The current cattle genome reference assembly, based on a single Hereford cow fails to capture the whole spectrumof genetic variations within the species. Structural variations (SVs), defined as genomic variations longer than 50nucleotides, can have a potential impact on both complex and Mendelian phenotypic variations. However theyare difficult to detect using only standard approaches of either short or long-read sequence mapping to the currentbovine genome assembly. Thanks to the recent advances in long-read sequencing technologies coupled with thedevelopment of appropriate bioinformatics tools, it’s now possible to construct de novo genome assemblies for alarge number of animals across various cattle breeds. It also offers the opportunity to study a broader range of bothsmall and more complex genome-wide variations. Using these technologies, we have produced a comprehensivecattle pangenome incorporating genetic diversity from 64 high-quality de novo assemblies representing 14 Frenchbovine dairy and beef breeds. We applied a combination of complementary approaches to characterize a wide spectrumof SVs and we report the identification of several megabases of novel genome sequences that are absent in thecurrent cattle genome reference assembly. Further work is currently in progress to investigate the gene content ofthese non-reference sequences. This work was conducted in the SeqOccIn project, funded by the Occitanie region,FEDER, and Apis-Gene. Valentin Sorin’s PhD is supported by INRAE.