Bladder cancer is a common disease characterized by multiple recurrences and an invasive disease course in more than 10% of patients. It is of monoclonal or oligoclonal origin and genomic instability has been shown at certain loci. We used a 10,000 single nucleotide polymorphism (SNP) array with an average of 2,700 heterozygous SNPs to detect allelic imbalances (AI) in 37 microdissected bladder tumors from 17 patients. Eight tumors represented upstaging from Ta to T1, eight from T1 to T2+, and one from Ta to T2+. The AI was strongly stage-dependent as four chromosomal arms showed AI in > 50% of Ta samples, eight in T1, and twenty-two in T2+ samples. The tumors showed stage-dependent clonality as 61.3% of AIs were reconfirmed in later T1 tumors and 84.4% in muscle-invasive tumors. Novel unstable chromosomal areas were identified at chromosomes 6q, 10p, 16q, 20p, 20q, and 22q. The tumors separated into two distinct groups, highly stable tumors (all Ta tumors) and unstable tumors (2/3 muscle-invasive). All 11 unstable tumors had lost chromosome 17p areas and 90% chromosome 8 areas affecting Netrin-1/UNC5D/MAP2K4 genes as well as others. AI was present at the TP53 locus in 10 out of 11 unstable tumors, whereas 6 had homozygous TP53 mutations. Tumor distribution pattern reflected AI as seven out of eight patients with additional upper urinary tract tumors had genomic stable bladder tumors (P < 0.05). These data show the power of high-resolution SNP arrays for defining clinically relevant AIs.