Worldwide, the number of genebanks and the amount of seed stored in them has increased substantially over the past few decades. Most attention is focused on the likely benefits from conservation, but conserving germplasm involves costs whose nature and magnitude are largely unknown. Because more resources spent on conserving germplasm often means less spent on characterizing the collection or using the saved seeds in crop-improvement research, knowledge of the costs of germplasm conservation has important, possibly long run, R&D management, policy, and food-security consequences. Moreover, these costs place a lower bound on the benefits deemed likely to justify the expense of saving this seed. In this paper we compile and use a set of cost data for wheat and maize stored in the CIMMYT genebank to address a number of questions. What is the cost of storing an accession of either crop for one more year, or, equivalently what is the benefit in terms of cost savings from eliminating duplicate accessions from the genebank? Relatedly, what is the cost from introducing a new accession into the genebank, given the decision to store it is revisited after one year? Does it make economic sense for CIMMYT to discard accessions that may be available elsewhere? As an extension of this line of inquiry it is possible to value the benefits from either consolidating genebanks or at least networking existing banks to reduce or eliminate duplicate holdings not needed for backup safety purposes. We present estimates of the size and scale economies evident in the CIMMYT operation as a basis for assessing the economics of consolidation. Genebanks represent a commitment to conserve seeds for the very long-run. In this study we report on these long-run costs for the CIMMYT genebank¾costs that are sensitive to the interest rate used and the protocols for periodically replenishing accessions that are shared with others or regenerating accessions whose viability gradually diminishes with age. We estimate that under baseline assumptions the present value of conserving the existing accessions in perpetuity at CIMMYT is $7.95 million¾$4.42 million for storing the 17,000 maize accessions and $3.53 million for the 123,000 wheat samples. Maintaining the current level of effort to disseminate accessions free-of-charge to those who request them would cost an additional $3.07 million in perpetuity. Contrary to popular perception, conserving seeds (like R&D more generally) is much more of a labor or human-, not physical-capital intensive, undertaking. On an annualized basis, physical capital represents 22 percent of the costs of conservation, labor about 60 percent, with operational costs making up the remaining 18 percent. Much of the labor takes the form of a quasi-fixed input¾the human capital embodied in senior scientific and technical genebank staff is a lumpy labor input that is not especially sensitive to changes in the size of the holding.