Offshore wind runs about $5/MW of dataplate energy according to a report today on the BBC about a major project that's just been cancelled -- £5.4 billion ($8.6 billion) for an 1800MW capacity wind turbine array (Three hundred 6MW units). Offshore gets a little bit better capacity factor than land-based units, maybe 30% so that's 540MW average over a year or so. Expected lifespan of offshore wind turbines is about 15-20 years but the industry has been quite coy over failure rates and actual operating costs of offshore wind turbines. Decommissioning costs don't seem to be mentioned but for an array of that size it could be hundreds of millions. The strike price, the cost the grid would be obligated by law to buy this wind energy in at was set at £145 per MWh; apparently this wasn't enough of a return for the folks proposing the wind farm and the project has been abandoned.
The two EPR1400 nuclear reactors planned for Hinkley in England will produce about 3200MW baseload at a capacity factor of about 90% or so, averaging about 2700MW per annum, expected construction costs about £10 billion ($16 billion) and a working lifespan of 60 years minimum, probably more. The grid will buy this reliable and very predictable baseload at an agreed strike price of £90 per MWh assuming the project actually goes ahead, that price to include paying into a decommissioning fund on each kWh sold as well as covering the cost of fuel, operation, mid-life refurbishments etc.
Those are the grid-supply pricing targets an array of small reactors will have to meet; it would take more than 35 NuScale 45MW reactors to deliver the same generating capacity as an EPR1400 which costs (according to the Chinese who are closest to completing their EPR builds on time and on budget) about £5 billion a pop so each NuScale reactor needs to cost less than £150 million for them to be even marginally an economic prospect.
The $226 million grant from the DoE isn't to build a reactor, it's to fund further development and help NuScale to get a licence to build and operate a prototype in maybe ten years time -- that process could cost a billion dollars in itself.