Axially loaded piles should be designed to provide adequate strength against a bearing capacity failure. In many engineering problems, pile groups are loaded eccentrically, that is, the external load is not applied on the center of the axial capacities of the individual piles, as shown in Fig. 1. Under centered load, the difference between the axial capacity of a group of piles and that of the single pile, Nu, multiplied by the number of piles, p, has been traditionally characterized by an efficiency factor, g, depending on both pile and soil type : Qu ¼ g p Nu ð۱Þ The proper value of the efficiency may be selected on the basis of the available experimental evidence [2, 3, 6, 7, 23, 32, 34, 35]. The above concept applies to the failure mode corresponding to individual pile capacities. However, collapse of pile groups may occur also by failure of the overall block of soil containing the piles [11, 31]. Fleming et al.  claim that block failure for granular soils occurs when the base area is much smaller than the side area (Ab/As1). As a result, group of closely spaced long piles are more likely to fail as a block than groups of short piles at relatively large spacing. For fine-grained soils in undrained conditions, block failure is even more likely to occur. Experiments carried out by many researchers (e.g., [6, 7]) suggested that groups with piles at spacing smaller than a critical value (say 2–۴ times the pile diameter d) fail as ‘blocks’. Independent calculation of both modes of failure should be therefore carried out, and the bearing capacity of the pile group taken as the lesser of the two capacities. In case of eccentric loads (Fig. 1), the most widespread approach is that of considering the achievement of the axial capacity (in compression or in uplift) on the outermost pile as the ultimate limit state of the pile group, as recommended, for example, by AASHTO Bridge Design Specifications . The implication in design of such an assumption is described in Fig. 2, referring to a 1 9 4 group of identical piles connected by a rigid cap and loaded by a vertical force acting along the axis of the first pile. For the sake of simplicity, the four piles are considered as linear elastic, perfectly plastic independent springs, with equal strength in compression (Nu) and in uplift (-Su). Under these hypotheses, the load distribution varies linearly with the distance along the cap until the achievement of the axial capacity on the first pile. According to the common approach, the axial capacity of the pile group would be Qu = 10/7Nu. However, the achievement of the axial capacity on pile 1 does not represent a failure condition for the whole group and can be viewed just as the onset of yielding. At this point, the pile group is still capable to carry a further increase in the external load taking advantage from the ductility of the system. For example, an external load Qu = 2Nu (that is 40% larger than 10/7Nu) might be equilibrated by a load distribution where piles from 1 to 3 achieve the axial capacity in compression and pile 4 that in uplift. Such a load distribution does not violate the failure criterion adopted for the piles and, thus, is a lower bound solution of the problem under examination. Therefore, the common approach is unduly conservative.