The scaling down of meta-atoms or metamolecules (collectively denoted as metaunits) is a long-lasting issue from the time when the concept of metamaterials was first suggested . According to the effective medium theory, which is the foundational concept of metamaterials, the structural sizes of meta-units should be much smaller than the working wavelengths (e.g., << 1/5 wavelength). At relatively low frequency regimes (e.g., microwave and terahertz), the conventional monolithic lithography can readily address the materialization of metamaterials . However, it is still challenging to fabricate optical metamaterials (metamaterials working at optical frequencies such as the visible and near-infrared regimes) through the lithographic approaches . This serves as the rationale for using colloidal self-assembly as a strategy for the realization of optical metamaterials . Colloidal self-assembly can address various critical issues associated with the materialization of optical metamaterials, such as achieving nanogaps over a large area, increasing true 3D structural complexities, and cost-effective processing, which all are difficult to attain through monolithic lithography . Nevertheless, colloidal self-assembly is still a toolset underutilized by optical engineers . Here, the design principle of the colloidally self-assembled optical metamaterials exhibiting unnatural refractions, the practical challenge of relevant experiments, and the future opportunities are critically reviewed.