Large amounts of dust (> 10 8 M ⊙ ) have recently been discovered in high redshifts quasars 1,2 and galaxies 3−5 , corresponding to a time when the Universe was less than one-tenth of its present age. The stellar winds produced by stars in the late stages of their evolution (on the asymptotic giant branch of the Hertzsprung-Russell diagram) are though to be the main source of dust in galaxies, but they cannot produce that dust on a short-enough timescale 6 (< 1 Gyr) to explain the results in the high-redshift galaxies.Supernova explosions of massive stars (type II) are also a potential source, with models predicting 0.2-4 M ⊙ of dust 7−10 . As massive stars evolve rapidly, on timescales of a few Myr, these supernovae could be responsible for the high-redshift dust. Observations 11−13 of supernova remnants in the Milky Way, however, have hitherto revealed only 10 −7 − 10 −3 M ⊙ of dust each, which is insufficient to explain the high-redshift data. Here we report the detection of ∼ 2 − 4 M ⊙ of cold dust in the youngest known Galactic remnant, Cassiopeia A. This observation implies that supernovae are at least as important as stellar winds in producing dust in our Galaxy and would have been the dominant source of dust at high redshifts.Over the past three decades, many searches for dust in supernova remnants (SNR) have been made in the mid and far-infrared (6-100µm). Remnants must be studied when they are young, before they have swept up large masses of interstellar material which makes it difficult to distinguish dust formed in the ejecta from that present in the ISM prior to the explosion. The handful of Galactic remnants which are both young and close enough (Cas A, Kepler and Tycho) have been studied with the Infrared Astronomical Satellite (IRAS) and the Infrared Space Observatory (ISO), but although dust at 100-200 K has been detected, the dust mass deduced is only 10 −7 − 10 −3 M ⊙ , many orders of magnitude lower than the solar mass quantities predicted. 11−13 The formation of dust in the recent supernova 1987A has been implied indirectly from the fading of the silicate line and increase in 10µm emission, 14 although the quantity is heavily dependant on the assumptions made 1