BLAST is one of the most widely used tools in bio informatics to find biologically similar sequences from the genome databases for a given query sequence. Although elaborate dynamic programming and heuristic search in the BLAST rapidly identify significant similar sequences, performance is still an important issue as the size of genome databases and sequences to be compared increases. We propose HGBS (hardware-oriented grid BLAST system) that significantly improves the BLAST performance by applying the grid computing technique. HGBS utilizes a PC-based grid environment constructed by a specialized hardware chip, not a grid middleware such as the Globus toolkit, to provide enhanced performance, good extensibility, and lower implementation cost. Performance analysis shows that there is a significant performance advantage in HGBS; for huge number of BLAST operations in 89-prokaryote genomes comparison, HGBS connecting 27 PCs needs just a week to complete the comparisons, but a single server system takes about 33 weeks.