||The availability of a host of software systems for the architects, the engineers and more generally for the construction practitioners has changed the business practice and modified the construction processes. Most of the project data are stored in some electronic form and most of the activities from the brief down to facility management resort to these data. The term Virtual Enterprise (VE) is more and more used to refer to this new complex information technology infrastructure. Exchange and sharing of technical data between construction project's partners, all along the project life cycle, is an important requirement for all participants in a VE. During the last few years an important effort has been undertaken in different research projects to define the software infrastructure of the future VE. While data modeling was mainly based on STEP methodology and norms, software solutions in these works were extremely varied. Studies cover distribution of data and applications, concurrent access as well as security, authentication and data ownership. Among these works we can cite those of NIIIP (and its derived works in SHIIP and SPARS) in US and European ESPRIT projects like VEGA and GENIAL. Nevertheless, the problem of sharing and concurrently accessing data by different partners during a product life cycle still far from being solved.In this paper we present a new approach to build a software infrastructure for the VE. It is based on a platform of persistent-distributed and shared memory called PerDiS. In this approach memory is shared between all applications, even located at different site or running at different times. Coherent caching of data improves performance and availability, ensures that applications have a consistent view of data, and free developers from manually managing object location. Persistence by reachability, based on a distributed garbage collector frees programmers from dealing with explicit data storage. PerDiS offers transactional mechanisms and checkpointing as well as notification when data is dirty. Applications can defines their security policy and data access right based on a task/role model. Distribution granularity can be tuned at application level and/or PerDiS platform level. This allows an easy porting of existing applications where default distribution granularity can be adopted and shift this delicate problem from programming/object level to the applications/VE level for new application development. To port STEP existing applications, we implement the STEP Data Access Interface (SDAI) on the top of PerDiS after extending its main concepts to meet the requirements of distribution, concurrent access and data security. We will show some experiment results and performance measurements.* This work is partially sponsored by PerDiS Esprit Project (22533).