Aims: Whether oral squamous cell carcinoma (OSCC) from distinct anatomic sites shares malignant architecture or shows site-specific tumor microenvironment (TME) organization remains unclear. Here, we used spatial transcriptomics to define malignant states and TME composition in HPV-negative OSCC. Methodology: Visium data from 27 HPV-negative OSCC samples from 18 patients was analysed: tongue (n=9), alveolar ridge (n=8), floor of mouth (n=5), and buccal mucosa (n=5). Malignant regions were identified by deconvolution, pathology review, and deep learning. Malignant niches were annotated by pathway/gene-set analysis. SpatialInferCNV inferred copy-number (CN) variation, and graph-based analysis defined cellular neighborhoods. Results: Among 61,989 spots, 27,335 malignant spots resolved into edge, fibrovascular, and core niches, with no significant subsite-specific differences in abundance. Tumor-core regions were enriched for keratinization, PTEN signaling, and differentiation (P<0.01). Edge and fibrovascular niches shared EMT, proliferation, and matrix-remodeling programs; the edge was more basal-like, whereas the fibrovascular niche showed greater MHC class I antigen presentation (P<0.01). SpatialInferCNV identified recurrent 3q/8q/11q gains and 3p/8p/10p losses, with no CN burden differences across niches (P=0.26). In contrast, non-malignant compartments varied by anatomic site, with enrichment of interferon-associated TAMs and interferon-inflammatory CAFs (P=0.029 and 0.08) in alveolar ridge, OXPHOS-high myCAFs (P=0.007) in tongue, and reduced TLS-organizing B cells in buccal tumors (P=0.054). Neighborhood analysis linked malignant states to distinct immune/stromal niches. Conclusion: OSCC shows conserved malignant programs and CN architecture across subsites, while immune/stromal compartments vary by anatomic site. These findings suggest that site-specific biology is driven largely by TME organization rather than malignant-state composition.