Abstract : D0 is a pioneer in grid computing for large scale production activities involving the handling of collider data samples. A data grid (SAM) has been used since the start of Tevatron Run II as the sole means of data transport (enabling local offsite analysis). The focus of the computational grid (SAM-grid) so far has been on production activities. Integration of SAM-Grid with other grids, like LCG and OSG are ongoing projects. All Monte Carlo data are produced off-site. In 2005 and 2007 large fractions of the Run IIa and Run IIb data sets respectively (1 billion events) were reprocessed using native SAM-Grid, LCG and OSG resources. The value of grid computing to the D0 experiment is conservatively estimated at roughly $4M/year. Evolution towards full grid computing and lessons learned from these activities will be discussed.