Numerous studies have been conducted to investigate the genetic basis of Fusarium wilt (FW) resistance using bi-parental and association mapping populations in cotton. In this study, a multi-parent advanced generation inter-cross (MAGIC) population of 550 recombinant inbred lines (RILs), together with their 11 Upland cotton (Gossypium hirsutum) parents, was used to identify QTLs for FW race 4 (FOV4) resistance. Among the parents, Acala Ultima, M240, and Stoneville 474 were the most resistant, while Deltapine Acala 90, Coker 315, and Stoneville 825 were the most susceptible. Twenty-two MAGIC lines were consistently resistant to FOV4. Through a genome-wide association study (GWAS) based on 473,516 polymorphic SNPs, a major FOV4 resistance QTL within a narrow region on chromosomes D02 was detected, allowing identification of 14 candidate genes. Additionally, a meta-analysis of 101 published FW resistance QTLs showed a D subgenome and individual chromosome bias and no correlation between homeologous chromosome pairs. This study represents the first GWAS study using a largest genetic population and the most comprehensive meta-analysis for FW resistance in cotton. The results illustrated that 550 lines are not enough for high resolution mapping to pinpoint a candidate gene, and experimental errors in phenotyping cotton for FW resistance further compromise the accuracy and precision in QTL localization and identification of candidate genes. This study identified FOV4 resistant parents, MAGIC lines, and the first major QTL for FOV4 resistance, providing useful information for developing FOV4 resistant cultivars and further genomic studies towards identification of causal genes for FOV4 resistance in cotton.