The usage of economics in grid systems designs has been very popular recently and a large number of system designers have attempted to use economic concepts to engineer the grid market mechanism that would enable an efficient allocation of resources. However, only few of these designs have been actually implemented and, in fact, very few, if any, have been successfully adopted by end users. The argument developed in this article is that the reason behind so many different systems being developed and their relative lack of success is the generally adopted 'top-down' approach to economics-based grid design. In contrast, this article argues that that grid market environments should be built bottom-up instead. In this new paradigm, the role of a grid middleware ceases to be prescriptive, in the sense that it carries a (or a set of) particular predefined market mechanism, and becomes a foundation on which the individual choices and interactions of users build an evolving, self-organised, self-adaptive and allocatively efficient market. A multi-agent/peer-to-peer based generic middleware is propose to support the paradigm.