Genomic selection - the prediction of breeding values using DNA polymorphisms - is a disruptive method that has widely been adopted by plant breeders to increase the genetic progress. It was recently shown that other sources of molecular variations such as those resulting from transcriptomics or metabolomics could be used to make accurate predictions of complex traits. These endophenotypic variations have the advantage of capturing the expressed genotypes and as a consequence the complex regulatory networks that can occur in the different layers between the genome and the phenotype. However, obtaining omics data for genomic or endophenotypic predictions at large scales, such as those typically experienced in plant breeding, remains difficult because of the costs of genotyping or endophenotypic characterization. As an alternative, we proposed to use near-infrared spectroscopy (NIRS) as a high-throughput, low cost and non-destructive tool to indirectly capture endophenotypic variants and compute relationship matrices for predicting complex traits. We coined this new approach "phenomic selection" (PS). We tested PS on wheat (Triticum aestivum L.) using NIRS on various tissues (grains, flour, leaves) on panels with various diversity levels. We showed that one could reach predictions as accurate as with molecular markers, for developmental and productivity traits, even in environments radically different from the one in which NIRS were collected. Our work constitutes a proof of concept and provides new perspectives for the breeding community, as PS is theoretically applicable to any organism at low cost and does not require any molecular information.