The following post is based on my contribution to the Dagstul Seminar "Is the Future of Preservation Cloudy?" in November 2012:
A growing number of archival institutions are turning towards POSE (Pay Once, Store Eternally) or Endowment models for funding their long-term digital archiving and preservation activities. The endowment model has a number of seductive advantages. First, it fits in nicely with a project-oriented digitisation efforts, as the endowment costs can be included in a project budget and do not have to be added to annual running budgets. Endowment models also allow simple budget calculations based on total storage volume, which in turn support business models based on archival services. As archival institutions face pressure to become self-sustaining, such business models are in great demand.
However, there may be a number of dangerous assumptions behind simple endowment models. There is a pervading view that data centres with endowment models are like pension plans, in which incoming endowments (workers) will pay for old data (retirees). This analogy is clearly false because, unlike unfortunate pensioners, old data never dies. The reply to this is usually, "but the old data is so much smaller than the new data." This is true, but the only way in which an endowment model can handle the ever-increasing volumes of new data is by basing the cost model on a careful analysis of storage costs, such as detailed in Rosenthal et al.. Too many endowment models simply assume that per volume storage costs will continue to decrease (Kryder's law) forever, which simply cannot be the case. Rather, the storage capacity per unit cost, which is presently in an exponential growth phase, will eventually reach a stationary phase and level off, just as every other exponential growth scenario in nature. It is incumbant upon an endowment model to at least attempt to predict at what rate storage capacity will continue to grow and when the stationary phase will occur.
Commercial storage providers such as Google and Amazon are well-aware of these difficulties, and offer business models that are highly advantageous to themselves as a result (the decreases in service costs offered over the past five years are still much higher than the real decrease in storage costs). The question is, are libraries, archives, and other data centres equally aware? Endowment models are complex, and probably more expensive than we think. The inevitable conclusion is that we can no longer afford to archive everything.