The headline arguments for building a new generation of nuclear power stations are familiar. The main lines are that nuclear is carbon-free in operation and that it might improve security of supply.

The security argument is weak. A decision to proceed might be made in 2007, followed by a commercial choice of reactor and supplier, a safety licensing process, a public inquiry and a period for construction and commissioning. If all these were to run smoothly, the first power from a new reactor might be produced around 2018 - or more likely around 2020. A series of reactors through the 2020s might significantly improve security of supply, but so might other measures that might be at least as cost-effective.

The climate change argument is, in principle, much more persuasive. Over a time horizon to 2030 and beyond, nuclear could make a significant dent in UK carbon emissions. But an important question is how nuclear power might get financed and built, given that no direct government money will be involved.

Nuclear is a large-scale technology, requiring large unit sizes and substantial numbers of reactors if costs are to be at their lowest. This is a serious inflexibility. Nuclear power is an extreme case of 'lumpy' investment.

The nuclear industry argues that it is worth taking the nuclear road only if there is a commitment to eight or 10 reactors. These would generate about 10,000 MW, equivalent to about 20 per cent of peak electricity demand. This represents some £15 billion or more.

This would affect investment in conventional gas-powered generation, which competes directly with nuclear. If markets expected a major programme of nuclear investment, they would be unlikely to commit to any further gas-based investment.

The worst-case scenario following a commitment to nuclear new-build would be a sterilisation of non-nuclear investment while the nuclear programme itself stalled. Such a scenario is far from a remote chance - the last time a UK government committed to 10 nuclear stations (Margaret Thatcher's in 1979) only one station was built, Sizewell, and then only after 15 years. If that were to happen again, security of supply would substantially worsen in the 2010s.

The risks attaching to nuclear investment are several. There are three types, corresponding to the nuclear life-cycle - construction risk, electricity market risk and decommissioning and waste risk - plus a more general and pervasive political/regulatory risk.

In the face of such risks, the cost of capital for nuclear projects will be higher than for conventional projects. The premium over a low-risk rate cannot be determined in the absence of detailed project plans but an inflation-adjusted cost of 10 per cent is likely on early projects. For a capital-intensive project such a relatively high cost of capital is a serious handicap.

None of this means that a new nuclear programme could not proceed. But the benefits are delayed and the inflexibility pronounced. The risk for the nuclear industry is that if it argues for 10,000 MW or nothing, it may get nothing. If nuclear is to have the chance to make a serious long-term contribution to climate change, a more flexible and incremental case would seem better than current proposals.

Observer 4th December 2005