instead of 100 large ones, private insurance wouldn't be a problem. You get around this problem by creating the largest insurance pool possible, which in this case would be nationalized insurance covering all 100 nuclear power plants.

Statistically, per unit of energy generated, nuclear power is the safest power source man has invented.