Abstract We present results from Atacama Large Millimeter/submillimeter Array (ALMA) 1.2 mm continuum observations of a sample of 27 star-forming galaxies at z = 2.1–2.5 from the MOSFIRE Deep Evolution Field survey with metallicity and star formation rate measurements from optical emission lines. Using stacks of Spitzer, Herschel, and ALMA photometry (rest frame ∼8–400 μm), we examine the infrared (IR) spectral energy distributions (SED) of z ∼ 2.3 subsolar-metallicity (∼0.5 Z _⊙) luminous infrared galaxies (LIRGs). We find that the data agree well with an average template of higher-luminosity local low-metallicity dwarf galaxies (reduced χ ² = 1.8). When compared with the commonly used templates for solar-metallicity local galaxies or high-redshift LIRGs and ultraluminous IR galaxies, even in the most favorable case (with reduced χ ² = 2.8), the templates are rejected at >98% confidence. The broader and hotter IR SED of both the local dwarfs and high-redshift subsolar-metallicity galaxies may result from different grain properties or a harder/more intense ionizing radiation field that increases the dust temperature. The obscured star formation rate (SFR) indicated by the far-IR emission of the subsolar-metallicity galaxies is only ∼60% of the total SFR, considerably lower than that of the local LIRGs with ∼96%–97% obscured fractions. Due to the evolving IR SED shape, the local LIRG templates fit to mid-IR data overestimate the Rayleigh–Jeans tail measurements by a factor of 2–20. These templates underestimate IR luminosities if fit to the observed ALMA fluxes by >0.4 dex. At a given stellar mass or metallicity, dust masses at z ∼ 2.3 are an order of magnitude higher than z ∼ 0. Given the predicted molecular gas fractions, the observed z ∼ 2.3 dust-to-stellar mass ratios suggest lower dust-to-molecular gas masses than in local galaxies with similar metallicities.