The origin of dust in galaxies is still a mystery (1, 2, 3, 4). The majority of the refractory elements are produced in supernova explosions but it is unclear how and where dust grains condense and grow, and how they avoid destruction in the harsh environments of star-forming 3 Figure 3 shows the resulting confidence interval for the two parameters a max and α around the best fit values of a min = 0.001 µm, a max = 4.2 µm and α = 3.6. It is evident that only size distributions extending to grain radii that are significantly larger than that of MW interstellar medium (21, 22) dust ( 0.25 µm) can reproduce the supernova extinction curve (Figure 2). The 2 σ lower limit on the maximum grain size is a max > 0.7 µm. We cannot perform a similar analysis of the late epoch because the intrinsic line profile at this epoch is unknown and likely highly affected by extinction (13). However, we note that the blueshift velocities change only marginally with wavelength (Extended Data Figure 6), suggestive of large grains also at this epoch. Figure 4 illustrates the continuous build-up of dust as a function of time. The increasing attenuation of the lines is accompanied by increasing emission in the near-infrared (NIR) spectra, from a slight excess over a supernova blackbody fit at early times to total dominance at the late epoch. We fitted the spectra with black bodies which for the NIR excess yield a constant blackbody radius of (1.0 ± 0.2) × 10 16 cm at the early epochs, and a temperature that declines from ∼ 2,300 K to ∼ 1,600 K from day 26 onwards. At the late epoch, we obtain a black-body radius of (5.7 ± 0.2) × 10 16 cm and a temperature of ∼ 1,100 K. The high temperatures detected at the early epochs suggest that the NIR excess is due to thermal emission from carbonaceous dust, rather than silicate dust, which has a lower condensation temperature of ∼ 1, 500 K (1). The high temperatures rule out suggestions that the NIR emission is due to pre-existing dust or a dust echo (11) (Figure 1), the accelerated dust formation occurring at later times ( Figure 4) and at larger radius is possibly facilitated by the bulk ejecta material, which travels on average at a velocity of ∼ 7, 500 km s −1 at early epochs (Extended Data Figure 4).Our detection of large grains soon after the supernova explosion suggests a remarkably rapid and efficient mechanism for dust nucleation and growth. The underlying physics is poorly understood but may involve a two-stage process governed by early dust formation in a cool, dense shell, 5 followed by accelerated dust formation involving ejecta material. For Type IIP supernovae, the growth of dust grains can be sustained up to 5 years past explosion (25). The dense CSM around Type IIn supernovae may provide conditions to facilitate dust growth beyond that. The process appears generic, in that other Type IIn supernovae like SN 1995N, SN 1998S, SN 2005ip, and SN 2006jd exhibited similar observed NIR properties (8, 10, 26, 27) and growing dust masses, consistent with the trend revealed here for SN 2010...