One of the main issues in view of the realization of a DEMOnstration fusion reactor is the availability of a sufficient external supply of tritium (T) to start operation. T is an unstable nuclide, which is almost absent in nature and is currently available as by-product in e.g. CANDU, whose operation in the next decades (both in terms of life extension of existing reactors and construction of new ones) is at the moment under debate. During DEMO operation, T will be generated on-site by breeding blanket, employing the neutrons originating from D-T reaction. However, it is considered that a certain initial amount of T is needed to start operation, the so-called start-up inventory. An alternative approach consists of obtaining the start-up inventory exploiting reactions occurring in a D -D plasma, which generate T both directly in the plasma and via breeding in the breeding blanket. In the present paper, the conditions under which D -D start-up becomes a favorable option for a power plant are discussed. The analysis mainly focuses on the EU-DEMO reactor concept, for which design data are sufficient for a fairly quantitative evaluation of the relevant parameters. It is found that the unavoidable presence of elements requiring saturation before they are able to release significant amounts of T clamps the T production rate to the same order of magnitude as D -D reaction rate. Thus, under very optimistic assumptions, several hundreds of full-current D -D discharges are necessary for T to be available for plasma fueling, but more realistic estimates let this number raise up to several thousands.
